Fan blade damage identification system and fan

Abstract

The invention provides a draught fan blade damage identification system and a draught fan, and belongs to the technical field of draught fans, and the draught fan blade damage identification system comprises a sound array module which comprises a plurality of sound sensors arranged in an array and is configured to collect sound data when draught fan blades operate; the image acquisition module is configured to acquire image data of the fan blade when the fan blade runs to a preset position; the processing module is respectively connected with the sound array module and the image acquisition module, the processing module comprises a processor end, and the processor end is provided with a main core and an auxiliary core; wherein the main core is configured to perform abnormity identification on the sound data, and send an enable signal to the auxiliary core through the shared memory under the condition that the sound data is abnormal; and the auxiliary core is configured to respond to the enable signal and identify the damage type of the fan blade based on the image data. According to the fan blade damage identification system provided by the embodiment of the invention, the damage of the fan blade can be accurately identified in the running state of the fan.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine technology, specifically relating to a wind turbine blade damage identification system and a wind turbine. Background Technology

[0002] Currently, damage identification of wind turbine blades typically employs acoustic monitoring and visual inspection methods. However, the noise generated by wind turbine blades is greatly affected by environmental factors such as wind speed, temperature, and external noise, which can mask the acoustic signals generated by blade damage and affect the accuracy of damage identification. Visual inspection methods usually rely on drones for flight monitoring, which requires the wind turbine blades to be stopped and is affected by factors such as wind speed, making real-time detection difficult. Therefore, identifying wind turbine blade damage while ensuring that the blades are in operation is an urgent problem to be solved. Summary of the Invention

[0003] In view of the above problems, this application provides a wind turbine blade damage identification system and a wind turbine to overcome or at least partially solve the above problems.

[0004] A first aspect of this application provides a wind turbine blade damage identification system, comprising: The acoustic array module includes multiple arrayed acoustic sensors configured to collect sound data during the operation of the wind turbine blades. The image acquisition module is configured to acquire image data of the wind turbine blades when the wind turbine blades move to a preset position; The processing module is connected to both the acoustic array module and the image acquisition module. The processing module includes a processor with a main core and a sub-core. The main core is configured to identify anomalies in the sound data and, in the event of an anomaly in the sound data, send an enable signal to the sub-core via shared memory. The sub-core is configured to respond to the enable signal and identify the damage type of the wind turbine blade based on the image data.

[0005] Furthermore, the processing module includes a programmable logic terminal, which is equipped with a first acoustic signal processing unit and a camera driving unit; wherein, The first sound signal processing unit is configured to determine the position of the wind turbine blades in real time based on the sound data, and send a pulse trigger signal to the camera driving unit when the position of the wind turbine blades is at the preset position; The camera driving unit is configured to drive the image acquisition module to acquire the surface image of the wind turbine blade in response to the pulse trigger signal.

[0006] Furthermore, the main core is equipped with a second acoustic signal processing unit, and the sub-core is equipped with an image processing unit; wherein, The second sound signal processing unit is configured to, in response to the wind turbine blades running to the preset position, perform a first classification on the sound data based on a pre-trained sound model, determine whether there is an abnormality in the wind turbine blades based on the first classification result, and send the enable signal to the image processing unit if there is an abnormality. The image processing unit is configured to, in response to the enable signal, perform a second classification on the image data based on a pre-trained image model, and determine the damage type of the wind turbine blades based on the second classification result; The sound model is obtained by deep learning on a first preset model based on multiple sound data samples. The multiple sound data samples include sound data samples when the wind turbine blades are abnormal and sound data samples when the wind turbine blades are operating normally. The image model is obtained by deep learning on a second preset model based on multiple wind turbine blade image samples. The multiple wind turbine blade image samples include wind turbine blade image samples when the wind turbine blades have different types of damage, and wind turbine blade image samples when the wind turbine blades do not have damage.

[0007] Furthermore, the image model is configured with image feature parameters, and the image processing unit uses the pre-trained image model to perform a second classification of the image data, including: The texture feature parameters of the image data are calculated using the gray-level co-occurrence matrix; The second classification result of the image data is obtained by comparing the texture feature parameters with the image feature parameters of the image model.

[0008] Furthermore, the processor module also includes a storage unit, which is communicatively connected to the first acoustic signal processing unit, the camera driving unit, the second acoustic signal processing unit, and the image processing unit, respectively; wherein, The first acoustic signal processing unit is further configured to preprocess the acoustic data collected by the acoustic array module and store the preprocessed acoustic data in the storage unit; The camera driving unit is further configured to store the image data acquired by the image acquisition module into the storage unit; The second sound signal processing unit is further configured to extract the sound data corresponding to the wind turbine blade running to a preset position from the storage unit when performing the first classification on the sound data, so as to perform the first classification on the sound data corresponding to the preset position; The image processing unit is further configured to extract image data to be classified into the second category from the storage unit in response to the enable signal.

[0009] Furthermore, the storage unit includes a first storage space and a second storage space; wherein the first storage space is configured to store the sound data, and the second storage space is configured to store the image data; The first storage space and the second storage space are independent of each other.

[0010] Furthermore, the main core is configured with an asymmetric multiprocessing mode; The main core is further configured to, after startup, activate the secondary core via the shared memory based on the asymmetric multiprocessing mode, so that the secondary core and the main core can execute different tasks independently. After being started, the main core responds to the pulse trigger signal and extracts the sound data from the storage unit for anomaly identification. After being activated, the sub-core, in response to the enable signal, extracts the image data from the storage unit to identify the damage type.

[0011] Furthermore, the first acoustic signal processing unit is also configured to send the pulse trigger signal to the image processing unit when the wind turbine blades run to the preset position; The image processing unit is further configured to respond to the pulse trigger signal to determine whether there is an abnormality in the wind turbine blades. If there is an abnormality and the second acoustic signal processing unit determines that the blades are also abnormal, the image data is classified in a second way based on the pre-trained image model.

[0012] Furthermore, the system also includes an alarm module, which is connected to the processing module and configured to send an alarm signal and damage information when the wind turbine blades are identified as being damaged.

[0013] Furthermore, the processing module is connected to the acoustic array module and the image acquisition module on multiple wind turbines; The main core is also configured to detect the memory levels of the main core and the sub-core in real time, and dynamically adjust the task allocation of the main core and the sub-core through the shared memory when the memory levels of the main core or the sub-core are insufficient. The task allocation is used to assign anomaly recognition tasks for the sound data and recognition tasks for the image data.

[0014] In a second aspect of this application, a wind turbine is provided, including a tower, wind turbine blades, and the wind turbine blade damage identification system described in the first aspect of this application.

[0015] Furthermore, it also includes a medium substrate located on the tower. The medium substrate is on which the acoustic array module and image acquisition module of the wind turbine blade damage identification system are arranged, and the installation height of the medium substrate is higher than the lowest point of the projection of the tip of the wind turbine blade on the tower.

[0016] Furthermore, the medium substrate is cubic in shape, and the acoustic array module and the image acquisition module are located on the same surface of the medium substrate.

[0017] Furthermore, the camera in the image acquisition module has a wide-angle lens with a field of view greater than or equal to 60 degrees.

[0018] The wind turbine blade damage identification system and wind turbine provided in this embodiment can first collect sound data of the wind turbine blade running time in real time by forming an acoustic array module composed of multiple arrayed acoustic sensors. Since the processing module is connected to the acoustic array module and the image acquisition module respectively, and the processing module includes a processor, the processor is equipped with a main core and a sub-core, and the main core is configured to identify anomalies in the sound data. When the sound data is abnormal, an enable signal is sent to the sub-core through shared memory. By working together with multiple acoustic sensors, the interference of environmental noise can be effectively reduced, and the acquisition quality of the sound signal and the accuracy of damage identification can be improved.

[0019] Secondly, the image acquisition module collects image data of the wind turbine blades when they are running in a preset position, enabling non-stop detection. This avoids wind power resource loss due to downtime detection, improves detection efficiency, and, by responding to the enable signal through the sub-core, identifies the damage type of the wind turbine blades based on the image data. This further verifies that the wind turbine blades have been damaged, avoiding wind turbine blade failures caused by incorrect damage identification. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of a wind turbine blade damage identification system provided in an embodiment of this application; Figure 2 This is a schematic diagram of another wind turbine blade damage identification system provided in the embodiments of this application; Figure 3 This is a communication schematic diagram of a wind turbine blade damage identification system provided in an embodiment of this application; Figure 4 This is a front view schematic diagram of a fan provided in an embodiment of this application: Figure 5 This is a side view schematic diagram of a fan provided in an embodiment of this application; Figure label: 1-Wind turbine blade; 2-Tower; 3-Medium substrate; 4-Sound sensor; 5-Image acquisition module. Detailed Implementation

[0022] Exemplary embodiments of this application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of this application are shown in the drawings, it should be understood that this application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of this application and to fully convey the scope of this application to those skilled in the art.

[0023] In related technologies, acoustic monitoring solutions mainly monitor the health status of wind turbine blades by capturing sound wave signals generated during their operation. Wind turbine blades generate vibrations and noise under the action of wind. These sound signals can be captured and analyzed by sensors to determine whether there are structural problems such as cracks or delamination in the blades.

[0024] While visual inspection solutions can quickly and remotely acquire images of blades, reducing the safety hazards and workload of manual inspections and avoiding the risks of high-altitude operations, they require manual operation of drones for flight monitoring, which places high demands on human operators.

[0025] In view of this, this embodiment provides a wind turbine blade damage identification system and a wind turbine. The system collects sound data of the wind turbine blades in real time through an acoustic array module and image data of the wind turbine blades through an image acquisition module. The main core in the processing module performs anomaly identification on the sound data. In case of anomalies, the secondary core in the processing module determines the type of damage to the wind turbine blades based on the image information, so that the staff can determine whether to stop the operation of the wind turbine blades.

[0026] Reference Figure 1 , Figure 1 This is a schematic diagram of a wind turbine blade damage identification system provided in an embodiment of this application. Figure 1 From this, we can know that Figure 1 include: An acoustic array module includes multiple arrayed acoustic sensors configured to collect acoustic data during wind turbine blade operation; an image acquisition module is configured to acquire image data of the wind turbine blade when it reaches a preset position; a processing module is connected to both the acoustic array module and the image acquisition module, and includes a processor with a main core and a sub-core deployed thereon; wherein the main core is configured to perform anomaly identification on the acoustic data, and in the event of anomalies in the acoustic data, send an enable signal to the sub-core via shared memory; the sub-core is configured to respond to the enable signal and identify the damage type of the wind turbine blade based on the image data.

[0027] In this embodiment, since the wind turbine blades operate in a natural environment, the sources of sound in the natural environment are quite complex. Furthermore, the wind turbine blades themselves generate considerable noise during operation. Therefore, by using multiple arrayed sound sensors to form a sound array module to collect sound data during the operation of the wind turbine blades, the wind noise of the wind turbine blades themselves and the interference of other noises in the natural environment can be reduced by fusing the sound data collected by multiple sensors.

[0028] In some examples, the array arrangement of the sound sensors can be rectangular, circular, linear, or spherical, as long as the sound sensors can accurately and comprehensively collect the sound signals generated by the wind turbine blades during operation.

[0029] The image acquisition module is configured to acquire image data of the wind turbine blades when the blades reach a preset position. In this embodiment, the image acquisition module can be a high-definition camera or an infrared image acquisition device or other electronic device capable of acquiring images. The preset position is the position where the image acquisition device can successfully acquire images of the wind turbine blades. Since the wind turbine blades are constantly rotating during operation, and a wind turbine includes multiple blades, in order to ensure that the operation of the wind turbine blades is not affected, the image acquisition module can acquire image data of each wind turbine blade. The image acquisition module is generally arranged on other components of the wind turbine besides the wind turbine blades. Therefore, in order to ensure the effectiveness and reliability of the acquired image data of the wind turbine blades, the image acquisition module can acquire image data of the wind turbine blades only when the blades reach the preset position to avoid acquiring invalid image data.

[0030] The processing module is an electronic device used to process audio and image data. It includes a processor, which can be understood as the hardware component of the processing module. The processor has a main core and a secondary core. The main core is configured to identify anomalies in the audio data and, in the event of an anomaly, sends an enable signal to the secondary core via shared memory. This shared memory can be OCM (On-Chip Memory). On-chip memory enables lower access latency and increases response speed. Therefore, it ensures that the secondary core responds to the enable signal more quickly, enabling the identification of wind turbine blade damage types based on image data.

[0031] The main core performs anomaly detection on the sound data as follows: First, it extracts features from the sound data, specifically using methods such as Mel-frequency cepstral coefficients (MFCCs) and Gammatone cepstral coefficients (GTCCs). The extracted features effectively characterize the spectral properties of the sound signal and are robust to environmental noise. Then, a pre-trained sound model is used to analyze the extracted features to determine if the sound data is abnormal. The sound model can include autoencoders, convolutional neural networks, etc. For example, an autoencoder learns the feature representation of normal sound data. When the features of the input sound data differ significantly from those of normal data, the model's reconstruction error increases significantly, thus indicating an anomaly. When the main core detects an anomaly in the sound data, it sends an enable signal to the sub-core via shared memory, notifying the sub-core to initiate the image data damage detection process.

[0032] The sub-core performs damage type identification on image data as follows: After receiving the enable signal, the sub-core first preprocesses the image data, including grayscale conversion, noise reduction, and edge enhancement, to improve image quality and facilitate subsequent damage identification. Then, the sub-core uses deep learning algorithms to perform target detection and segmentation on the blades in the image, extracting the blade regions. After extracting the blade regions, the sub-core further extracts features from the blade image to identify the damage type. Feature extraction methods include Gray-Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP), which can characterize the texture and shape information of the blade surface. Finally, the sub-core inputs the extracted damage features into a pre-trained image model to classify and identify the damage type. The image model can be a Support Vector Machine (SVM), Convolutional Neural Network (CNN), etc. Then, the output of the image model can determine the blade damage type, which can include cracks, surface damage, corrosion, etc.

[0033] Therefore, the wind turbine blade damage identification system provided in this embodiment can identify wind turbine blade damage by using a sound array module composed of multiple arrayed sound sensors to collect sound data of the wind turbine blade running time in real time. The image acquisition module collects image data of the wind turbine blade when it is running at a preset position. The main core identifies anomalies in the sound data. When the sound data is abnormal, it sends an enable signal to the sub-core through shared memory. The sub-core responds to the enable signal and identifies the damage type of the wind turbine blade based on the image data. By utilizing the collaborative work of the main core and the sub-core, the system can quickly and accurately identify wind turbine blade damage, effectively improving the maintenance efficiency and safety of wind turbine blades.

[0034] In one specific embodiment, the processing module includes a programmable logic terminal, on which a first acoustic signal processing unit and a camera driving unit are deployed; wherein, the first acoustic signal processing unit is configured to determine the position of the wind turbine blade in real time based on the acoustic data, and send a pulse trigger signal to the camera driving unit when the position of the wind turbine blade is at the preset position; the camera driving unit is configured to drive the image acquisition module to acquire the surface image of the wind turbine blade in response to the pulse trigger signal.

[0035] In this embodiment, refer to Figure 2 , Figure 2 This is a schematic diagram of another wind turbine blade damage identification system provided in the embodiments of this application. Figure 2 As can be seen, the processing module includes a programmable logic unit (PLU), which can be understood as the software part of the processing module. The PLU is equipped with a first acoustic signal processing unit and a camera driver unit. Due to the complex operating environment of the wind turbine blades, the acquired sound data may contain a large amount of background noise, such as wind noise, thunder, and rain noise. To improve signal quality, the original sound data can be denoised using an acoustic signal preprocessing module. For example, using a Butterworth bandpass filter can effectively remove low-frequency wind noise and high-frequency electronic noise. Then, the first acoustic signal processing unit can determine the position of the blades based on the denoised sound data using a sound source localization algorithm. When the determined blade position approaches or reaches a preset position, the first acoustic signal processing unit sends a pulse trigger signal to the camera driver unit. The preset position refers to the optimal position where the image acquisition module can accurately capture the surface image of the wind turbine blade when it reaches a specific angle or position during rotation. The preset position can be set according to the rotation cycle of the wind turbine blades and the requirements of damage detection.

[0036] The camera driver unit responds to a pulse trigger signal. Upon receiving the pulse trigger signal, the camera driver unit initializes the image acquisition module. This includes setting the camera parameters in the image acquisition module, such as resolution, frame rate, and exposure time. For example, when using the Phase One IQ3 camera, its resolution needs to be set to 100 million pixels to ensure that even small defects on the blades can be captured. Simultaneously, the camera driver unit triggers the camera's shutter via the pulse trigger signal, completing the image capture of the wind turbine blade surface.

[0037] In one specific embodiment, the main core is equipped with a second acoustic signal processing unit, and the sub-core is equipped with an image processing unit. The second acoustic signal processing unit is configured to, in response to the wind turbine blade reaching the preset position, perform a first classification on the acoustic data based on a pre-trained acoustic model, determine whether the wind turbine blade exhibits an abnormality based on the first classification result, and, if an abnormality is found, send an enable signal to the image processing unit. The image processing unit is configured to, in response to the enable signal, perform a second classification on the image data based on a pre-trained image model, and determine the damage type of the wind turbine blade based on the second classification result. The acoustic model is obtained by deep learning on a first preset model based on multiple acoustic data samples, including acoustic data samples when the wind turbine blade exhibits an abnormality and acoustic data samples when the wind turbine blade is operating normally. The image model is obtained by deep learning on a second preset model based on multiple wind turbine blade image samples, including wind turbine blade image samples when the wind turbine blade exhibits different damage types and wind turbine blade image samples when the wind turbine blade is undamaged.

[0038] In this embodiment, refer to Figure 2 It is known that the main core is equipped with a second sound signal processing unit, and the sub-core is equipped with an image processing unit. Since the sound model is obtained by deep learning on the first preset model based on multiple sound data samples, the multiple sound data samples include sound data samples when the wind turbine blades are abnormal and sound data samples when the wind turbine blades are running normally. Therefore, when the wind turbine blades run to the preset position, the second sound signal processing unit performs the first classification of the sound data based on the pre-trained sound model. The first classification is the way to detect whether the wind turbine blades are abnormal.

[0039] The first classification specifically compares the sound data when the wind turbine blades move to a preset position with the sound data samples in the sound model. If the sound data matches the sound data samples when the wind turbine blades are normal, the wind turbine blades are determined to be normal. If the sound data matches the sound data samples when the wind turbine blades are abnormal, the wind turbine blades are determined to be abnormal. In the case of an abnormality, an enable signal is sent to the image processing unit.

[0040] Since the image model is obtained by deep learning on the second preset model based on multiple wind turbine blade image samples, the multiple wind turbine blade image samples include wind turbine blade image samples with different types of damage, as well as wind turbine blade image samples without damage. Therefore, the second classification is specifically achieved by the image processing unit responding to the enable signal. It can compare the image data samples existing in the pre-trained image model. If the image data matches the image data sample of the wind turbine blade crack damage type, it indicates that the wind turbine blade damage type is crack. If the image data matches the image data sample of the wind turbine blade normal, it indicates that the wind turbine blade is not damaged. It may be that the sound data is wrong or that the collected image data is wrong. Secondary verification can be performed as needed. For example, the image data and sound data of the next cycle can be collected and then verified according to the sound model and image model.

[0041] In one specific embodiment, the image model is configured with image feature parameters, and the image processing unit uses the pre-trained image model to perform a second classification of the image data, including: calculating the texture feature parameters of the image data through the gray-level co-occurrence matrix; and comparing the texture feature parameters with the image feature parameters of the image model to obtain a second classification result of the image data.

[0042] In this embodiment, since the image model is configured with image feature parameters, the image feature parameters include the feature parameters corresponding to the wind turbine blade image samples with different damage types and the normal wind turbine blade image samples. The feature parameters can include color features, texture features, shape features, etc. The image processing unit uses the pre-trained image model to perform a second classification on the image data by calculating the texture feature parameters of the image through the gray-level co-occurrence matrix. Specifically, different directions and distances of the image in the image data can be selected, and the gray-level co-occurrence matrix under each direction and distance can be calculated. Texture feature parameters such as contrast, second moment of angle (energy), correlation and inverse difference are extracted from the gray-level co-occurrence matrix. Texture feature parameters can reflect the texture characteristics of an image. For example, contrast reflects the clarity of the image and the depth of the grooves. Since the image model is based on multiple wind turbine blade image samples and is obtained by deep learning on the second preset model, the multiple wind turbine blade image samples include wind turbine blade image samples with different damage types and wind turbine blade image samples without damage. Therefore, the extracted texture feature parameters can be input into the image model and compared with the image feature parameters of the image model to determine the damage type of the wind turbine blade. Finally, the image model outputs the damage type of the wind turbine blade, such as cracks, surface damage, corrosion, etc.

[0043] In one specific embodiment, the processor module further includes a storage unit, which is communicatively connected to the first acoustic signal processing unit, the camera driving unit, the second acoustic signal processing unit, and the image processing unit. The first acoustic signal processing unit is further configured to preprocess the sound data acquired by the acoustic array module and store the preprocessed sound data in the storage unit. The camera driving unit is further configured to store the image data acquired by the image acquisition module in the storage unit. The second acoustic signal processing unit is further configured to, when performing the first classification on the sound data, extract the sound data corresponding to the wind turbine blades running to a preset position from the storage unit to perform the first classification on the sound data corresponding to the preset position. The image processing unit is further configured to, in response to the enable signal, extract the image data to be classified in the second classification from the storage unit.

[0044] In this embodiment, refer to Figure 2The processing module also includes a storage unit, which is communicatively connected to the first acoustic signal processing unit, the camera driver unit, the second acoustic signal processing unit, and the image processing unit. The first acoustic signal processing unit preprocesses the acoustic data collected by multiple arrayed acoustic sensors during wind turbine blade operation. The acoustic data includes normal operating sounds and abnormal sounds caused by damage. Preprocessing includes filtering, amplification, and noise reduction of the acoustic data. The preprocessed acoustic data is then stored in the storage unit. The camera driver unit also stores the image data collected by the image acquisition module in the storage unit.

[0045] When performing the first classification of sound data, the second sound signal processing unit directly extracts the sound data corresponding to the wind turbine blades running to a preset position from the storage unit, extracts feature parameters from the sound data, such as frequency, amplitude, and waveform, and performs the first classification on the sound data corresponding to the preset position. In response to the enable signal, the image processing unit also extracts the image data to be classified in the second classification from the storage unit, completing the second classification of the image data.

[0046] In one specific embodiment, the storage unit includes a first storage space and a second storage space; wherein the first storage space is configured to store the sound data, and the second storage space is configured to store the image data; wherein the first storage space and the second storage space are independent of each other.

[0047] In this embodiment, refer to Figure 2 The storage unit includes a first storage space and a second storage space. The first storage space is configured to store audio data, and the second storage space is configured to store image data. The first storage space and the second storage space are independent of each other, which can effectively isolate audio data and image data and prevent data interference and conflict.

[0048] In one specific embodiment, the main core is configured with an asymmetric multiprocessing mode; wherein, the main core is further configured to, upon startup, invoke the secondary core to start via the shared memory based on the asymmetric multiprocessing mode, so that the secondary core and the main core execute different tasks independently; after being started, the main core, in response to the pulse trigger signal, extracts the sound data from the storage unit for anomaly identification; after being started, the secondary core, in response to the enable signal, extracts the image data from the storage unit for damage type identification.

[0049] In this embodiment, the main core is configured in asymmetric multiprocessing mode. In asymmetric multiprocessing mode, the task allocation between processor cores is asymmetric, and each core can run different tasks independently.

[0050] Therefore, after the main core starts up, the secondary core is activated via shared memory based on an asymmetric multiprocessing model. The main core and secondary core can execute different tasks. By assigning specific tasks to the main core and secondary core, the asymmetric processing model can fully leverage the advantages of each core and improve resource utilization. The main core can focus on processing audio data, while the secondary core can focus on processing image data, thus achieving more efficient collaboration between the main core and secondary core. For example, when processing wind turbine blade damage identification, the main core can quickly identify anomalies in the audio data, while the secondary core can simultaneously perform detailed damage type identification on the image data.

[0051] In one specific embodiment, the first acoustic signal processing unit is further configured to send the pulse trigger signal to the image processing unit when the wind turbine blades run to the preset position; the image processing unit is further configured to, in response to the pulse trigger signal, determine whether there is an abnormality in the wind turbine blades, and if there is an abnormality and the second acoustic signal processing unit determines that the blades are also abnormal, perform a second classification on the image data based on a pre-trained image model.

[0052] In this embodiment, in order to ensure the reliability of blade damage identification, the first acoustic signal processing unit can also send a pulse signal to the image processing unit when the wind turbine blade runs to a preset position. The image processing unit can respond to the pulse signal to determine whether there is an abnormality in the wind turbine blade. Specifically, it compares the image data sample with the damaged image data sample in the image model. If the image processing unit determines that the wind turbine blade is abnormal and the second acoustic signal processing unit also determines that the wind turbine blade is abnormal, the image data is then classified in the second way according to the image model.

[0053] The first sound signal processing unit determines that the wind turbine blades have reached the preset position by analyzing the time-frequency characteristics of the sound signal, such as short-time Fourier transform (STFT) or wavelet transform. It can identify the specific sound characteristics when the wind turbine blades reach the preset position, or consider that the wind turbine blades have reached the preset position when the intensity of the sound signal at a specific frequency exceeds the preset threshold.

[0054] Machine learning algorithms, such as support vector machines or neural networks, can also be used to classify and identify sound signals to determine whether the wind turbine blades have reached the preset position. For example, during the training phase, sound signal samples when the wind turbine blades reach the preset position can be collected, and feature extraction and labeling of the sound signal samples can be performed to build a sound position model. During the operation of the wind turbine, the real-time collected sound signals are input into the trained sound position model, and the sound position model will output the judgment result of whether the blades have reached the preset position. Based on the judgment result, it can be determined whether the wind turbine blades have reached the preset position.

[0055] In one specific embodiment, the system further includes an alarm module connected to the processing module and configured to send an alarm signal and damage information when the wind turbine blades are identified as being damaged.

[0056] In this embodiment, refer to Figure 2 It also includes an alarm module, which is connected to the processing module. When the processing module determines that damage has been detected on the wind turbine blades, it will issue an alarm signal. Furthermore, the alarm module can also upload damage information of the damaged wind turbine blades based on image and sound data provided by the processing module, allowing staff to determine the location and severity of the damage. In addition, Figure 2 It may also include a power supply module, which can supply power to the alarm module, processing module, sound array module and image acquisition module.

[0057] In one specific embodiment, the processing module is connected to the acoustic array module and the image acquisition module on multiple wind turbines; wherein, the main core is further configured to detect the memory amount of the main core and the sub-core in real time, and dynamically adjust the task allocation of the main core and the sub-core through the shared memory when the memory amount of the main core or the sub-core is insufficient; wherein, the task allocation is used to allocate the anomaly recognition task of the sound data and the recognition task of the image data.

[0058] In this embodiment, the processing module is also connected to the acoustic array module and image acquisition module on multiple wind turbines, so that the processing module can simultaneously process the blade damage identification task of multiple wind turbines.

[0059] Since the main core and the secondary core process audio data and image data respectively, in order to ensure the efficiency of processing tasks, the main core can detect the memory amount of the main core and the secondary core in real time. When the memory amount of the main core or the secondary core is insufficient, the task allocation of the main core and the secondary core is dynamically adjusted through shared memory.

[0060] For example, if the main core is overloaded with tasks and cannot handle them, while the secondary core still has plenty of memory, the anomaly recognition task for the sound data that the main core needs to process can be partially distributed to the secondary core. When the sound data from the main core is distributed to the secondary core, the secondary core retrieves the sound data from the storage unit via shared memory and performs the anomaly recognition task. When an anomaly is detected in the sound data, the secondary core retrieves the image data corresponding to the sound data from the storage unit to perform wind turbine blade damage recognition. The main core does not need to send an enable signal to the secondary core via shared memory to start the secondary core to recognize the type of blade damage.

[0061] Correspondingly, when the secondary core's memory is insufficient, the image data recognition task of the secondary core can be allocated to the primary core for processing. When the primary core identifies anomalies in the sound data, it can first determine whether the corresponding image data is within the image data recognition task that the primary core needs to execute. If it is, the primary core retrieves the corresponding image data from the storage unit through the shared inner layer via the secondary core to perform wind turbine blade damage recognition. If it is not, the primary core sends an enable signal to the secondary core, and the secondary core retrieves the corresponding image data from the storage unit to perform wind turbine blade damage recognition. Through dynamic task allocation between the primary and secondary cores, it can be ensured that the anomaly recognition tasks for sound data and the recognition tasks for image data can be executed efficiently.

[0062] For example, the following will refer to Figure 2 and Figure 3 This embodiment provides a complete description of how the wind turbine blade damage identification system identifies wind turbine blade damage: Figure 3 This is a communication schematic diagram of a wind turbine blade damage identification system provided in an embodiment of this application. Figure 3 As can be seen from this, taking one cycle as an example, one rotation of one fan blade is one cycle. Figure 2 First, the acoustic array module detects the sound of the wind turbine blades as they rotate one revolution and transmits the collected sound data to the storage unit. The second acoustic signal processing unit then detects and processes the sound data in the storage unit in real time. When the wind turbine blade reaches the center position (the preset position), the acoustic signal processing module sends a trigger pulse signal to the image acquisition module. The image acquisition module then captures an image of the wind turbine blade's surface and stores the image data in the storage unit. The second acoustic signal processing unit then processes the sound data corresponding to the wind turbine blade reaching the center position to determine if the sound data is abnormal. This can be done by comparing it with samples of different sound data in a pre-trained sound model. If the sound data is determined to be abnormal, the damage type of the wind turbine blade needs to be further determined. The second acoustic signal processing unit sends an enable signal to the image processing unit. The image processing unit responds to the enable signal and processes the image data in the storage unit to determine the degree and type of damage to the wind turbine blade. This can also be done by comparing it with samples of different graphic data in a pre-trained image model to determine the damage type of the wind turbine blade.

[0063] When damage to the wind turbine blades is confirmed, the processing module can also send an alarm signal to the alarm module to remind the staff that the wind turbine blades are damaged. At the same time, the processing module can also send the type and extent of damage to the wind turbine blades to the alarm module so that the staff can determine whether the wind turbine needs to be stopped immediately for maintenance.

[0064] Therefore, the wind turbine blade damage identification system provided in this embodiment, when applied to wind turbine blade damage identification, can comprehensively understand the operating status of the wind turbine blades through dual detection of sound and image data, improving the accuracy of damage identification. Furthermore, the image acquisition module can acquire image data when the wind turbine blades reach a preset position, enabling non-stop detection and avoiding wind power resource loss due to shutdown detection. An alarm is only triggered when both the second sound signal processing unit and the image processing unit detect an anomaly, effectively reducing false alarms. This not only improves the efficiency and accuracy of wind turbine blade damage identification but also reduces the false alarm rate, further enhancing the adaptability and flexibility of wind turbine blade damage identification.

[0065] This application also provides a wind turbine, including a tower, wind turbine blades, and the wind turbine blade damage identification system described in the above embodiments.

[0066] In this embodiment, refer to Figure 4 , Figure 4 This is a front view schematic diagram of a fan provided in an embodiment of this application. From Figure 4 As can be seen from the diagram, the wind turbine includes a tower 2, wind turbine blades 1, and a wind turbine blade damage identification system.

[0067] In one specific embodiment, the system further includes a medium substrate 3 located on the tower. The medium substrate 3 is equipped with the acoustic array module and image acquisition module 5 of the wind turbine blade damage identification system, and the installation height of the medium substrate is higher than the lowest point of the projection of the tip of the wind turbine blade onto the tower.

[0068] In this embodiment, the wind turbine also includes a medium substrate 3, which is located on the tower 2. The installation height of the medium substrate 3 is higher than the lowest point of the projection of the tip of the wind turbine blade 1 onto the tower 2. This ensures that the detection equipment covers the entire operating range of the wind turbine blade 1, which is helpful for comprehensive detection of the wind turbine blade 1. Secondly, installing the monitoring equipment at a higher position can reduce ground reflection and interference from nearby objects, thereby improving the accuracy of sound data and image data acquisition.

[0069] In one specific embodiment, the medium substrate 3 is cubic, and the acoustic array module and the image acquisition module are located on the same surface of the medium substrate 3.

[0070] In this embodiment, placing the acoustic array module and the image acquisition module 5 on the same surface of the dielectric substrate ensures that they acquire data in a highly synchronized manner over time, reduces electromagnetic and signal interference between them, and improves the accuracy of data acquisition. For example, when detecting damage to the wind turbine blade 1, the abnormal sound can be more accurately correlated with the damage location in the image data, which helps reduce data deviations caused by positional differences between the acoustic array module and the image acquisition module 5, and improves the consistency and reliability of acoustic and image data acquisition.

[0071] Specific reference Figure 4 Multiple sound sensors 4 in the sound array module can be arranged in an array on the medium substrate 3, and the image acquisition module 5 can be placed at the center of the multiple sound sensors 4. In this way, the multiple sound sensors can be arranged in a ring array around the image acquisition module.

[0072] In one specific embodiment, the camera in the image acquisition module 5 is a wide-angle lens with a field of view greater than or equal to 60 degrees.

[0073] In this embodiment, to ensure detection accuracy, the range of the wind turbine blade images captured by the camera in the image acquisition module 5 is expanded, referring to... Figure 5 , Figure 5 This is a side view schematic diagram of a wind turbine provided in an embodiment of this application. A camera is set to have a shooting angle θ, which refers to the angle of the camera lens relative to the object being photographed. The shooting angle determines the direction and angle from which the camera captures the image. The field of view angle refers to the range of vision that the camera lens can capture. The field of view angle determines the size of the scene that the camera can see. To ensure that the camera's field of view angle captures image data of the tip of the wind turbine blade 1, and further acquires images of the entire wind turbine blade 1 over a larger area, the field of view angle φ needs to be considered when selecting the camera. To improve the monitoring effect, the blade length is L, the distance between the blade and the tower is S, the height of the blade's rotating shaft is H, and the height of the camera's placement position from the ground is h. Therefore, the height difference between the tip of the wind turbine blade 1 and the center position of the camera is L-(Hh). From the distance S between the wind turbine blade 1 and the camera and the field of view angle φ, it can be known that the shooting angle θ=φ / 2-arctan[(L+hH) / S]. Therefore, a wide-angle lens with a field of view angle φ greater than or equal to 60 degrees can be selected.

[0074] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0075] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods and apparatus according to embodiments of the present invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing module of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processing module of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0076] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0078] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0079] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0080] The present invention provides a detailed description of a wind turbine blade damage identification system and a wind turbine. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A wind turbine blade damage identification system, characterized in that, include: The acoustic array module includes multiple arrayed acoustic sensors configured to collect sound data during the operation of the wind turbine blades. The image acquisition module is configured to acquire image data of the wind turbine blades when the wind turbine blades move to a preset position; The processing module is connected to both the acoustic array module and the image acquisition module. The processing module includes a processor with a main core and a sub-core. The main core is configured to identify anomalies in the sound data and, in the event of an anomaly in the sound data, send an enable signal to the sub-core via shared memory. The sub-core is configured to respond to the enable signal and identify the damage type of the wind turbine blade based on the image data.

2. The system according to claim 1, characterized in that, The processing module includes a programmable logic terminal, which is equipped with a first acoustic signal processing unit and a camera driving unit; wherein... The first sound signal processing unit is configured to determine the position of the wind turbine blades in real time based on the sound data, and send a pulse trigger signal to the camera driving unit when the position of the wind turbine blades is at the preset position; The camera driving unit is configured to drive the image acquisition module to acquire the surface image of the wind turbine blade in response to the pulse trigger signal.

3. The system according to claim 2, characterized in that, The main core is equipped with a second acoustic signal processing unit, and the sub-core is equipped with an image processing unit; wherein... The second sound signal processing unit is configured to, in response to the wind turbine blades running to the preset position, perform a first classification on the sound data based on a pre-trained sound model, determine whether there is an abnormality in the wind turbine blades based on the first classification result, and send the enable signal to the image processing unit if there is an abnormality. The image processing unit is configured to, in response to the enable signal, perform a second classification on the image data based on a pre-trained image model, and determine the damage type of the wind turbine blades based on the second classification result; The sound model is obtained by deep learning on a first preset model based on multiple sound data samples. The multiple sound data samples include sound data samples when the wind turbine blades are abnormal and sound data samples when the wind turbine blades are operating normally. The image model is obtained by deep learning on a second preset model based on multiple wind turbine blade image samples. The multiple wind turbine blade image samples include wind turbine blade image samples when the wind turbine blades have different types of damage, and wind turbine blade image samples when the wind turbine blades do not have damage.

4. The system according to claim 3, characterized in that, The image model is configured with image feature parameters, and the image processing unit uses the pre-trained image model to perform a second classification on the image data, including: The texture feature parameters of the image data are calculated using the gray-level co-occurrence matrix; The second classification result of the image data is obtained by comparing the texture feature parameters with the image feature parameters of the image model.

5. The system according to claim 3, characterized in that, The processor module also includes a storage unit, which is communicatively connected to the first acoustic signal processing unit, the camera driving unit, the second acoustic signal processing unit, and the image processing unit, respectively; wherein, The first acoustic signal processing unit is further configured to preprocess the acoustic data collected by the acoustic array module and store the preprocessed acoustic data in the storage unit; The camera driving unit is further configured to store the image data acquired by the image acquisition module into the storage unit; The second sound signal processing unit is further configured to extract the sound data corresponding to the wind turbine blade running to a preset position from the storage unit when performing the first classification on the sound data, so as to perform the first classification on the sound data corresponding to the preset position; The image processing unit is further configured to extract image data to be classified into the second category from the storage unit in response to the enable signal.

6. The system according to claim 5, characterized in that, The storage unit includes a first storage space and a second storage space; wherein the first storage space is configured to store the sound data, and the second storage space is configured to store the image data; The first storage space and the second storage space are independent of each other.

7. The system according to claim 5, characterized in that, The main core is configured with an asymmetric multiprocessing mode; The main core is also configured to, after startup, activate the secondary core through the shared memory based on the asymmetric multiprocessing mode, so that the secondary core and the main core can execute different tasks independently. When the main core is started, it responds to the pulse trigger signal and extracts the sound data from the storage unit for anomaly identification. After being activated, the sub-core, in response to the enable signal, extracts the image data from the storage unit to identify the damage type.

8. The system according to claim 3, characterized in that, The first acoustic signal processing unit is further configured to send the pulse trigger signal to the image processing unit when the wind turbine blades run to the preset position; The image processing unit is further configured to respond to the pulse trigger signal to determine whether there is an abnormality in the wind turbine blades. If there is an abnormality and the second acoustic signal processing unit determines that the blades are also abnormal, the image data is classified in a second way based on the pre-trained image model.

9. The system according to claim 1, characterized in that, The system also includes an alarm module, which is electrically connected to the processing module and is configured to send an alarm signal and damage information when the wind turbine blades are identified as being damaged.

10. The system according to claim 1, characterized in that, The processing module is connected to the acoustic array module and the image acquisition module on multiple wind turbines; The main core is also configured to detect the memory levels of the main core and the sub-core in real time, and dynamically adjust the task allocation of the main core and the sub-core through the shared memory when the memory levels of the main core or the sub-core are insufficient. The task allocation is used to assign anomaly recognition tasks for the sound data and recognition tasks for the image data.

11. A fan, characterized in that, Includes towers, wind turbine blades, and the wind turbine blade damage identification system according to any one of claims 1-10.

12. The fan according to claim 11, characterized in that, It also includes a medium substrate located on the tower. The medium substrate is on which the acoustic array module and image acquisition module of the wind turbine blade damage identification system are arranged. The installation height of the medium substrate is higher than the lowest point of the projection of the tip of the wind turbine blade on the tower.

13. The fan according to claim 12, characterized in that, The medium substrate is cubic in shape, and the acoustic array module and the image acquisition module are located on the same surface of the medium substrate.

14. The fan according to claim 13, characterized in that, The camera in the image acquisition module has a wide-angle lens with a field of view of 60 degrees or greater.

Images