Method and system for monitoring a winding process, method for manufacturing a transformer winding, and method and system for manufacturing a transformer

Abstract

A computer vision system (40) processes at least one image (42) captured during a winding process showing a transformer winding (20) to monitor the winding process. The computer vision system (40) detects a difference between a result of an object classification and design data (49) of the transformer winding (20).

Description

Technical Field

[0001] Embodiments of the present invention relate to the manufacture of windings for power system components. In particular, embodiments of the present invention relate to a quality control method and system operable to detect errors during a winding process performed on a winding device. Embodiments of the present invention also relate to manufacturing systems and methods including the quality control method or system according to embodiments. Background Technology

[0002] Transformers are essential components of power systems. Transformers can have various configurations depending on their intended use. Manufacturing transformers designed for power generation, transmission, and / or distribution is a highly complex task due to the specific requirements imposed on them. This is particularly true for the manufacture of transformer windings.

[0003] The transformer windings are designed to optimize their performance for a given amount of copper used. The design is also optimized to occupy a small footprint while maintaining high reliability, efficient cooling, and insulation. This complicates the manufacturing process.

[0004] Fortunately, errors during the winding process of forming transformer windings are rare, but any fault can have serious negative impacts on power supply, operational safety, and worker safety.

[0005] Similar challenges exist in the winding process of forming the windings of other power system components, such as reactors.

[0006] Therefore, quality control of transformer windings or other power system components is an important issue in reducing the risk of failures with catastrophic consequences.

[0007] Unfortunately, such quality control is not easily performed using conventional techniques. Given the complexity of the winding process, errors are difficult to detect. Many errors, such as those related to the position of crossovers or deviations from design data, are hidden and invisible within the transformer windings. The challenges in error detection are particularly pronounced when the winding equipment provides both rotational and translational degrees of freedom, resulting in the transformer windings being both rotated around their axis and translated during manufacturing.

[0008] In light of the above, a method and system are needed to mitigate the risks associated with the use of power system component windings that are not formed to their specifications in transformers, transformer cores, or other power system components. In particular, a method and system are needed that allows for the detection of quality problems during the winding process of forming the windings. Summary of the Invention

[0009] According to various aspects of the invention, methods and systems as described in the independent claims are provided. Dependent claims define preferred embodiments.

[0010] According to one aspect of the invention, a method is provided for monitoring the winding process of a winding (such as a transformer winding) forming a power system component on a winding device. The method includes: processing by a computer vision system at least one image captured during the winding process, showing at least a portion of the winding (e.g., a transformer winding), wherein processing the at least one image includes performing object classification. The method further includes: detecting discrepancies between the results of the object classification detected by the computer vision system and design data of the winding (e.g., a transformer winding).

[0011] Methods for monitoring the winding process allow for the detection of errors while the winding (e.g., transformer winding) is still forming. This enables mitigation and / or corrective actions to be taken during the manufacturing of the transformer winding.

[0012] Methods for monitoring the winding process utilize computer vision systems in conjunction with design data to detect discrepancies. These discrepancies are indicators that errors have occurred or may have occurred. Therefore, computer vision systems provide assistance during the winding process of forming transformer windings.

[0013] The method may include: a computer vision system performing at least one action in response to a detected difference.

[0014] Therefore, when a difference is detected by the computer vision system, one or more actions can be performed automatically or semi-automatically.

[0015] At least one action may include: the computer vision system causing the human-machine interface (HMI) to output information based on the detected differences.

[0016] Therefore, the detected difference can be used as a trigger for an output action that can warn the operator of the winding equipment that an error has occurred or may have occurred during the winding process.

[0017] This information may include one, several, or all of the following: alarms and / or alerts; information indicating the location of the detected difference; information indicating the time when the detected difference occurred; information about the root cause of the detected difference; and instructions for correcting the detected difference.

[0018] Therefore, the detected difference can be used as a trigger for an output action that can alert the operator of the winding equipment and optionally provide additional information that helps the operator identify the root cause of the error or assists the operator in taking appropriate or corrective mitigation actions to correct the error in the winding process.

[0019] At least one action may include a relief action and / or a corrective action.

[0020] Therefore, the detected differences can be used as triggers for actions that can be performed automatically or semi-automatically to mitigate and / or correct errors already identified by the computer vision system.

[0021] This action may include stopping the rotation and / or translation of the winding equipment.

[0022] This facilitates visual inspection or correction of the winding process.

[0023] The winding apparatus may include a winding support on which windings (e.g., transformer windings) are formed. The winding apparatus may include one or more motors that drive the winding support. The winding apparatus may include: a first motor that rotates the winding support; and a second motor that translates the winding support, preferably along the axis of rotation of the winding support.

[0024] This facilitates the manufacture of windings (e.g., transformer windings). Additionally, this configuration of the winding equipment is well-suited for monitoring methods using computer vision systems. Image acquisition devices (such as cameras) can be fixed in place because the winding equipment is shifted during the formation of the transformer windings so that its top remains at approximately the same height.

[0025] The method may further include: determining at least one image acquisition time for at least one image during the winding process.

[0026] Thus, a computer vision system can determine at which image acquisition times during the winding process at which at least one image provides particularly good information to facilitate error detection. For illustration, one or more image acquisition times may depend on how the shielding covering the conductors and / or filler of the winding (e.g., transformer windings) is positioned relative to the image acquisition device.

[0027] The method may further include: performing image acquisition device control and / or image selection based on at least one determined image acquisition time.

[0028] Therefore, image acquisition can be controlled in such a way that at least one image provides particularly good information to facilitate error detection. Alternatively or additionally, images particularly suitable for detecting errors occurring during the winding process can be selected from the image stream provided by the image acquisition device (e.g., because the shielding covering the conductor or filler is properly positioned relative to the image acquisition device).

[0029] At least one image acquisition time can be determined based on at least one position of the winding device. At least one image acquisition time can be determined based on the rotational and / or translational position of the winding device. Design data can be used in conjunction with multiple positions of the winding device to determine at least one image acquisition time. For illustration, design data can be used to determine which rotational and / or translational positions of the winding device's shield, conductors, filler, and / or crossovers and / or other winding components or structural features are appropriately positioned within the field of view of the image acquisition device.

[0030] At least one image acquisition time can be determined based on an image stream acquired by at least one image acquisition device. The computer vision system can use image processing techniques (such as edge detection techniques, Fourier-based techniques, and / or contrast enhancement techniques) to identify the image in the stream that is most likely to provide information about whether an error occurred during the winding process.

[0031] A computer vision system may use a first machine learning (ML) model to determine at least one image acquisition time. The first ML model may have a first input layer that receives the image pixels from the stream. The first ML model may have a first output layer that outputs values ​​indicating whether further analysis should be performed on any of the pixels in the image. Alternatively or additionally, the first input layer may receive the location of a winding device, such as the rotational and / or translational position of a winding support on which transformer windings are formed.

[0032] Processing at least one image may include identifying regions of interest (ROIs) in which object classification is to be performed.

[0033] Therefore, object classification can focus on areas where differences (if any) are likely to be visible.

[0034] The boundaries of an ROI may include the edges of the shielding of a transformer winding.

[0035] Therefore, object classification can be focused on the area between adjacent shielding components inserted between continuous discs of transformer windings.

[0036] Alternatively or additionally, the boundaries of the ROI may include the inner and outer circumferential edges of the layer formed by the conductor and filler.

[0037] Therefore, object classification can focus on areas where winding errors are more likely to occur.

[0038] Identifying ROIs may include performing edge detection.

[0039] This allows for efficient identification of ROIs.

[0040] Identifying a Region of Interest (ROI) may involve applying at least one ROI classifier to at least one image. Several ROI classifiers may be applied to at least one image. Each of the one or more ROI classifiers may indicate, for each pixel, whether the corresponding pixel is classified as inside or outside the ROI.

[0041] One or more ROI classifiers may include at least one second ML model. The at least one second ML model may have a second input layer that receives at least one pixel of an image. The at least one second ML model may have a second output layer that outputs a value for each pixel indicating whether the corresponding pixel is in the ROI.

[0042] At least one second ML model may include a deep learning ML model.

[0043] Therefore, object classification tasks can be performed reliably.

[0044] At least one second ML model may include a U-Net ML model.

[0045] Therefore, ROI identification tasks can be performed reliably.

[0046] Object classification may include executing at least one third ML model. The at least one third ML model may have a third input layer that receives pixels located within an ROI. The at least one third ML model may have a third output layer that outputs an object classifier indicating which of several object categories a corresponding pixel in the ROI is assigned to. The third output layer may indicate the probability that an object belongs to a given object category (such as filler, conductor, shield, and / or intersection).

[0047] The third ML model may include, but is not limited to, random forest (RF) models.

[0048] Therefore, object classification tasks can be performed in a way that provides particularly good results during inference.

[0049] Object classification can classify at least conductors, fillers, and shielding components in at least one image.

[0050] Alternatively or additionally, object classification can be applied to intersections. Object classification can be applied to internal intersections and external intersections.

[0051] Alternatively or additionally, object classification may classify other components and / or other structural features of the transformer windings.

[0052] This resulted in object categories suitable for comparison with design data.

[0053] Several classifiers can be used, each associated with a given object category (such as filler, conductor, shield, and / or cross) and trained to identify that given object category.

[0054] Object classification can define a sequence of objects, including at least conductors and fillers arranged in a radial direction (e.g., from inside to outside or from outside to inside), as identified in the ROI of an image.

[0055] Processing at least one image may include one or more position measurements using a winding device. The one or more position measurements may include rotational and / or translational position measurements. One or more position measurements may be received from a sensor. Alternatively, one or more position measurements may be received from the controller of the winding device.

[0056] One or more location measurements can be used to determine which images require further analysis, such as by performing ROI determination and object classification. Alternatively or additionally, one or more location measurements can be used to assist in ROI determination.

[0057] Processing at least one image may include executing at least one ML model having an input layer that receives at least a portion of the at least one image and / or the location of the winding device, wherein the at least one ML model has an output layer that provides at least one of the following: defining at least one image acquisition time for the at least one image to be processed; an ROI in which object classification is to be performed; and / or an object classifier indicating which of several predetermined object categories an object is assigned to.

[0058] The method may further include: determining a target sequence of objects along the radial direction of the transformer winding based on design data.

[0059] Thus, the design data is converted into a format that can be easily compared with the results of object classification.

[0060] Design data may include at least one cross graph.

[0061] Design data can define the winding as a three-dimensional volumetric component. Design data can be in various formats, such as one or more cross diagrams, but is not limited to these.

[0062] Determining the target sequence of objects may include generating a machine-readable representation indicating the target sequence based on at least one cross graph. The machine-readable representation may include, but is not limited to, an XML representation.

[0063] Detecting differences can include comparing the results of target sequence and object classification.

[0064] The target sequence can be automatically determined by a computer vision system or a processing system that is communicatively connected to it.

[0065] A winding (e.g., a transformer winding) may comprise several layers (also referred to in the art as a disc). When the winding is formed on a winding device, at least one image may show at least one of the layers on the upper surface of the winding. The image of the uppermost layer on the upper surface of the winding may be processed in the method to detect differences.

[0066] During the winding process, at least one image can be processed repeatedly on a continuous basis (i.e., continuously) and the results of object classification can be compared with the design data.

[0067] When this method uses one or more ML models (such as one, several, or all of the first, second, and third ML models mentioned above), labeled image data can be used to train the ML models. Labeled image data can include synthetically generated images derived from real-world images that demonstrate errors during the winding process. The generation of synthetically generated images can include one or more image modification operations, such as rotation, translation, mirroring, adding distortion, etc.

[0068] One or more ML models can be trained using supervised or semi-supervised training.

[0069] The method may further include: training an ML model using labeled image data.

[0070] Training can be performed by a separate computer system that is distinct from the computer vision system, or it can be performed by the computer vision system itself.

[0071] Training an ML model may involve synthetically generating images from real-world images that show errors during the winding process. The generation of synthetically generated images may include one or more image modification operations, such as rotation, translation, mirroring, adding distortion, etc.

[0072] Synthetically generated images allow for a more balanced training set, thereby improving results when ML models are used for inference during subsequent field use in computer vision systems.

[0073] A trainable ML model can perform at least one classification task. The at least one classification task can be or may include one or more of the following: classifying an image as an image to be processed; classifying pixels as belonging to a Region of Interest (ROI); performing object classification.

[0074] ML models may include at least one of the following: deep learning models, such as the U-Net model; random forest model; recurrent neural network (RNN); convolutional neural network (CNN).

[0075] Synthetically generating images may include using adversarial ML models to generate synthetic images.

[0076] Transformer windings can be transformer windings from the following sources: power generation and transmission systems (such as high-voltage and medium-voltage transmission systems), distribution systems, railway transformers, or reactor or inductor systems, or shunt reactors (such as variable reactors or other shunt reactors). According to another aspect of the invention, a method is provided for training at least one ML model for monitoring the winding process of windings (such as transformer windings) forming power system components on winding equipment.

[0077] The method may include: using image data showing the markings of the windings when they are formed (such as transformer windings) to train one or more ML models (such as one, several or all of the first, second and third ML models mentioned above).

[0078] Training methods may include: synthetically generating at least some training images from real-world images that demonstrate errors during the winding process. Generating the synthetically generated images may include one or more image modification operations, such as rotation, translation, mirroring, adding distortion, etc.

[0079] One or more ML models can be trained using supervised or semi-supervised training.

[0080] Training can be performed by a separate computer system that is distinct from the computer vision system, or it can be performed by the computer vision system itself.

[0081] Synthetically generated images allow for a more balanced training set, thereby improving results when ML models are used for inference during subsequent field use in computer vision systems.

[0082] Training methods may include training at least one ML model to perform at least one classification task. The at least one classification task may be, or may include, one or more of the following: classifying an image as an image to be processed; classifying pixels as belonging to a Region of Interest (ROI); performing object classification.

[0083] At least one ML model may include at least one of the following: deep learning models, such as the Unet model; random forest models; recurrent neural networks (RNNs); and convolutional neural networks (CNNs).

[0084] Synthetically generating images may include using adversarial ML models to generate synthetic images.

[0085] The training method can be performed by a computer vision system that executes (multiple) trained ML models during inference, or by a computing system separate from the computer vision system.

[0086] Training methods may include deploying (multiple) trained ML models to a computer vision system for monitoring the winding process of forming transformer windings on a winding device.

[0087] According to another aspect of the invention, a method for manufacturing windings (such as transformer windings) for power system components is provided. The method includes: controlling a winding device on which windings (e.g., transformer windings) are formed by at least one control device. The method also includes performing a method for monitoring the winding process according to any of the embodiments while the transformer windings are being formed.

[0088] The method may further include: correcting at least one error that occurred during the winding process in response to a detected difference.

[0089] The method may further include: discarding transformer windings in response to detected differences.

[0090] This method can be executed automatically by a system comprising a winding device and a computer vision system operable to perform the monitoring methods discussed in detail herein. The computer vision system can be operated if it is coupled to the controller of the winding device.

[0091] According to another aspect of the present invention, a method for manufacturing a transformer core is provided. The method includes assembling a transformer core comprising at least one transformer winding manufactured using a method for manufacturing transformer windings according to an embodiment.

[0092] Assembling a transformer core may include assembling at least one transformer winding with a yoke.

[0093] The yoke may include a laminate comprising several layers. This allows for the efficient transfer of magnetic flux within the transformer core.

[0094] The transformer core can be one of the following: power transmission systems (such as high-voltage and medium-voltage power transmission systems), power distribution systems, railway transformers, or reactor or inductor systems.

[0095] According to another aspect of the present invention, a method for manufacturing a transformer is provided. The method includes assembling a transformer comprising at least one transformer winding manufactured using the method for manufacturing transformer windings according to an embodiment.

[0096] Assembling a transformer may include assembling at least one transformer winding with a yoke.

[0097] The yoke may include a laminate comprising several layers. This allows for the efficient transfer of magnetic flux within the transformer core.

[0098] Assembling a transformer may include arranging at least one transformer winding in a transformer box.

[0099] Assembling a transformer may involve filling the interior of the transformer tank with an insulating fluid, such as insulating oil.

[0100] Transformers can be the following types of transformers: power transmission systems (such as high-voltage and medium-voltage power transmission systems), power distribution systems, or railway systems, or reactors (such as variable reactors or shunt reactors).

[0101] According to another aspect of the invention, a system is provided configured to monitor the winding process of windings (such as transformer windings) forming components of a power system. The system includes a computer vision system configured to: process at least one image captured during the winding process, showing at least a portion of the winding (e.g., a transformer winding), to perform object classification; detect a difference between the result of the object classification and design data of the winding; and perform an action in response to the detected difference.

[0102] Systems used to monitor the winding process allow for the detection of errors while the windings of power system components are still forming. This enables mitigation and / or corrective actions to be taken during the manufacturing of transformer windings.

[0103] Systems used to monitor the winding process utilize computer vision systems in conjunction with design data to detect discrepancies. Discrepancies are indicators that errors have occurred or may have occurred. Therefore, computer vision systems assist in the winding process of forming the windings of power system components (e.g., transformer windings).

[0104] The system may further include a human-machine interface (HMI). The computer vision system is configured to control the HMI in response to detected differences.

[0105] The system can be configured such that the information may include one, several, or all of the following: alarms and / or alerts; information indicating the location of the detected difference; information indicating the time when the detected difference occurred; information about the root cause of the detected difference; and instructions for correcting the detected difference.

[0106] The system can be configured such that at least one action may include a mitigation action and / or a corrective action. The system can be configured to generate and output control signals or control commands for the winding equipment or HMI.

[0107] The system can be configured such that actions may include stopping the rotation and / or translation of the winding device.

[0108] The system may further include a winding device. The winding device may include a winding support on which windings (e.g., transformer windings) are formed. The winding device may include one or more motors that drive the winding support. The winding device may include: a first motor that rotates the winding support; and a second motor that translates the winding support, preferably along the axis of rotation of the winding support.

[0109] The system can be configured to determine at least one image acquisition time for at least one image during the winding process.

[0110] The system can be configured to perform image acquisition device control and / or image selection based on at least one determined image acquisition time.

[0111] The system can be configured to determine at least one image acquisition time based on at least one position of the winding device. At least one image acquisition time can be determined based on the rotational and / or translational position of the winding device. Design data can be used in conjunction with multiple positions of the winding device to determine at least one image acquisition time. For illustration, design data can be used to evaluate which rotational and / or translational positions of the winding device's shielding, conductors, fillers, and / or crossovers and / or other components or structural features are appropriately positioned within the field of view of the image acquisition device.

[0112] The system can be configured to determine at least one image acquisition time based on an image stream acquired by at least one image acquisition device. The computer vision system can use image processing techniques (such as edge detection techniques, Fourier-based techniques, and / or contrast enhancement techniques) to identify the images in the stream that are most likely to provide information about whether an error occurred during the winding process.

[0113] The system can be configured such that a computer vision system can use a first machine learning (ML) model to determine at least one image acquisition time. The first ML model may have a first input layer that receives the image pixels of the stream. The first ML model may have a first output layer that outputs values ​​indicating whether further analysis should be performed on any of the images and, if so, on which of the images. Alternatively or additionally, the first input layer may receive the location of the winding device, such as the rotational and / or translational position of the winding support on which transformer windings are formed.

[0114] The system can be configured to identify regions of interest (ROIs) in which object classification is to be performed.

[0115] The system can be configured such that the boundary of the ROI may include the edge of the shield of the transformer winding.

[0116] Alternatively or additionally, the system may be configured such that the boundaries of the ROI may include the inner and outer circumferential edges of a layer formed of conductors and fillers.

[0117] The system can be configured to perform edge detection to detect ROI.

[0118] The system can be configured to apply at least one ROI classifier to at least one image. Several ROI classifiers can be applied to at least one image. Each of the one or more ROI classifiers can indicate for each pixel whether the corresponding pixel is classified as inside or outside the ROI.

[0119] One or more ROI classifiers may include at least one second ML model. The at least one second ML model may have a second input layer that receives at least one pixel of an image. The at least one second ML model may have a second output layer that outputs a value for each pixel indicating whether the corresponding pixel is in the ROI.

[0120] At least one second ML model may include a deep learning ML model.

[0121] At least one second ML model may include a U-Net ML model.

[0122] The system can be configured such that object classification may include executing at least one third ML model. The at least one third ML model may have a third input layer that receives pixels located within an ROI. The at least one third ML model may have a third output layer that outputs an object classifier indicating which of several object categories a corresponding pixel in the ROI is assigned to.

[0123] The third ML model may include, but is not limited to, random forest (RF) models.

[0124] The system can be configured such that object classification can classify at least conductors, fillers, and shielding elements in at least one image; and / or other components and / or structural features of the winding.

[0125] Alternatively or additionally, the system can be configured such that object classification can classify intersections. Object classification can classify internal intersections and external intersections.

[0126] Object classification can define a sequence of objects, including at least conductors and fillers arranged in a radial direction (e.g., from inside to outside or from outside to inside), as identified in the ROI of an image.

[0127] The system can be configured such that processing at least one image may include one or more position measurements using a winding device. The one or more position measurements may include rotational and / or translational position measurements. One or more position measurements may be received from a sensor. Alternatively, one or more position measurements may be received from a controller of the winding device.

[0128] The system can be configured to use one or more location measurements to determine which images require further analysis, such as by performing ROI determination and object classification. Alternatively or additionally, the system can be configured to use one or more location measurements to assist in ROI determination.

[0129] The system can be configured such that processing at least one image may include executing at least one ML model having an input layer that receives at least a portion of the at least one image and / or the location of the winding device, wherein the at least one ML model has an output layer that provides at least one of the following: defining at least one image acquisition time for the at least one image to be processed; an ROI in which object classification is to be performed; and / or defining an object classifier of one of several predetermined object categories.

[0130] The system can be configured to determine a target sequence of objects along the radial direction of the transformer winding based on design data.

[0131] Design data may include at least one cross graph.

[0132] Design data can define windings (e.g., transformer windings) as three-dimensional volume components. Design data can have various formats, such as one or more cross diagrams, but is not limited to these.

[0133] The system can be configured such that determining the target sequence of objects may include generating a machine-readable representation indicating the target sequence based on at least one cross graph. The machine-readable representation may include, but is not limited to, an XML representation.

[0134] The system can be configured such that detecting differences can include comparing the results of target sequence and object classification.

[0135] A computer vision system can be configured to determine a sequence of targets. Alternatively, the system may further include a processing system communicatively coupled to the computer vision system and configured to determine the target sequence and provide it to the computer vision system.

[0136] This system can be configured to monitor the winding process of windings used in power systems, such as transformer windings, which comprise several layers (also referred to in the art as discs). When the transformer winding is formed on the winding equipment, at least one image can show at least one of the layers on the upper surface of the transformer winding. The image of the uppermost layer on the upper surface of the transformer winding can be processed in this method to detect differences.

[0137] The system can be configured to repeatedly process at least one image on a continuous basis (i.e., continuously) during the winding process and compare the results of object classification with design data.

[0138] When the system is configured to use one or more ML models (such as one, several, or all of the first, second, and third ML models mentioned above), labeled image data can be used to train the ML models. Labeled image data can include synthetically generated images derived from real-world images that demonstrate errors during the winding process. The generation of synthetically generated images can include one or more image modification operations, such as rotation, mirroring, adding distortion, etc.

[0139] ML models may include at least one of the following: deep learning models, such as the U-Net model; random forest model; recurrent neural network (RNN); convolutional neural network (CNN).

[0140] Transformer windings can be the following: power transmission systems (such as high-voltage and medium-voltage power transmission systems), power distribution systems, railway transformers, or reactor or inductor systems (such as shunt reactors and / or variable reactors).

[0141] According to another aspect of the invention, a manufacturing system for manufacturing transformer windings, transformer cores, or transformers is provided. The system includes: a winding apparatus; and, according to an embodiment, a system configured to monitor the winding process forming the transformer windings.

[0142] According to a further embodiment, machine-readable instruction code is provided that, when executed by at least one programmable circuit, causes the programmable circuit to perform the method according to the embodiment.

[0143] According to a further embodiment, a non-transitory storage medium is provided thereon storing machine-readable instruction code that, when executed by at least one programmable circuit, causes the programmable circuit to perform the method according to the embodiment.

[0144] According to another aspect of the invention, a transformer winding, transformer core, or transformer manufactured using the method according to the embodiment is provided.

[0145] The effects achieved by the system according to the embodiments correspond to the effects disclosed in detail in association with the method according to the embodiments.

[0146] Embodiments of the present invention are disclosed through the following list of embodiments: Example 1: A method for monitoring the winding process of forming a transformer winding on a winding device, the method comprising: processing at least one image captured during the winding process by a computer vision system, showing at least a portion of the transformer winding, wherein processing the at least one image includes performing object classification; detecting a difference between the result of the object classification and design data of the transformer winding by the computer vision system; and performing at least one action by the computer vision system in response to the detected difference.

[0147] Example 2: According to the method of Example 1, wherein the at least one action includes: the computer vision system causing the human-machine interface (HMI) to output information depending on the detected differences.

[0148] Example 3: According to the method of Example 2, the information includes one, several, or all of the following: alarms and / or warnings; information indicating the location of the detected difference; information indicating the time when the detected difference occurred; information about the root cause of the detected difference; and instructions for correcting the detected difference.

[0149] Example 4: The method according to any of the foregoing examples, wherein the at least one action includes one or both of the following: a mitigation action; a correction action.

[0150] Example 5: According to any of the methods described in the foregoing embodiments, the action includes stopping the rotation and / or translation of the winding device.

[0151] Example 6: The method according to any of the foregoing embodiments further includes: determining at least one image acquisition time of the at least one image during the winding process; and performing image acquisition device control and / or image selection based on the determined at least one image acquisition time.

[0152] Example 7: According to any of the methods described in the foregoing embodiments, processing the at least one image includes identifying a region of interest (ROI) in which object classification is to be performed, optionally the boundary of the ROI may include the edge of the shield of a transformer winding.

[0153] Example 8: According to any of the preceding embodiments, the method of processing the at least one image includes: using one or more location measurements of a winding device; and / or executing at least one machine learning (ML) model, the at least one ML model having: an input layer that receives at least a portion of the at least one image; and an output layer that provides a region of interest (ROI) in which object classification is to be performed, and / or an object classifier.

[0154] Example 9: According to any of the methods described in the foregoing examples, the detection of differences includes: determining a target sequence of objects along the radial direction of the transformer winding based on design data; and comparing the target sequence with the results of object classification.

[0155] Example 10: According to the method described in any of the foregoing embodiments, the object classification is performed on the following: at least conductors, fillers and shielding in the at least one image; and / or crosses; and / or other components of the transformer winding; and / or other structural features of the transformer winding.

[0156] Example 11: A method for manufacturing a transformer winding, the method comprising: controlling a winding device thereon to form a transformer winding by at least one control device; and performing the method according to any one of the foregoing embodiments while forming the transformer winding.

[0157] Example 12: A method for manufacturing a transformer core or a transformer, the method comprising: manufacturing at least one transformer winding by means of the method according to Example 11; and assembling a transformer core or transformer including said at least one transformer winding.

[0158] Example 13: A system configured to monitor the winding process of forming a transformer winding, the system comprising: a computer vision system configured to: process at least one image captured during the winding process, showing at least a portion of the transformer winding, to perform object classification; detect a difference between the result of the object classification and design data of the transformer winding; and perform an action in response to the detected difference.

[0159] Example 14: The system according to Example 13 further includes a human-machine interface (HMI), wherein a computer vision system is configured to control the HMI in response to detected differences.

[0160] Example 15: A manufacturing system for manufacturing transformer windings, transformer cores, or transformers, the system comprising: winding equipment; and the system according to Example 13 or Example 14.

[0161] Various effects and advantages are achieved through embodiments of the present invention. These systems and methods allow for the detection of errors while the winding process is ongoing. This allows for corrective and / or mitigating actions to be taken during the winding process. Even if the error is irreparable, these systems and methods offer advantages because the winding process can be terminated earlier (i.e., before the transformer winding is completed), thereby reducing the amount of material required for the base.

[0162] If the winding process is complete, these systems and methods can also detect errors located inside the transformer windings. This offers a significant advantage over techniques that analyze the transformer windings after the winding process is complete.

[0163] These systems and methods are configured to automatically monitor the winding process using images acquired by one or more image acquisition devices, such as cameras.

[0164] These systems form components of transformers (such as transformers in transmission or distribution systems) with complex configurations and structures. Attached Figure Description

[0165] Embodiments of the invention will be described with reference to the accompanying drawings, wherein similar or identical reference numerals denote elements having similar or identical configurations and / or functions.

[0166] Figure 1 The system includes a monitoring system.

[0167] Figure 2 The monitoring system is shown.

[0168] Figure 3 A flowchart is shown.

[0169] Figure 4 A flowchart is shown.

[0170] Figure 5 The monitoring system is shown.

[0171] Figure 6 Images illustrating ROI detection and object classification are shown.

[0172] Figure 7 An image illustrating ROI detection is shown.

[0173] Figure 8 A block diagram of the monitoring system is shown.

[0174] Figure 9 A flowchart is shown.

[0175] Figure 10 A flowchart is shown.

[0176] Figure 11 A transformer is shown.

[0177] Figure 12 The assembly system is shown.

[0178] Figure 13 A flowchart is shown. Detailed Implementation

[0179] Embodiments of the invention will be described with reference to the accompanying drawings. In the drawings, similar or identical reference numerals denote elements having similar or identical configurations and / or functions.

[0180] Although embodiments will be described in association with transformer windings configured to be installed in a transformer in a power transmission or distribution system, embodiments are not limited thereto. Although embodiments will be described in association with certain winding configurations, such as transformer windings comprising several discs with spacers arranged between the discs, embodiments are not limited thereto.

[0181] Unless otherwise expressly stated, the features of the embodiments may be combined with each other.

[0182] Figure 1 A system 10 according to an embodiment is shown. System 10 is configured to perform a winding process to form a transformer winding. System 10 is configured to monitor the winding process in which the transformer winding 20 is formed. System 10 includes a computer vision system 40 for monitoring the winding process to detect errors that may occur during the winding process.

[0183] System 10 includes a winding device 30. The winding device 30 includes a winding support on which the transformer winding is supported during the winding process. The winding device 30 can be configured to rotate the transformer winding 20 about an axis 31 while forming the transformer winding 20. The winding device 30 can be configured to translately displace a formed portion of the transformer winding 20 along the axis 31. Displacement along the axis 31 can be performed after the layers (or discs) of the transformer winding 20 have been completed. The distance by which the formed portion of the transformer winding 20 is displaced can correspond to the height of each disc formed during one full rotation of the winding device, possibly including spacers inserted between adjacent layers.

[0184] The winding device 30 may include a first motor 32 configured to rotate the transformer windings during the winding process.

[0185] The winding device 30 may include a second motor 33 configured to translate the transformer winding 20 during the winding process. The translation can be performed intermittently, for example, incrementally or stepwise. The translation can be performed after a full revolution around axis 31. As will be described in more detail below, the translation along axis 31 also provides the effect that, during the winding process, the image captured by the image acquisition device 41 for visual quality monitoring is held at approximately the same height above the top layer 21 of the partially wound transformer winding 20.

[0186] The winding device 30 may include one or more position sensors 34. The one or more position sensors 34 may be configured to measure the rotational and / or translational position of the winding device 30. The image processing system 50 of the computer vision system 40 may use these position measurements to determine, for example, which images require further processing, and / or perform ROI detection.

[0187] As a supplement or alternative to receiving position measurements from (multiple) position sensors 34, the computer vision system may receive position information from the winding device controller 35.

[0188] The computer vision system includes an image processing system 50. The image processing system 50 will be described in more detail below.

[0189] Image processing system 50 is typically configured to process images 42 captured by one or more image acquisition devices 41. One or more image acquisition devices 41 may be mounted in a fixed manner relative to the frame of winding device 30. Translational displacement achieved by the winding device along axis 31 prevents excessive change in the angle between the optical axis of image acquisition device 40 and the top layer 21 of the transformer winding 20 being wound.

[0190] The image processing system 50 is configured to perform at least object classification. Object classification can specify a range of different objects, such as conductors and fillers seen radially from the inner edge to the outer edge of the topmost layer 21 of the formed portion of the transformer winding 20.

[0191] Image processing system 50 may optionally be configured to perform additional processing on image 42. For illustration, image processing system 50 may be configured to perform preprocessing to achieve one or more of the following: adjusting perspective; enhancing contrast; identifying periodic features; eliminating light reflections, etc.

[0192] Alternatively or additionally, the image processing system 50 may be configured to perform region of interest (ROI) determination. ROI determination may be implemented as a classifier. The classifier can provide a binary output indicating whether a given pixel is included in the ROI. Object classification can then be restricted to the ROI. When the transformer winding 20 comprises several layers with shielding arranged between them, the ROI classifier can operate to distinguish pixels representing conductors or fillers from pixels representing shielding.

[0193] Image processing system 50 uses the results of object classification and design data 49 specifying the target design of transformer winding 20 to detect discrepancies that indicate errors occurring during the winding process. Image processing system 50, or a separate computing system coupled to it, can process design data 49 into a format that can be compared with the object classification results. For illustration, design data 49, including one or more cross-plots, can be automatically converted into a list that specifies which objects are expected to be present in any ROI within the field of view of image acquisition device 41 for the desired rotational and translational position of winding device 30.

[0194] When the image processing system 50 detects a difference between the result of object classification and the result expected according to the design data, the image processing system 50 may trigger an action. This action may include output via a human-machine interface (HMI) 37. The output may indicate that a difference has been detected, but may also include additional information, such as information indicating the location of the detected difference and / or suggestions on how the identified difference can be corrected (if such correction is possible). Additionally or alternatively, this output may affect the operation of the winding device 30. For illustration, the image processing system 50 may trigger the winding device controller 35 to stop rotation and / or translation in response to a detected difference. This facilitates visual inspection and possible corrective actions by a human operator.

[0195] The additional features of the image processing system according to the embodiments will now be described in more detail.

[0196] Figure 2 An image processing system 50 is shown. The image processing system 50 has at least one first interface 51 for receiving an image 42. The received image 42 shows at least a portion of the topmost layer 21 of the transformer winding 20 during various stages of the winding process. The at least one first interface 51 may be a data interface configured to receive the image 42 as digital data. The at least one first interface 51 may also be configured to receive design data 49 specifying a desired target configuration of the transformer winding 20.

[0197] The results of the processing performed by the image processing system 50 can be output via at least one first interface 51. Alternatively, the image processing system 50 may have at least one second interface 52 configured to communicatively connect to the winding device controller 35 and / or the HMI 37. The image processing system 50 may generate and output signals or commands to trigger actions, such as information provided by the HMI 37 regarding detected differences and / or causing the winding device controller 35 to react to detected differences.

[0198] Image processing system 50 may include storage system 53. Images (such as image streams received from image acquisition devices 41) may be temporarily stored in storage system 53 for processing and further analysis. Storage system 53 may also include parameters for processing image 42, such as parameters of an ROI classifier trained to identify ROIs in the image and / or parameters of an object classifier trained to distinguish different objects (such as conductors, fillers, shielding, and / or other components or structural features of crosses and / or windings).

[0199] Image processing system 50 may include one or more processing circuits 60. The one or more processing circuits 60 may be programmable circuits. The one or more processing circuits may include, but are not limited to, any or any combination of integrated circuits, integrated semiconductor circuits, processors, controllers, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and circuits including qubits and / or quantum gates.

[0200] Image processing system 50 may include an image timing selector 61. Image timing selector 61 may receive as input the locations of (multiple) winding devices and / or the image stream included in image 42. Image timing selector 61 may then determine which images require further processing to perform object classification. Image timing selector 61 may select already captured images from the image stream for more detailed processing. Alternatively or additionally, image timing selector 61 may predict at which image acquisition times to capture images, which may be particularly useful for detecting potential discrepancies between the detection and design data. Image processing system 50 may cause (multiple) image acquisition devices 41 to capture images based on the output of image timing selector 61.

[0201] Image processing system 50 may include ROI detector 62. ROI detector 62 may include a first classifier, also referred to herein as an ROI classifier. ROI detector 62 receives pixels of an image, possibly pixels from preprocessed images. ROI detector 62 may classify each pixel as belonging to an ROI or not belonging to any ROI. ROI detector 62 may use edge detection techniques. Alternatively or additionally, ROI detector 62 may include one or more trained ML models suitable for performing image segmentation tasks. Deep learning models are an example of such ML models, but implementation of ROI detector 62 is not limited to this.

[0202] Image processing system 50 may include object classifier 63. Object classifier 63 is configured to classify each pixel in at least one ROI as belonging to one of several object categories, which represent different types of physical components included in transformer winding 20. For illustration, these object categories may include those for: conductors (it should be understood that conductors may be covered with dielectric material), filler between conductors that may be arranged within layers (also called discs) of transformer winding, or shielding that may be arranged between adjacent layers of transformer winding. This list is not exhaustive. Additional or alternative object categories may be provided to reflect the component and / or structural characteristics of the winding.

[0203] More than one object classifier 63 may be used. For illustration, one object classifier may be provided, which is specifically adapted to distinguish conductors, fillers and shielding; and another object classifier, which is specifically adapted to identify intersections and / or distinguish inner intersections from sequential intersections.

[0204] The image processing system 50 may include a difference detector 64 configured to detect differences between objects classified by the object classifier 63 and a desired sequence of objects, such as differences in the radial direction along the top layer 21 of the transformer winding, based on design data 49.

[0205] Image processing system 50 may include output generator 65. Output generator 65 may be configured to generate a command or signal that triggers an action in response to the output of difference detector 64. The action may include providing information related to the detected difference (such as a warning or alarm), and / or mitigation or correction actions.

[0206] Figure 3 This is a flowchart of method 80. Method 80 can be executed automatically by computer vision system 40 or image processing system 50 of computer vision system 40.

[0207] At process block 81, image preprocessing is performed. Image preprocessing may include correcting perspective effects (e.g., by applying a transform matrix to the image), edge enhancement, contrast enhancement, etc. Image preprocessing may include performing at least one transformation of the image, depending on the intrinsic and / or extrinsic camera parameters of the image acquisition device's camera. The intrinsic and / or extrinsic camera parameters may be known (e.g., provided by the camera manufacturer) or determined in a calibration routine.

[0208] At process block 82, the ROI is detected. ROI detection may include edge detection. The ROI can be detected such that it is defined by the edge of the shield on the top layer 21 and by the radial inner and outer circumferences of the top layer 21.

[0209] At process block 83, object classification is performed. Object classification distinguishes different physical components present in transformer winding 20. Object classification determines the sequence in which conductors and fillers are arranged along the radial direction of transformer winding 20. Object classification may additionally or alternatively distinguish different structural features (such as crossovers) present in transformer winding 20.

[0210] At process block 84, a possible difference between the object classification result and the design data is detected. An action can be executed in response to the detection of this difference.

[0211] Sensor signals can be used at various stages of the process. Sensor signals can indicate the position of the winding device. Sensor signals can indicate the translational and / or rotational position of the winding device.

[0212] The winding device position can be used in various ways. For illustration, the sensor signal can determine which image preprocessing to apply to accommodate the corresponding height difference between the image acquisition device 41 and the top layer 21. The winding device position can also be used when performing ROI detection.

[0213] The winding device location can also be used when determining which sequence of objects is expected to exist in an image of a given winding device location (e.g., in a ROI) based on design data.

[0214] Figure 4 This is a flowchart of method 90. Method 90 can be executed automatically by computer vision system 40 or image processing system 50 of computer vision system 40.

[0215] At process block 91, the positions of (multiple) winding devices are obtained. The positions of (multiple) winding devices can be obtained from (multiple) sensors 34 or the winding device controller 35.

[0216] At process block 92, an object sequence is determined based on one or more images 42 (e.g., in the top layer 21). The object sequence can be determined along a radial direction from the inner circumference to the outer circumference of the top layer 21. The object sequence can be determined using, for example, ROI detection and object classification, as previously described.

[0217] At process block 93, it is determined whether the identified sequence of objects and the design data are consistent with each other. If consistency exists (as is typically the case), the process returns to process block 91. Various process blocks can be repeated for multiple rotary winding device positions and (if any) multiple translational winding device positions.

[0218] At process block 94, an action is triggered if inconsistencies are detected between the sequences of objects in the design data. As previously described, this action may include controlling the HMI and / or triggering corrective and / or mitigation actions.

[0219] In the computer vision system 40, and more specifically in the image processing system 50, various classifiers may be implemented using techniques such as edge detection (for example, detecting the boundaries between different objects within an ROI), periodicity detection (for example, detecting the periodic arrangement of layers on the outer cylindrical surface of a transformer winding and / or detecting sequences of objects within an ROI); but are not limited thereto.

[0220] One or more classifiers can be implemented using trained ML models. More than one ML model can be used. For illustration, image processing system 50 may include a first ML model configured to determine at which image acquisition times the most relevant images are captured for further analysis. Alternatively or additionally, image processing system 50 may include a second ML model configured to detect Relative Opinions (ROLs). Alternatively or additionally, image processing system 50 may include a third ML model configured to perform object classification.

[0221] Several trained ML models or other classifiers can be deployed and used to perform any of these functions. To illustrate, several ROI classifiers can be applied to the same image data to achieve more accurate detection of potential errors during the winding process.

[0222] Figure 5 An image processing system 50 is illustrated, comprising at least one ML model 101. The ML model 100 may include an input layer 101, an output layer 102, and a hidden layer 103. The specific configuration of the ML model 100 may depend on the functions it will perform. Similarly, the input layer 101 and the output layer 102 depend on the functions that the ML model 100 has in the computer vision system.

[0223] A computer vision system may use a first ML model to determine at least one image acquisition time of an image to be processed to perform object classification. In this case, the first ML model may have a first input layer 101 that receives image pixels from an image stream. The first ML model may have a first output layer 102 that outputs values ​​indicating whether further analysis should be performed on any of the images. Alternatively or additionally, the first input layer may receive the location of a winding device, such as the rotational and / or translational position of a winding support on which transformer windings are formed.

[0224] Identifying a Region of Interest (ROI) may involve applying at least one ROI classifier to at least one image. Several ROI classifiers may be applied to at least one image. Each of the one or more ROI classifiers may indicate, for each pixel, whether the corresponding pixel is classified as inside or outside the ROI.

[0225] One or more ROI classifiers may include at least one second ML model. The at least one second ML model may have a second input layer 101 that receives at least one pixel of an image and / or receives (multiple) winding device locations. The at least one second ML model may have a second output layer 102 that outputs a value for each pixel indicating whether the corresponding pixel is in the ROI. Alternatively or additionally, the second input layer may receive winding device locations, such as the rotational and / or translational positions of winding supports on which transformer windings are formed.

[0226] At least one second ML model 100 may include a deep learning ML model, such as the U-Net ML model.

[0227] At least one third ML model can be used to perform object classification. The at least one third ML model may have a third input layer 101 that receives pixels located within a Region of Interest (ROI). The at least one third ML model may have a third output layer 102 that outputs an object classification result indicating which of several object categories a corresponding pixel in the ROI has been assigned to. The third ML model may include, but is not limited to, a random forest (RF) model.

[0228] Figure 6 The image 120 shown is captured by the image acquisition device 41 while the transformer winding 20 is being formed on the winding device 30.

[0229] As mentioned above, optional preprocessing can be performed on the raw image data. Optional preprocessing can be used to take into account perspective effects caused by the arrangement of the optical axis relative to the topmost layer 21. Optional preprocessing can also enhance certain features, such as edges and / or contrast and / or periodicity.

[0230] ROI detection identifies ROI 125 in image 120. ROI 125 can be defined by the edge 126 of shield 121. ROI 125 can also be defined radially by the inner and outer circumferential boundaries of the innermost and outermost conductors seen in image 120.

[0231] Object classification is performed to categorize objects such as conductor 122 and filler 123. A sequence of these object types can be determined along a radial direction (i.e., radially inward or outward). The consistency of this object category sequence with the design data is then checked. A target sequence of object categories expected when conforming to the design data can be determined. The object category sequence determined based on image 120 can be compared with the target sequence.

[0232] Figure 7 Another image with ROI 125 is shown (in this case, determined using a trained U-Net ML model as a classifier). ROI detection works reliably. In particular, the top layer 21 can be reliably distinguished from the outer cylindrical surface of the already wound transformer windings. The edges of the shield can be reliably determined.

[0233] It is not necessary to perform object classification for every image acquired or available by the image acquisition device 41. The image acquisition time for images to be analyzed in detail (particularly by performing object classification) can be determined by the computer vision system 40 (e.g., by its image processing system 50).

[0234] Figure 8 A functional block diagram 130 is shown, illustrating the operation of the computer vision system 40 to determine the image acquisition time of the images it captures around further analysis.

[0235] Multiple winding device locations 131 and / or image streams 132 can be used as inputs to determine the relevant image acquisition time 133. Alternatively, design data can be taken into account to determine which winding device locations 131 can detect minor occlusion of the relevant feature, which is then translated into corresponding image acquisition times during the winding process.

[0236] The determined image acquisition time can be used in various ways for the images to be analyzed in detail.

[0237] Camera control operation 134 can be performed to cause image acquisition device 41 to perform image acquisition at a desired time. Alternatively or additionally, at block 135, images can be selected from an image stream (which can be continuously acquired) according to the determined image acquisition time(s).

[0238] As an alternative to determining a specific image acquisition time, image processing can be performed on a continuous basis. However, it is usually not necessary (but is possible) to analyze every image frame.

[0239] Figure 9 This is a flowchart of process 140. Process 140 can be executed by computer vision system 40 or a separate computing system connected to computer vision system 40. Process 140 transforms design data into a format that can be easily compared with the results of object classification.

[0240] At process block 141, design data is obtained. Design data can be retrieved from a database. The design data may be specific to the transformer in which the transformer windings will be used. The design data may define the three-dimensional configuration of the transformer windings. The design data may include cross diagrams, but may have alternative formats, such as instruction codes executed by the winding equipment or its controller.

[0241] At process block 142, the design data is processed to infer the target sequence of object categories for the corresponding winding device locations based on the design data and the state of the winding process (which determines the location of the winding device).

[0242] The target sequence of object categories (also referred to as the target sequence of objects in this paper) can be compared with the results of object classification performed on at least one image to detect possible differences.

[0243] Computer vision systems may include one or more ML models, but other processing techniques may also be used. One challenge when using one or more ML models is providing a suitable training set, as images showing errors during the winding process are, fortunately, rare compared to images illustrating perfect operation. The methods and systems disclosed in this paper can synthetically generate additional images showing winding errors, thereby providing a more balanced training dataset. This improves accuracy during inference.

[0244] ML model training can be performed by a computer vision system. ML model training can be repeated during the field use of the computer vision system. ML model training can also be performed by a computing system separate from the computer vision system.

[0245] Figure 10 A flowchart of method 150 is shown. Method 150 can be executed to train one or more ML models that can be used in a method or system for monitoring the winding process.

[0246] At process block 151, the labeled training images are obtained. It may be possible, but is not necessary, to capture the labeled training images on the same winding device used in the ML model during the inference phase.

[0247] At process block 152, training images labeled as indicating winding errors are identified. Additional training images are synthetically generated. Generating additional synthetic images with labels indicating winding errors may include transformations of the images labeled as indicating winding errors, such as rotation, translation, distortion, and / or reflection.

[0248] At process block 153, ML model training is performed using images that are generated, especially synthetically.

[0249] At process block 154, the trained ML model can be deployed to the computer vision system for inference during field use of the computer vision system.

[0250] ML models can be ROI detectors or object classifiers, but are not limited to these.

[0251] According to an embodiment, a method for forming transformer windings on a winding device includes monitoring the winding process while performing the winding process using computer vision techniques described in detail herein.

[0252] The resulting windings can be assembled to form a transformer core or transformer, such as a transmission transformer, or distribution transformer, shunt reactor, and / or variable reactor for use in, for example, high-voltage or medium-voltage transmission networks.

[0253] Figure 11 A transformer 160 is schematically shown, comprising at least one transformer winding 163 formed using a method according to an embodiment. The at least one transformer winding 163 is assembled with a yoke 162 to form a transformer core. The transformer 160 may include additional components such as a transformer housing 161, insulating oil within the transformer housing 161, bushings (not shown), a transformer breather (not shown), etc.

[0254] Figure 12 This is a block diagram of a transformer core manufacturing system 170 configured to manufacture transformers. Manufacturing system 170 includes a winding device 30 and a yoke manufacturing system 171. Manufacturing system 170 includes a computer vision system 40 according to an embodiment.

[0255] Under the monitoring of computer vision system 40, the transformer windings formed on winding equipment 30 are used in conjunction with other transformer components (such as yokes) to assemble transformer cores or transformers. This assembly can be performed by assembly system 172.

[0256] Figure 13 This is a flowchart of method 180. Method 180 can be executed by manufacturing system 170.

[0257] At process block 181, visual quality control is performed during the winding process, during which the transformer windings are formed on the winding equipment.

[0258] At process block 182, it is determined whether any discrepancies have been detected during the winding process. Such discrepancies would indicate that the object seen by the computer vision system does not match or does not perfectly match the object expected according to the design data.

[0259] If no discrepancy is detected at process block 183, the transformer windings can be used to assemble the transformer core or transformer.

[0260] If a difference has been detected at process block 184, the transformer winding can be discarded.

[0261] In response to the detection of a difference, if a difference has been detected but subsequently corrected by an appropriate correction action, the transformer winding may still be potentially used at process block 183.

[0262] Although embodiments have been described with reference to the accompanying drawings, modifications and alterations may be implemented in other embodiments.

[0263] For the purpose of illustration, although embodiments in which transformer windings may include several discs and shielding between the discs have been described, the techniques disclosed herein are not limited to monitoring the winding process of such transformer windings.

[0264] To further illustrate, although an embodiment in which an image captured by a single image acquisition device is analyzed has been described, these techniques are also applicable when images are captured by more than one image acquisition device. Using several image acquisition devices can be beneficial for reading the potential adverse effects of using specific occlusions.

[0265] For further explanation, although an embodiment in which a trained ML model can be used to perform certain processing operations has been described, other processing techniques may be used.

[0266] To illustrate further, although an embodiment in which a classifier is used to perform ROI detection has been described, other techniques that do not involve a classifier can be used to perform ROI detection if it is to be performed.

[0267] For further explanation, although embodiments have been described in connection with monitoring the winding process of transformer windings, these techniques can also be used to monitor the manufacture of windings used in reactance or inductance systems, such as shunt reactors or variable reactors.

[0268] Various effects and advantages are achieved through embodiments of the present invention. For illustration, the embodiments allow for error detection while the winding process is ongoing. This allows for corrective and / or mitigating actions to be taken during the winding process. The risk of transformer failure caused by winding errors is reduced.

[0269] The methods, systems, and apparatuses are particularly suitable for use with power transformers in power transmission systems, such as high-voltage, medium-voltage, or distribution systems. These methods, systems, and apparatuses are not limited to these applications. For illustration, the techniques disclosed herein can also be used with railway transformers.

[0270] The description and drawings illustrating aspects and embodiments of the invention should not be considered limiting—the claims define the protected invention. In other words, although the invention has been illustrated and described in detail in the drawings and foregoing description, such illustrations and descriptions are to be considered illustrative or exemplary rather than limiting. Various mechanical, compositional, structural, electrical, and operational changes may be made without departing from the spirit and scope of the description and claims. In some instances, well-known circuits, structures, and techniques have not been shown in detail so as not to obscure the invention. Therefore, it will be understood that changes and modifications can be made by those skilled in the art within the scope and spirit of the appended claims. In particular, the invention covers further embodiments having any combination of features from the different embodiments described above and below.

[0271] This disclosure also covers all further features shown separately in the accompanying drawings, although they may not have been described in the preceding or following description. Furthermore, single alternatives to the embodiments described in the drawings and description, and single alternatives to their features, may be excluded from the subject matter of the invention or from the disclosed subject matter. This disclosure includes the subject matter consisting of features defined in the claims or exemplary embodiments, and the subject matter including said features.

[0272] The word "comprising" does not exclude other elements or steps, and the indefinite articles "a" or "an" do not exclude multiple. A single unit or step may perform the function of several features recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that combinations of these measures cannot be advantageously used. Components described as being connected or linked may be electrically or mechanically directly connected, or they may be indirectly connected via one or more intermediate components. Any reference signs in the claims should not be construed as limiting the scope.

[0273] Machine-readable instruction code may be stored / distributed on suitable media, such as optical storage media or solid-state media supplied with or as part of other hardware, but may also be distributed in other forms, such as via wide area networks or other wired or wireless telecommunications systems. Furthermore, machine-readable instruction code may also be a data structure product or signal used to embody a particular method (such as the method according to an embodiment).

Claims

1. A method for monitoring the winding process of forming a transformer winding (20) on a winding device (30), the method comprising: At least one image (42) captured during the winding process by a computer vision system (40) showing at least a portion of the transformer winding (20), wherein processing the at least one image (42) includes performing object classification; The computer vision system (40) detects the difference between the object classification result and the design data (49) of the transformer winding (20), wherein detecting the difference includes: determining a target sequence of objects along the radial direction of the transformer winding (20) based on the design data (49); and comparing the target sequence with the object classification result; The computer vision system (40) performs at least one action in response to a detected difference; and The at least one image acquisition time of the at least one image (42) during the winding process is determined based on the translational and rotational positions of the winding device (30) in conjunction with the design data (49).

2. The method according to claim 1, wherein The at least one action includes: the computer vision system (40) causing the human-machine interface (HMI) (37) to output information depending on the detected differences.

3. The method according to claim 2, wherein The information includes one, several, or all of the following: Alarms and / or warnings; Information indicating the location of the detected differences; Information indicating the time when the detected difference occurred; Information regarding the root cause of the detected differences; Instructions for correcting the detected differences.

4. The method according to any one of claims 1 to 3, wherein The at least one action includes one or both of the following: Relief actions; Correction action.

5. The method according to any one of claims 1 to 3, wherein The action includes stopping the rotation and / or translation of the winding device (30).

6. The method according to any one of claims 1 to 3, further comprising: Image acquisition device control and / or image selection are performed based on at least one determined image acquisition time.

7. The method according to any one of claims 1 to 3, wherein Processing the at least one image (42) includes identifying a region of interest (ROI) (125) in which the object classification is to be performed, optionally the boundary of the ROI (125) includes the edge (126) of the shield of the transformer winding (20).

8. The method according to any one of claims 1 to 3, wherein, Processing the at least one image (42) includes: Measured values ​​using one or more locations of the winding device (30); and / or Execute at least one machine learning ML model (100), the at least one machine learning ML model having: an input layer (101) receiving at least a portion of the at least one image (42); and an output layer (102), wherein the output layer provides a region of interest (ROI) (125) in which the object classification is to be performed, and / or an object classifier.

9. The method according to any one of claims 1 to 3, wherein The object classification is categorized as follows: At least the conductor, filler and shielding in the at least one image (42), and / or Cross, and / or Other components of the transformer winding (20), and / or Other structural features of the transformer winding (20).

10. A method for manufacturing a transformer winding (20), comprising: The winding device (30) on which the transformer winding (20) is formed is controlled by at least one control device (35). as well as The method according to any one of claims 1 to 9 is performed while the transformer winding (20) is being formed.

11. A method for manufacturing a transformer core (162, 163) or a transformer (160), comprising: At least one transformer winding (20) is manufactured by the method according to claim 10; and Assemble a transformer core (162, 163) or transformer (160) including at least one transformer winding (20).

12. A system configured to monitor the winding process of forming a transformer winding (20) on a winding device (30), the system comprising: The computer vision system (40) is configured as follows: At least one image (42) captured during the winding process, showing at least a portion of the transformer winding (20), is processed to perform object classification; The detection of the difference between the object classification result and the design data (49) of the transformer winding (20) includes: determining a target sequence of objects along the radial direction of the transformer winding (20) based on the design data (49); and comparing the target sequence with the object classification result. To perform an action in response to a detected difference; and The at least one image acquisition time of the at least one image (42) during the winding process is determined based on the translational and rotational positions of the winding device (30) in conjunction with the design data (49).

13. The system according to claim 12, further comprising: Human-machine interface (HMI) (37), wherein the computer vision system (40) is configured to control the HMI in response to the detected differences.

14. A manufacturing system for manufacturing transformer windings (20), the system comprising: Winding equipment (30); as well as The system according to claim 12 or claim 13.

Images