Fan deformation identification method, device and system

Abstract

The present disclosure relates to a fan deformation identification method, device and system, and relates to the technical field of image recognition. The method comprises: obtaining a measured image of a fan; the type of the measured image comprises a three-dimensional point image and a vibration signal image; inputting the three-dimensional point image and the vibration signal image into a pre-trained deformation identification model; obtaining an identification result output by the deformation identification model after identifying the deformation state of the fan based on the three-dimensional point image and the vibration signal image according to a preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of different original sample images; and determining whether the fan has deformed according to the identification result. The present scheme improves the accuracy of identifying the deformation state of the fan.

Description

Technical Field

[0001] This disclosure relates to the field of image recognition technology, and in particular to a method, apparatus and system for wind turbine deformation recognition. Background Technology

[0002] In related technologies, to accurately detect the deformation of wind turbine towers and blades and ensure the safe operation of wind turbines, methods such as radar are typically used to acquire deformation and vibration data of the towers and blades. Based on this data, deep neural networks are then used to determine whether deformation has occurred. However, due to the limited amount of deformation and vibration data of the wind turbine towers and blades under deformed conditions, it is difficult to complete the initial training of the deep neural network, resulting in low accuracy in identifying wind turbine deformation using the neural network model. Summary of the Invention

[0003] To overcome the problems existing in related technologies, this disclosure provides a method, device and system for wind turbine deformation identification.

[0004] According to a first aspect of the present disclosure, a wind turbine deformation identification method is provided, comprising:

[0005] Acquire measured images of the wind turbine; the types of measured images include three-dimensional point plots and vibration signal plots;

[0006] The three-dimensional point map and the vibration signal map are input into the pre-trained deformation recognition model;

[0007] The deformation recognition model obtains the recognition result after identifying the deformation state of the wind turbine based on the three-dimensional point map and the vibration signal map according to the preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of the different original sample images.

[0008] Based on the identification results, it is determined whether the fan has undergone deformation.

[0009] According to a second aspect of the present disclosure, a wind turbine deformation identification device is provided, comprising:

[0010] The first acquisition unit is used to acquire measured images of the wind turbine; the types of measured images include three-dimensional point plots and vibration signal plots.

[0011] The input unit is used to input the three-dimensional point map and the vibration signal map into the pre-trained deformation recognition model;

[0012] The second acquisition unit is used to acquire the recognition result output by the deformation recognition model after recognizing the deformation state of the wind turbine based on the three-dimensional point map and the vibration signal map according to the preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of the different original sample images.

[0013] A determining unit is used to determine whether the fan has deformed based on the identification result.

[0014] According to a third aspect of the present disclosure, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method as described in any one of the first aspects.

[0015] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method as described in any one of the first aspects.

[0016] According to a fifth aspect of the present disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method as described in any one of the first aspects.

[0017] The technical solutions provided by the embodiments of this disclosure can include the following beneficial effects: acquiring measured images of the wind turbine; the types of measured images include three-dimensional point maps and vibration signal maps; inputting the three-dimensional point maps and vibration signal maps into a pre-trained deformation recognition model; obtaining the recognition result output by the deformation recognition model after recognizing the deformation state of the wind turbine based on the three-dimensional point maps and vibration signal maps according to a preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of the different original sample images; determining whether the wind turbine has deformed based on the recognition result. By generating more wind turbine images with deformation by using shape features and appearance features from different images, the feature vector set of the prototype network model is enriched, thereby improving the accuracy of the prototype network model in recognizing the deformation state of the wind turbine.

[0018] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0019] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0020] Figure 1 This is a flowchart illustrating a wind turbine deformation identification method according to an exemplary embodiment.

[0021] Figure 2 This is a flowchart illustrating another method for identifying wind turbine deformation according to an exemplary embodiment.

[0022] Figure 3 This is a block diagram illustrating a wind turbine deformation recognition device according to an exemplary embodiment.

[0023] Figure 4 This is a block diagram illustrating an apparatus for wind turbine deformation identification according to an exemplary embodiment. Detailed Implementation

[0024] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0025] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the present disclosure. The singular forms “a” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0026] It should be understood that although the terms first, second, third, etc., may be used to describe various information in embodiments of this disclosure, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first information may also be referred to as second information without departing from the scope of embodiments of this disclosure, and similarly, second information may also be referred to as first information. Depending on the context, the words “if” and “suppose” as used herein may be interpreted as “when”, “when”, or “in response to a determination”.

[0027] Furthermore, various forms of processes shown in the embodiments of this disclosure can be used to reorder, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and no limitation is imposed herein.

[0028] In related technologies, to accurately detect the deformation of wind turbine towers and blades and ensure the safe operation of wind turbines, methods such as radar are typically used to acquire deformation and vibration data of the towers and blades. Based on this data, deep neural networks are then used to determine whether deformation has occurred. However, due to the limited amount of deformation and vibration data of the wind turbine towers and blades under deformed conditions, it is difficult to complete the initial training of the deep neural network, resulting in low accuracy in identifying wind turbine deformation using the neural network model.

[0029] One commonly used method for wind turbine deformation detection is the use of deep neural networks (DNNs). However, DNNs require a large amount of data, and the high cost of label collection remains a major challenge. Furthermore, some types of data may be rare and difficult to collect, leading to poor learning performance of deep neural networks. FSL (Few-Shot Learning) addresses the classification problem with limited test class labels. It is generally assumed that when the support set contains only a small number of labeled samples as reference, the FSL model must classify images of new classes. However, trained FSL models typically exhibit poor generalization ability. This generalization error mainly involves the gap between the training and test data, and the unconventionality of the test data samples exacerbates this gap. Various researches have emerged to address the high generalization error problem, most of which involve training augmentation to improve the generalization ability of FSL models. However, this augmentation involves training the model on increased amounts of data, consuming additional computational resources on each additional trained FSL model. Moreover, since the gap between the test data and the augmented training data remains significant, the performance improvement from such augmentation methods remains limited.

[0030] To address the aforementioned issues, this disclosure provides a method, apparatus, and system for wind turbine deformation recognition. The method involves acquiring measured images of the wind turbine; these images include 3D point maps and vibration signal maps; inputting the 3D point maps and vibration signal maps into a pre-trained deformation recognition model; obtaining the recognition result output by the deformation recognition model based on a preset feature vector set, identifying the wind turbine's deformation state using the 3D point maps and vibration signal maps; the feature vectors in the feature vector set are obtained based on the shape and appearance features of different original sample images; and determining whether the wind turbine has undergone deformation based on the recognition result. By generating more wind turbine images showing deformation through the shape and appearance features from different images, the feature vector set of the prototype network model is enriched, thereby improving the accuracy of the prototype network model in recognizing the wind turbine's deformation state.

[0031] Figure 1 This is a flowchart illustrating a wind turbine deformation identification method according to an exemplary embodiment, such as... Figure 1As shown, it should be noted that the wind turbine deformation identification method of this embodiment is applied in a wind turbine deformation identification device. For example... Figure 1 As shown, the method may include the following steps:

[0032] Step 101: Obtain the measured image of the wind turbine.

[0033] The types of measured images include three-dimensional point plots and vibration signal plots.

[0034] In some embodiments of this disclosure, the measured images can be any one or more of the wind turbine tower and wind turbine blades.

[0035] In one embodiment, the three-dimensional point map can be obtained by scanning the wind turbine using radar equipment.

[0036] Step 102: Input the three-dimensional point map and vibration signal map into the pre-trained deformation recognition model.

[0037] It is understandable that there is a correlation between the 3D point plot and the vibration signal plot and the deformation state of the fan. Therefore, the deformation state of the fan can be determined based on the 3D point plot and the vibration signal plot.

[0038] In one embodiment, after obtaining the on-site measured overall deformation image (i.e., 3D point map) and vibration signal of the wind turbine, a support dataset and a test dataset for the deformation recognition model can be established based on this data, according to the requirements of few-shot learning. The established dataset can have its number of categories and corresponding sample counts changed according to actual needs. For example, a 4-way 3-shot support dataset (containing 4 different categories, with 3 samples in each category) can be established by randomly sampling data based on requirements and the actual number of available samples. The deformation recognition model is then trained using the support dataset and the test dataset.

[0039] In some embodiments of this disclosure, the deformation recognition model can be a model that can directly obtain sample embedding representation information, that is, a model that can directly obtain sample feature vectors. For example, it can be a ProtoNet model or a Matching Networks model.

[0040] Step 103: Obtain the recognition result output by the deformation recognition model after recognizing the deformation state of the wind turbine based on the preset feature vector set, the three-dimensional point map, and the vibration signal map.

[0041] The feature vectors in the feature vector set are obtained based on the shape and appearance features of the different original sample images.

[0042] It should be noted that when the original sample image is a three-dimensional point map, the above-mentioned shape features may include any one or more of the contour edges and corner points, and the above-mentioned appearance features may include any one or more of the texture and reflectivity; when the original sample image is a vibration signal map, the above-mentioned shape features may include any one or more of the waveform and spectral features, and the above-mentioned appearance features may include any one or more of the distribution and noise characteristics.

[0043] Understandably, when the number of original samples is small, it is necessary to increase the number of samples to improve the accuracy of the deformation recognition model in identifying the deformation state of wind turbines. Therefore, shape features and appearance features can be extracted from the original sample images separately, and the shape features and appearance features of different original sample images can be fused to obtain more wind turbine sample images with deformation. Based on the wind turbine sample images obtained by the above fusion, the feature vector set of the deformation recognition model is enriched, so that the deformation recognition model can more accurately identify the deformation state of wind turbines, improve the generalization of the deformation recognition model, and does not require more computing resources and data requirements.

[0044] Step 104: Determine whether the fan has deformed based on the identification results.

[0045] In one embodiment, if the identification result is the deformation parameter of the fan, the deformation parameter is compared with a preset threshold. If the deformation parameter is greater than the preset threshold, it is determined that the fan has deformed.

[0046] In some embodiments of this disclosure, such as Figure 2 As shown, the feature vector set of the deformation recognition model is obtained through the following steps:

[0047] Step 201: Obtain the original sample image of the wind turbine.

[0048] The original sample images include the first sample measured image and the second sample measured image.

[0049] It should be noted that the first and second sample measured images belong to the same type. For example, both the first and second sample measured images are three-dimensional point plots, or both are vibration signal plots, and both are wind turbine blade images, or both are wind turbine tower images, or both are overall wind turbine images.

[0050] Step 202: Use a pre-trained data merging generator to extract the shape and appearance features of the first and second sample measured images.

[0051] In some embodiments of this disclosure, the data merging generator can be trained through the following steps:

[0052] Step a1: Input the original sample image of the wind turbine into the generator to be trained to obtain the fused image output by the generator after fusing the original sample image.

[0053] In one embodiment, the AAE (Adaptive Arithmetic Encoder) algorithm trains this data merging generator based on a train-test split dataset. The generator uses a method that extracts different dimensional features (shape features and style features) from two images, then merges the two dimensions from the different images to obtain a new image, where image x provides the shape features of the generated image, and image y provides the style features of the generated image.

[0054] Step a2: Calculate the loss value between the fused image and the original sample image using the following comprehensive loss function:

[0055]

[0056] Among them, L total Let D be the discriminator of the GANs, G be the generator of the GANs, and k be the loss function. R and k FM For hyperparameters, This is the loss function for GANs, used to drive the competition between the generator and the discriminator. This is the loss function for the shape feature extractor, used to calculate the loss value between the shape features and their corresponding original sample images. This is the loss function for the appearance feature extractor, used to determine the loss value between the appearance feature and its corresponding original sample image.

[0057] It should be noted that, in the embodiments of this disclosure, when determining the loss value between the fused image and the original sample image, not only is the competition between the GANs algorithm generator and the discriminator considered, that is, not only is the loss function of GANs calculated, but also the loss between the appearance features in the fused image and the corresponding original sample image, as well as the loss between the shape features and the corresponding original sample image, which effectively improves the accuracy of the loss value and thus improves the training efficiency of the data merging generator.

[0058] In some embodiments of this disclosure, It is calculated using the following formula:

[0059]

[0060]

[0061]

[0062] Where, Φ e y is the feature extractor, x is the original sample image corresponding to the extracted shape features, and y is the original sample image corresponding to the extracted appearance features.

[0063] Step a3: If the loss value is greater than the preset threshold, the generator is adjusted according to the loss value, and the process returns to the step of inputting the original sample image of the wind turbine into the generator to be trained, until the loss value is less than the preset threshold.

[0064] Step 203: Use the data merging generator to perform M fusion processes on the shape features and appearance features of the measured images of different samples to obtain M first wind turbine images.

[0065] Where M is an integer greater than 1.

[0066] Understandably, the data merging generator can obtain a different wind turbine image each time it performs the fusion process. By performing M fusion processes, M first wind turbine images can be obtained.

[0067] Step 204: Select the first wind turbine image with the smallest difference from the original sample image from the M first wind turbine images to obtain the target wind turbine image.

[0068] Understandably, in order to ensure that the difference between the final generated sample and the original sample image is as small as possible, the first wind turbine image with the smallest difference from the original sample image can be selected from the M first wind turbine images to obtain the target wind turbine image, which can then be used for further fusion with the original sample image to obtain a wind turbine image with even smaller difference from the original sample image.

[0069] In some embodiments of this disclosure, step 204 may specifically include the following steps:

[0070] For each of the M images of the first wind turbine, the divergence fraction of the first wind turbine image is calculated using the following formula:

[0071]

[0072] in, Here, is the divergence score, KL is the relative entropy, also known as the Kullback-Leibler divergence, used to measure the difference between one probability distribution and another reference probability distribution. x is the original sample image corresponding to the shape features of the first wind turbine image, y is the original sample image corresponding to the appearance features, G is the generator of the Generative Adversarial Network (GAN), and h... dΦ is a function for computing the hidden representation of the generated image. f For feature extractors;

[0073] The target wind turbine image is obtained by selecting the wind turbine image with the smallest divergence score from among the M wind turbine images.

[0074] In one embodiment, a neighbor selector can be used to select the sample with the smallest class difference (i.e., the lowest divergence score) between the generated and original sample images from the candidate sample pool (i.e., support set) of M first wind turbine images for subsequent generation.

[0075] In other embodiments of this disclosure, the above divergence fraction can also be calculated using JS (Jensen-Shannon) divergence.

[0076] Step 205: Use a data merging generator to perform N fusion processes on the target wind turbine image and the original sample image to obtain N second wind turbine images.

[0077] Where N is an integer greater than 1.

[0078] In one embodiment, a data merging generator is used to merge the target wind turbine image, which has a low degree of difference from the original sample image, with the original sample image N times to obtain N second wind turbine images.

[0079] Understandably, the second wind turbine image obtained by the fusion process in step 205 has a lower degree of similarity to the original sample image than the target wind turbine image.

[0080] In some embodiments of this disclosure, step 205 may specifically include the following steps:

[0081] Obtain the shape features of the target wind turbine image, as well as the appearance features from the original sample image;

[0082] A data merging generator is used to perform N fusion processes on the shape features of the target wind turbine image and the appearance features in the original sample image to obtain N second wind turbine images.

[0083] Understandably, since the deformation recognition model needs to accurately identify the deformation state of the wind turbine, it is advisable to allow the second wind turbine image to differ more in shape features from the original sample image. In other words, feature enhancement processing is performed on the shape features in the newly generated second wind turbine image, improving its robustness and thus enhancing the performance of the prototype network model.

[0084] Step 206: Determine the average feature vector based on N second wind turbine images, and store the average feature vector in the feature vector set of the deformation recognition model.

[0085] It should be noted that since the deformation recognition model can be updated by storing the above average feature vector into the feature vector set of the deformation recognition model, this scheme can be used as a plug-and-play model optimization method that can be adapted to all pre-trained few-shot learning models that can directly obtain sample embedding representation information (i.e., feature vectors), and has good portability and generalization.

[0086] In some embodiments of this disclosure, determining the average feature vector based on N second wind turbine images may specifically include the following steps:

[0087] Generate feature vectors for each of the N second wind turbine images, resulting in N feature vectors;

[0088] Calculate the average of the N eigenvectors to obtain the average eigenvector.

[0089] It is understandable that by calculating the average of the feature vectors of N second wind turbine images, the average feature vector can be obtained, which can correct the bias of the few-sample learning FSL model, that is, correct the bias between the second wind turbine image and the expected wind turbine image, and reduce the misjudgment rate of the deformation recognition model.

[0090] According to the wind turbine deformation recognition method proposed in this disclosure, measured images of the wind turbine are acquired. These measured images include 3D point maps and vibration signal maps. The 3D point maps and vibration signal maps are input into a pre-trained deformation recognition model. The model then identifies the deformation state of the wind turbine based on the 3D point maps and vibration signal maps using a preset feature vector set. The feature vectors in the feature vector set are obtained based on the shape and appearance features of the original sample images. The method determines whether the wind turbine has deformed based on the recognition results. By generating more wind turbine images with deformation using shape and appearance features from different images, the feature vector set of the prototype network model is enriched, thereby improving the accuracy of the prototype network model in recognizing the wind turbine deformation state. Furthermore, since the above scheme only requires operation on the original sample images during the testing phase, no additional dataset is needed during training, which not only reduces the training computation cost but also reduces the difficulty of collecting sample database data.

[0091] Figure 3 This is a block diagram illustrating a wind turbine deformation identification device according to an exemplary embodiment. (Refer to...) Figure 3 The device includes a first acquisition unit 301, an input unit 302, a second acquisition unit 303, and a determination unit 304.

[0092] The first acquisition unit 301 is used to acquire measured images of the wind turbine; the types of measured images include three-dimensional point plots and vibration signal plots.

[0093] Input unit 302 is used to input the three-dimensional point map and vibration signal map into the pre-trained deformation recognition model;

[0094] The second acquisition unit 303 is used to acquire the recognition result output by the deformation recognition model after recognizing the deformation state of the wind turbine based on the three-dimensional point map and vibration signal map according to the preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of the different original sample images.

[0095] The determining unit 304 is used to determine whether the fan has deformed based on the identification result.

[0096] In some embodiments of this disclosure, the apparatus may further include a first training unit, which is used to train a deformation recognition model. Specifically, the first training unit may be used for:

[0097] Obtain the original sample images of the wind turbine; the original sample images include the first sample measured image and the second sample measured image;

[0098] A pre-trained data merging generator is used to extract the shape and appearance features of the first and second sample measured images, respectively.

[0099] The shape and appearance features of the measured images of different samples are fused M times using a data merging generator to obtain M first wind turbine images; where M is an integer greater than 1.

[0100] The target wind turbine image is obtained by selecting the first wind turbine image with the smallest difference from the original sample image from the M first wind turbine images;

[0101] A data merging generator is used to perform N fusion processes on the target wind turbine image and the original sample image to obtain N second wind turbine images; where N is an integer greater than 1.

[0102] The average feature vector is determined based on N images of the second wind turbine, and then stored in the feature vector set of the deformation recognition model.

[0103] In some embodiments of this disclosure, the first training unit may further be used for:

[0104] For each of the M images of the first wind turbine, the divergence fraction of the first wind turbine image is calculated using the following formula:

[0105]

[0106] in, Let be the divergence score, KL be the relative entropy, x be the original sample image corresponding to the shape feature of the first wind turbine image, y be the original sample image corresponding to the appearance feature, G be the generator of the Generative Adversarial Network (GAN), and h be the vector. d Φ is a function for computing the hidden representation of the generated image. f For feature extractors;

[0107] The target wind turbine image is obtained by selecting the wind turbine image with the smallest divergence score from among the M wind turbine images.

[0108] In some embodiments of this disclosure, the first training unit may further be used for:

[0109] Obtain the shape features of the target wind turbine image, as well as the appearance features from the original sample image;

[0110] The shape features of the target wind turbine image and the appearance features of the original sample image are fused N times to obtain N second wind turbine images.

[0111] In some embodiments of this disclosure, the first training unit may further be used for:

[0112] Generate feature vectors for each of the N second wind turbine images, resulting in N feature vectors;

[0113] Calculate the average of the N eigenvectors to obtain the average eigenvector.

[0114] In some embodiments of this disclosure, the apparatus further includes a second training unit, which may specifically be used for:

[0115] The original sample image of the wind turbine is input into the generator to be trained, and the fused image output by the generator after fusing the original sample image is obtained.

[0116] The loss between the fused image and the original sample image is calculated using the following comprehensive loss function:

[0117]

[0118]

[0119] Among them, L total Let D be the discriminator of the GANs, G be the generator of the GANs, and k be the loss function. E and k FM For hyperparameters, The loss function for GANs, This is the loss function for the shape feature extractor, used to calculate the loss value between the shape features and their corresponding original sample images. This is the loss function for the appearance feature extractor, used to determine the loss value between the appearance feature and its corresponding original sample image;

[0120] If the loss value is greater than the preset threshold, the generator is adjusted according to the loss value, and the process returns to the step of inputting the original sample image of the wind turbine into the generator to be trained, until the loss value is less than the preset threshold.

[0121] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0122] According to the wind turbine deformation recognition device proposed in this embodiment, the device acquires measured images of the wind turbine; the types of measured images include three-dimensional point maps and vibration signal maps; the three-dimensional point maps and vibration signal maps are input into a pre-trained deformation recognition model; the deformation recognition model obtains the recognition result output by recognizing the deformation state of the wind turbine based on the three-dimensional point maps and vibration signal maps according to a preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of different original sample images; and the device determines whether the wind turbine has deformed based on the recognition result. By generating more wind turbine images with deformation by using shape features and appearance features from different images, the feature vector set of the prototype network model is enriched, thereby improving the accuracy of the prototype network model in recognizing the deformation state of the wind turbine.

[0123] Figure 4 This is a block diagram illustrating an apparatus for a wind turbine deformation recognition method according to an exemplary embodiment. For example, apparatus 400 may be an electronic device, such as a mobile phone, computer, digital broadcasting terminal, messaging device, game console, tablet device, medical device, fitness equipment, personal digital assistant, etc.

[0124] Reference Figure 4 The device 400 may include one or more of the following components: a processing component 402, a memory 404, a power component 406, a multimedia component 408, an audio component 410, an input / output (I / O) interface 412, a sensor component 414, and a communication component 416.

[0125] Processing component 402 typically controls the overall operation of device 400, such as operations associated with display, telephone calls, data communication, camera operation, and recording. Processing component 402 may include one or more processors 420 to execute instructions to perform all or part of the steps of the methods described above. Furthermore, processing component 402 may include one or more modules to facilitate interaction between processing component 402 and other components. For example, processing component 402 may include a multimedia module to facilitate interaction between multimedia component 408 and processing component 402.

[0126] Memory 404 is configured to store various types of data to support the operation of device 400. Examples of this data include instructions for any application or method operating on device 400, contact data, phonebook data, messages, pictures, videos, etc. Memory 404 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0127] The power supply component 406 provides power to the various components of the device 400. The power supply component 406 may include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power to the device 400.

[0128] Multimedia component 408 includes a screen that provides an output interface between the device 400 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors may sense not only the boundaries of the touch or swipe action but also the duration and pressure associated with the touch or swipe operation. In some embodiments, multimedia component 408 includes a front-facing camera and / or a rear-facing camera. When the device 400 is in an operating mode, such as a shooting mode or a video mode, the front-facing camera and / or the rear-facing camera may receive external multimedia data. Each front-facing camera and rear-facing camera may be a fixed optical lens system or have focal length and optical zoom capabilities.

[0129] Audio component 410 is configured to output and / or input audio signals. For example, audio component 410 includes a microphone (MIC) configured to receive external audio signals when device 400 is in an operating mode, such as call mode, recording mode, and voice recognition mode. The received audio signals may be further stored in memory 404 or transmitted via communication component 416. In some embodiments, audio component 410 also includes a speaker for outputting audio signals.

[0130] I / O interface 412 provides an interface between processing component 402 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include, but are not limited to, home buttons, volume buttons, power buttons, and lock buttons.

[0131] Sensor assembly 414 includes one or more sensors for providing status assessments of various aspects of device 400. For example, sensor assembly 414 may detect the on / off state of device 400, the relative positioning of components such as the display and keypad of device 400, changes in the position of device 400 or a component of device 400, the presence or absence of user contact with device 400, the orientation or acceleration / deceleration of device 400, and temperature changes of device 400. Sensor assembly 414 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. Sensor assembly 414 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, sensor assembly 414 may also include an accelerometer, a gyroscope, a magnetometer, a pressure sensor, or a temperature sensor.

[0132] Communication component 416 is configured to facilitate wired or wireless communication between device 400 and other devices. Device 400 can access wireless networks based on communication standards, such as WiFi, 2G, or 3G, or combinations thereof. In one exemplary embodiment, communication component 416 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, communication component 416 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0133] In an exemplary embodiment, the apparatus 400 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described above.

[0134] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 404 including instructions, which can be executed by a processor 420 of the device 400 to perform the above-described method. For example, the non-transitory computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device, etc.

[0135] In an exemplary embodiment, a computer program product is also provided, including a computer program that implements the above-described method when executed by the processor 420 of the device 400.

[0136] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of the invention are indicated by the following claims.

[0137] It should be understood that the present invention is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for identifying wind turbine deformation, characterized in that, include: Acquire measured images of the wind turbine; the types of measured images include three-dimensional point plots and vibration signal plots; The three-dimensional point map and the vibration signal map are input into the pre-trained deformation recognition model; The deformation recognition model obtains the recognition result after identifying the deformation state of the wind turbine based on the three-dimensional point map and the vibration signal map according to the preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of the different original sample images. Based on the identification results, determine whether the fan has undergone deformation; The feature vector set of the deformation recognition model is obtained through the following steps: Obtain the original sample images of the wind turbine; the original sample images include the first sample measured image and the second sample measured image; A pre-trained data merging generator is used to extract the shape and appearance features of the first sample measured image and the second sample measured image, respectively. The shape and appearance features of the different sample measured images are fused M times using the data merging generator to obtain M first wind turbine images; where M is an integer greater than 1. The target wind turbine image is obtained by selecting the first wind turbine image with the smallest difference from the original sample image from the M first wind turbine images. The data merging generator is used to perform N fusion processes on the target wind turbine image and the original sample image to obtain N second wind turbine images; where N is an integer greater than 1. The average feature vector is determined based on the N second wind turbine images, and the average feature vector is stored in the feature vector set of the deformation recognition model.

2. The wind turbine deformation identification method according to claim 1, characterized in that, The target wind turbine image is obtained by selecting the first wind turbine image with the smallest difference from the original sample image from the M first wind turbine images, including: For each of the M first wind turbine images, the divergence fraction of the first wind turbine image is calculated using the following formula: in, Let be the divergence score, KL be the relative entropy, x be the original sample image to which the shape feature corresponding to the first wind turbine image belongs, y be the original sample image to which the appearance feature belongs, and G be the generator of the Generative Adversarial Network (GAN). A function for computing the hidden representation of the generated image. For feature extractors; The target wind turbine image is obtained by selecting the wind turbine image with the smallest divergence score from the M wind turbine images.

3. The wind turbine deformation identification method according to claim 1, characterized in that, The data merging generator is used to perform N fusion processes on the target wind turbine image and the original sample image to obtain N second wind turbine images, including: Obtain the shape features of the target wind turbine image and the appearance features of the original sample image; The shape features of the target wind turbine image and the appearance features in the original sample image are fused N times to obtain the N second wind turbine images.

4. The wind turbine deformation identification method according to claim 1, characterized in that, The step of determining the average feature vector based on the N second wind turbine images includes: Each of the N second wind turbine images is generated with its own feature vector, resulting in N feature vectors; Calculate the average value of the N feature vectors to obtain the average feature vector.

5. The wind turbine deformation identification method according to claim 1, characterized in that, The data merging generator is trained through the following steps: The original sample image of the wind turbine is input into the generator to be trained, and the fused image output by the generator after fusing the original sample image is obtained. The loss value between the fused image and the original sample image is calculated using the following comprehensive loss function: in, Let D be the comprehensive loss function, D be the discriminator of the GANs, and G be the generator of the GANs. and For hyperparameters, The loss function for GANs, This is the loss function for the shape feature extractor, used to calculate the loss value between the shape features and their corresponding original sample images. This is the loss function for the appearance feature extractor, used to determine the loss value between the appearance feature and its corresponding original sample image; If the loss value is greater than a preset threshold, the generator is adjusted according to the loss value, and the process returns to the step of inputting the original sample image of the wind turbine into the generator to be trained, until the loss value is less than the preset threshold.

6. A fan deformation identification device, characterized in that, include: The first acquisition unit is used to acquire measured images of the wind turbine; the types of measured images include three-dimensional point plots and vibration signal plots. The input unit is used to input the three-dimensional point map and the vibration signal map into the pre-trained deformation recognition model; The second acquisition unit is used to acquire the recognition result output by the deformation recognition model after recognizing the deformation state of the wind turbine based on the three-dimensional point map and the vibration signal map according to the preset feature vector set; the feature vectors in the feature vector set are obtained based on the shape features and appearance features of the different original sample images. A determining unit is used to determine whether the fan has deformed based on the identification result; The second acquisition unit is also used to acquire the original sample image of the wind turbine; the original sample image includes the first sample measured image and the second sample measured image; A pre-trained data merging generator is used to extract the shape and appearance features of the first sample measured image and the second sample measured image, respectively. The shape and appearance features of the different sample measured images are fused M times using the data merging generator to obtain M first wind turbine images; where M is an integer greater than 1. The target wind turbine image is obtained by selecting the first wind turbine image with the smallest difference from the original sample image from the M first wind turbine images. The data merging generator is used to perform N fusion processes on the target wind turbine image and the original sample image to obtain N second wind turbine images; where N is an integer greater than 1. The average feature vector is determined based on the N second wind turbine images, and the average feature vector is stored in the feature vector set of the deformation recognition model.

7. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, The computer program, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images