Fan blade defect detection device and detection method

Abstract

The invention belongs to the technical field of fan detection, and provides a fan blade defect detection device and method, and the device comprises a pedestal, a composite mechanism, an adsorption structure, a supporting structure, and a processor. The composite mechanism comprises a detection module and a navigation positioning module; the adsorption structure is composed of a negative pressure cavity and a vacuum pump, and the vacuum pump vacuumizes the negative pressure cavity; the supporting structure is provided with a steering wheel and a bull eye wheel which are coplanar with the negative pressure cavity; and the processor is integrated on the base, electrically connected with all the parts and connected with the memory for storing related data. The vacuum pump drives the negative pressure adsorption device to serve as a carrying platform, noise is lower, adsorption force is higher, the device can be stably attached to the complex curved surface of the fan blade, reliable movement is achieved in the severe environment, and the limitation of manual inspection is broken through. The visual navigation module realizes accurate self-positioning and path planning, so that missing detection is avoided; the detection module simulates manual knocking to collect voiceprint signals, and early recognizes internal cracks of the blade through real-time comparison and analysis, so that the timeliness and accuracy of defect detection are greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of wind turbine testing technology, and specifically relates to a device and method for detecting defects in wind turbine blades. Background Technology

[0002] As the core component of energy conversion, the operating status of wind turbine blades directly determines the safety, stability, and power generation efficiency of the wind turbine. However, current blade inspection technology has many critical issues that urgently need to be addressed, which seriously restricts the quality and safety of wind farm operation and maintenance.

[0003] Blades operate in extreme natural environments such as strong winds, torrential rain, lightning, and extreme temperatures for extended periods. Their surfaces and interiors are highly susceptible to developing minute defects, and these defects can deteriorate into fatal flaws in just 1-2 months, placing stringent demands on the timeliness of inspections. However, current mainstream manual inspection methods typically have an inspection cycle of 3 months, significantly lagging behind the rate of defect deterioration. This makes it difficult to detect early defects in a timely manner, resulting in missed opportunities for optimal intervention.

[0004] The blade structure is characterized by its enormous size (tens of meters), narrow internal space, and dim lighting, posing natural obstacles to manual inspection. Inspectors must work from elevated platforms, using rubber mallets to tap and listen to the sound of the blades to assess their condition. This operation is not only extremely risky but also susceptible to subjective factors such as the operator's experience and sense of responsibility, making it difficult to comprehensively cover high-risk areas like the blade root, tip, leading edge, and trailing edge, easily leading to missed or false inspections. Furthermore, manual inspection involves costs related to personnel salaries, transportation, and equipment rental, and is inefficient, failing to meet the rapid inspection needs of large-scale wind farms.

[0005] More importantly, crack-like defects account for over 30% of all blade defects and are the main cause of blade breakage. Once a breakage occurs, it can cause huge economic losses and potentially lead to safety accidents. However, current inspection technologies lack effective early warning methods for internal blade cracks, a long-standing technical challenge for the industry. These problems prevent timely, accurate, and comprehensive detection of blade defects, seriously threatening wind turbine operational safety. Therefore, a more efficient and reliable inspection technology solution is urgently needed. Summary of the Invention

[0006] To address the problems in the background art, this invention proposes a device and method for detecting defects in wind turbine blades.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a device for detecting defects in wind turbine blades, comprising: Base; The composite mechanism includes a detection module and a navigation and positioning module mounted on the base surface; the detection module is used to strike the wind turbine blades and collect acoustic data; the navigation and positioning module is used to acquire images and positions of the wind turbine blades. The adsorption structure includes a negative pressure chamber and a vacuum pump. The negative pressure chamber is installed at the center of one surface of the base. The vacuum pump is installed on the surface of the base opposite to the negative pressure chamber and is used to evacuate the negative pressure chamber. The support structure includes a steering wheel and a bullseye wheel, which are distributed around the periphery of the negative pressure chamber and are mounted on the same surface of the base, coplanar with the negative pressure chamber. The processor, integrated within the base, is electrically connected to the composite mechanism, vacuum pump, and support structure, and is also connected to a memory for storing voiceprint data and images.

[0008] Furthermore, the detection module includes an excitation element and an acoustic fingerprint sensor electrically connected to the processor; The excitation element is mounted on the surface of the base and is located on one side of the negative pressure chamber; Several acoustic sensors are mounted on the surface of the base and are evenly distributed around the periphery of the exciter.

[0009] Furthermore, the navigation and positioning module includes a first camera, a second camera, and an infrared laser sensor; The first camera is mounted on the upper surface of the base to acquire a global image of the wind turbine blades; The second camera is mounted on the lower surface of the base to acquire local images of the wind turbine blades; Several infrared laser sensors are installed on the peripheral surface of the base for positioning the detection device; The first camera, the second camera, and the infrared laser sensor are all connected to the processor.

[0010] Furthermore, the actuating components include a spring, an electromagnet, a hammer, and a movable rod; The hammer is fixedly connected to a movable rod; The movable rod is slidably connected to the electromagnet and passes through the electromagnet; The spring is sleeved on the surface of the movable rod, located on the side of the electromagnet away from the hammer, with one end of the spring abutting against the end flange of the movable rod and the other end abutting against the electromagnet; When an electromagnet is energized and becomes magnetized, the hammer moves in the direction that compresses the spring.

[0011] Furthermore, the two steering wheels are symmetrically distributed around the negative pressure chamber; The spatial arrangement of the two steering wheels and the bullseye wheel forms a triangular support structure.

[0012] Furthermore, the steering wheel includes a steering motor, rollers, drive motor, base, and support; The base is fixed to the base plate; The support is rotatably connected to the base; The steering motor is mounted on the base surface and is used to drive the support to rotate; The rollers are mounted on the support and rotate. The drive motor is mounted on the side of the support and is used to drive the roller to rotate; Both the steering motor and the drive motor are electrically connected to the processor.

[0013] Furthermore, the steering motor drives the support to steer via gear transmission or pulley transmission.

[0014] The present invention also provides a detection method for the above-mentioned wind turbine blade defect detection device, comprising the following steps: The processor uses the location and images collected by the navigation and positioning module to plan the movement path on the wind turbine blades; The processor obtains the current position of the detection device on the wind turbine blades through the navigation and positioning module; At the current location, tap the fan blades to collect the generated acoustic data; and obtain an image of the fan blades by detecting the structure; and obtain the pressure value inside the negative pressure chamber; The processor uses acoustic fingerprint data, images of the wind turbine blades, and pressure values ​​within the negative pressure chamber to determine existing defects in the wind turbine blades and predict future defects.

[0015] Furthermore, this includes the following steps: After acquiring the acoustic data, the image of the fan blades, and the pressure value inside the negative pressure chamber, the processor timestamps the acoustic data, the image of the fan blades, and the pressure value inside the negative pressure chamber.

[0016] Furthermore, based on acoustic signature data, images of the wind turbine blades, and pressure values ​​within the negative pressure chamber, existing defects in the wind turbine blades are determined, and defects are predicted. This includes the following steps: The processor determines the defects that have occurred on the fan blades based on acoustic fingerprint data, images of the fan blades, and pressure values ​​in the negative pressure chamber. The processor inputs acoustic print data, images of wind turbine blades, pressure values ​​in the negative pressure chamber, and existing defects in the wind turbine blades into a multimodal fusion classification network. Through cross-modal feature comparison, it predicts potential defects in the wind turbine blades.

[0017] The beneficial effects of this invention are: 1. The system of this invention uses a vacuum pump-driven negative pressure adsorption device as a mounting platform. Compared with the traditional method of drawing negative pressure with a fan, it significantly reduces operating noise and generates stronger adsorption force, which can stably attach to the complex curved surface of large fan blades. It can reliably move in harsh environments such as high-altitude strong winds and low light, breaking through the bottleneck of manual inspection being unable to reach the hidden parts of the blades and the limited working environment. Secondly, the visual navigation module integrated in this invention can accurately complete autonomous positioning and path planning, ensuring that the robot efficiently covers the entire blade area and avoids the problem of missed inspections caused by subjective experience differences in manual inspection. 2. The detection module of this invention simulates the action of manual knocking. With the carried acoustic sensor, it can efficiently collect the acoustic signal generated by the vibration of the blade. The built-in memory can compress and store the acoustic data. At the same time, it can realize the real-time comparison and analysis of acoustic features through the built-in feature library, so as to realize the early identification and warning of fatal defects such as internal cracks of the blade. This makes up for the insufficiency of manual inspection cycle in covering the period from defect deterioration to the critical defect stage, and greatly improves the timeliness and accuracy of defect detection. 3. The invention has a high degree of automation, which reduces reliance on manual labor, lowers the safety risks of high-altitude operations and long-term inspection costs, and provides efficient and reliable technical support for the safe operation of wind turbine blades, with broad prospects for industrial application.

[0018] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description and the drawings. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A schematic diagram of the bottom structure of a wind turbine blade defect detection device according to the present invention is shown; Figure 2 A schematic diagram of the top structure of a wind turbine blade defect detection device according to the present invention is shown; Figure 3 A three-dimensional isometric view of the bottom surface of a wind turbine blade defect detection device according to the present invention is shown. Figure 4 A schematic diagram of the structure of the excitation element of the present invention is shown; Figure 5A flowchart of a detection method according to the present invention is shown.

[0021] In the diagram: 1. Base; 2. Composite mechanism; 201. Excitation element; 2011. Spring; 2012. Electromagnet; 2013. Hammer; 2014. Movable rod; 202. Acoustic sensor; 203. First camera; 204. Second camera; 205. Infrared laser sensor; 3. Steering wheel; 301. Steering motor; 302. Roller; 303. Drive motor; 304. Base; 305. Support; 4. Bullseye wheel; 5. Negative pressure chamber; 6. Vacuum pump; 7. Processor; 8. Memory. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] like Figure 1 As shown, a device for detecting defects in wind turbine blades includes a base 1, a composite mechanism 2, an adsorption structure, a support structure, and a processor 7.

[0024] The base 1 supports the composite mechanism 2, the adsorption structure, and the support structure, and integrates a processor 7 internally. This provides a stable installation reference and circuit connection space for each functional module, ensuring the compactness and operational stability of the overall structure of the device. Furthermore, as the structural foundation of the entire device, the base 1 is made of lightweight, high-strength materials, ensuring structural stability while reducing the overall weight of the device and facilitating movement on the blade surface.

[0025] The composite mechanism 2 is mounted on the surface of the base 1 and includes a detection module and a navigation and positioning module. The detection module is used to strike the fan blades and collect acoustic signature data, providing acoustic signature evidence for defect judgment. This detection module includes an exciter 201 and acoustic signature sensors 202 electrically connected to the processor 7. The exciter 201 is mounted on the surface of the base 1 and located on one side of the negative pressure chamber 5, used to generate a striking action. Several acoustic signature sensors 202 are evenly distributed along the periphery of the exciter 201 on the surface of the base 1, enabling omnidirectional acoustic signature data collection, improving the integrity and accuracy of the acoustic signature signal, and avoiding signal deviation caused by a single sensor.

[0026] Optionally, the actuating element 201 includes a spring 2011, an electromagnet 2012, a hammer 2013, and a movable rod 2014. The hammer 2013 is fixedly connected to the movable rod 2014, and the movable rod 2014 passes through the electromagnet 2012 and is slidably connected to it. The spring 2011 is sleeved on the end of the movable rod 2014 away from the hammer 2013, and its two ends abut against the end flange face of the movable rod 2014 and the electromagnet 2012, respectively.

[0027] When the processor 7 controls the electromagnet 2012 to be energized, the electromagnet 2012 generates a magnetic force that attracts the movable rod 2014, causing the hammer 2013 to move towards the compression spring 2011 until the spring 2011 is compressed to a preset deformation degree. When the electromagnet 2012 is de-energized, the spring 2011 releases its elastic potential energy, pushing the movable rod 2014 to quickly return to its original position, thereby causing the hammer 2013 to strike the blade surface. This design can precisely adjust the striking force and frequency by controlling the frequency and current of the electromagnet 2012, ensuring the consistency of each strike and providing a reliable basis for the comparative analysis of the voiceprint data collected by the voiceprint sensor 202.

[0028] It should be noted that the control circuit of the exciter 201 can be a microcontroller or an ARM (Advanced RISC Machine) device.

[0029] The navigation and positioning module includes a first camera 203, a second camera 204, and multiple sets of infrared laser sensors 205. The first camera 203 is mounted on the upper surface of the base 1, with its lens facing the blade's extension direction, capturing a global image of the blade. The processor 7 analyzes this global image to plan the device's movement path, ensuring detection covers the entire blade surface. The second camera 204 is mounted on the lower surface of the base 1, with its lens perpendicular to the blade surface, capturing high-definition local images of the detection points. This captures subtle morphological features of the blade surface, assisting the acoustic signature data collected by the acoustic signature sensor 202 in determining defect types. The infrared laser sensors 205 are evenly distributed around the base 1. By emitting lasers and receiving reflected signals, they calculate in real-time the distance between the device and reference points such as blade edges and surface protrusions, thereby determining the device's precise position on the blade surface and preventing the device from deviating from its path or colliding with blade edges during movement.

[0030] It should be noted that the navigation and localization module can generally be combined with SLAM (Simultaneous Localization and Mapping) technology, enabling the detection device to move from an unknown location in an unknown environment. During the movement, it performs self-localization based on position estimation and sensor data, while simultaneously building an incremental map. During operation, the detection device calculates changes in the optical flow field in real time to compensate for the effects of blade swaying and lighting variations in wind conditions. Simultaneously, visual navigation and the SLAM algorithm work in tandem; the former provides anchor point constraints, while the latter provides continuous trajectory estimation. Their fusion ensures that the robot achieves high-precision, autonomous navigation and localization in environments with a wide range of curved surfaces. This method overcomes the limitations of existing visual navigation, which is mainly applied to planar or indoor environments, ensuring reliable robot operation on complex blade surfaces.

[0031] The first camera 203, the second camera 204 and the infrared laser sensor 205 are all connected to the processor 7 via lines. The collected image data and position data can be transmitted to the processor 7 in real time, providing data support for the device's movement control and defect judgment.

[0032] The adsorption structure primarily ensures the stability of the detection device on the surface of the wind turbine blades (inclined or even vertical surfaces). It consists of two parts: a negative pressure chamber 5 and a vacuum pump 6. The negative pressure chamber 5 is designed as a circular cavity structure, installed at the center of the base 1 facing the blade surface. Its opening end can be sealed with a rubber gasket to form a sealed space. The vacuum pump 6 is installed on the other surface of the base 1 (the surface opposite the blades) and is connected to the inside of the negative pressure chamber 5 via an air pipe. When the device moves to the detection position, the processor 7 controls the vacuum pump 6 to start, evacuating the inside of the negative pressure chamber 5 to create a negative pressure zone between the negative pressure chamber 5 and the blade surface. Atmospheric pressure then tightly adsorbs the device onto the blade surface. The advantages of this adsorption method are as follows: On the one hand, by adjusting the pumping power of the vacuum pump 6, the negative pressure value in the negative pressure chamber 5 can be controlled, thereby adapting to blade surfaces with different curvatures and materials. This ensures firm adsorption while avoiding damage to the blade surface caused by excessive adsorption force. On the other hand, the negative pressure chamber 5 is located in the center of the base 1, and its adsorption force can be evenly distributed throughout the entire device. Combined with the supporting structure, this further enhances the stability of the device.

[0033] The support structure includes a steering wheel 3 and a bullseye wheel 4. All wheels are mounted coplanarly with the negative pressure chamber 5 on the same surface of the base 1 to ensure flatness when in contact with the blade surface. Specifically, the two steering wheels 3 are symmetrically distributed around the negative pressure chamber 5, while the bullseye wheel 4 is mounted on the other side of the base 1, forming a triangular support structure. The triangular structure itself has good stability, effectively preventing the device from tipping over during movement or testing; at the same time, this layout allows the adsorption force of the negative pressure chamber 5 to be evenly distributed to the three support points, further improving the adhesion stability between the device and the blade surface.

[0034] Optionally, the steering wheel 3, as the active moving component, comprises a base 304, a support 305, a steering motor 301, a drive motor 303, and a roller 302. The base 304 is fixed to the base 1, and the support 305 is rotatably connected to the base 304. The steering motor 301 is mounted on the surface of the base 304 and drives the support 305 to rotate via gear or pulley transmission, thereby causing the roller 302 to change direction. The drive motor 303 is mounted on the side of the support 305 and directly drives the roller 302 to rotate, providing power for the movement of the device. The processor 7 controls the steering motor 301 and drive motor 303 of the two steering wheels 3 to realize various movement actions such as forward, backward, and turning of the device. The bullseye wheel 4, as the driven component, adopts a universal rolling structure and can adaptively adjust its rolling direction according to the movement direction of the steering wheel 3, which reduces the frictional resistance during device movement and helps support the weight of the device, ensuring smooth movement.

[0035] The processor 7 is generally installed inside the base 1 and is connected to all electrical components of the composite mechanism 2, the adsorption structure, and the support structure via internal wiring. The processor 7 is also connected to a memory 8, which can store images, acoustic data, etc., of the wind turbine blades.

[0036] The processor 7 can receive position and image data transmitted from the first camera 203, the second camera 204, and the infrared laser sensor 205 of the navigation and positioning module, plan the detection path, and then control the steering wheel 3 (steering motor 301, drive motor 303) of the support structure to move the device. After the device reaches the detection position, it controls the vacuum pump 6 of the adsorption structure to start to achieve stable adsorption. According to the detection requirements, it controls the excitation element 201 of the detection module in the composite mechanism 2 to perform a tapping action, and synchronously triggers the acoustic sensor 202 to collect data.

[0037] The processor 7 can also receive voiceprint data from the voiceprint sensor 202, image data from the first camera 203 and the second camera 204, and pressure data from the negative pressure chamber 5 fed back by the vacuum pump 6. It preprocesses these data (such as noise reduction and image enhancement) and then analyzes them through a preset algorithm to determine whether there are defects in the blade and the type of defects.

[0038] In addition, processor 7 can input preprocessed multi-dimensional data (acoustic print, image, pressure) and identified defect information into a multimodal fusion classification network. Through cross-modal feature comparison, it can mine potential correlations between data, predict possible future defects in the blades, and provide forward-looking suggestions for maintenance.

[0039] like Figure 5 The image shows a detection method for a wind turbine blade defect detection device, comprising the following steps: S1: The processor 7 plans the movement path on the wind turbine blades based on the location and images collected by the navigation and positioning module.

[0040] S2: The processor 7 obtains the current position of the detection device on the wind turbine blades through the navigation and positioning module.

[0041] S3: At the current position, tap the fan blades to collect the generated acoustic data; and, obtain the image of the fan blades through the detection structure; and, obtain the pressure value inside the negative pressure chamber 5. Afterwards, the processor 7 can timestamp the acoustic data, the image of the fan blades, and the pressure value inside the negative pressure chamber 5 as needed.

[0042] S4: Processor 7 determines existing defects in the wind turbine blades and predicts future defects based on acoustic signature data, images of the wind turbine blades, and pressure values ​​within the negative pressure chamber 5. For example, processor 7 determines existing defects in the wind turbine blades based on acoustic signature data, images of the wind turbine blades, and pressure values ​​within the negative pressure chamber 5; then, processor 7 inputs the acoustic signature data, images of the wind turbine blades, pressure values ​​within the negative pressure chamber 5, and existing defects into a multimodal fusion classification network, and predicts potential defects in the wind turbine blades through cross-modal feature comparison.

[0043] It should be noted that, in the above embodiments, the processor 7 can also perform defect analysis on a single set of data. For example, the processing of voiceprint data can be performed using the following method: 1) Noise reduction is performed on the collected leaf acoustic data to filter out environmental noise and improve the signal-to-noise ratio of acoustic events.

[0044] 2) For the noise-reduced audio signal sample set in a healthy state, a contrastive learning approach is used to perform self-supervised learning on the pre-trained wav2vec speech feature extractor.

[0045] 3) Input the noise-reduced acoustic signal sample set of the health status into the above wav2vec speech feature extractor to obtain the health status feature library.

[0046] 4) Calculate the cosine distance between features in the feature library and use the 3sigma criterion to determine the distance threshold; 5) Calculate the output features of the wav2vec speech feature extractor for the test sample, calculate the cosine distance between the feature and all features in the feature library, find the minimum distance, and when the minimum distance is greater than the threshold, the test sample is abnormal; otherwise, it is normal.

[0047] The following is a detailed explanation of the fully automated inspection process for wind turbine blade defects based on the S1-S4 inspection methods:

[0048] 1) Before the detection begins, the device is placed at the initial detection position on the wind turbine blade. The processor 7 first activates the navigation and positioning module: the first camera 203 captures a global image of the blade, and the infrared laser sensor 205 collects initial position data. The processor 7 combines the global image and the initial position data to plan a detection path covering the entire surface of the blade and determine the position coordinates of each detection point. Subsequently, the processor 7 controls the steering wheel 3 (steering motor 301, drive motor 303) of the support structure to move, driving the device along the planned path. During the movement, the infrared laser sensor 205 provides real-time feedback of position data, and the processor 7 adjusts the steering and rotation speed of the steering wheel 3 based on the feedback to ensure that the device accurately reaches each preset detection position.

[0049] 2) Once the device reaches a certain detection position, the detection process enters the data acquisition phase, and the modules work together according to the following logic:

[0050] Adsorption and fixation: The processor 7 first controls the vacuum pump 6 to start, evacuating the negative pressure chamber 5. At the same time, the pressure sensor monitors the internal pressure value of the negative pressure chamber 5 in real time. When the pressure reaches the preset stable value, the vacuum pump 6 keeps running at a low speed to maintain the negative pressure and ensure stable adsorption of the device.

[0051] Voiceprint acquisition: The processor 7 controls the electromagnet 2012 of the detection module in the composite mechanism 2 to be energized, which drives the hammer 2013 to compress the spring 2011. Then the power is cut off, causing the hammer 2013 to strike the blade. At the same time as the hammer strikes, the voiceprint sensor 202 arranged around the hammer 2013 synchronously acquires the voiceprint data signal and converts the signal into digital data to be transmitted to the processor 7.

[0052] Image acquisition: Synchronous with acoustic fingerprint acquisition, the second camera 204 captures a local high-definition image of the blade surface at the current detection point, while the first camera 203 supplements the capture of global image segments of the current area. Both types of image data are transmitted to the processor 7 in real time to assist in defect judgment.

[0053] Data synchronization: The processor 7 timestamps the collected voiceprint data from the voiceprint sensor 202, the image data from the first camera 203 and the second camera 204, and the pressure data from the negative pressure chamber 5 to ensure that the three types of data correspond to the same detection point, thus providing data consistency assurance for subsequent analysis.

[0054] 3) After data acquisition, the processor 7 first compares the acoustic data collected by the acoustic sensor 202 with the preset "defect-free blade acoustic template". If there is a significant difference between the two (such as different acoustic frequencies and attenuation rates), it is preliminarily determined that there may be a defect at the detection point. Subsequently, the processor 7 performs image recognition on the local image captured by the second camera 204 to analyze whether there are morphological features such as cracks, dents, and delamination. Combining the acoustic differences and morphological features, the processor 7 determines the type (such as surface cracks, internal delamination), location (based on the positioning data of the infrared laser sensor 205), and size (based on the image ratio calculation) of the defect. At the same time, the pressure data of the negative pressure chamber 5 can verify whether the device is well attached. If the pressure is unstable, it indicates that there is a gap between the device and the blade surface. The adsorption state needs to be readjusted and the data collected again to avoid detection errors caused by poor attachment.

[0055] 4) Subsequently, the processor 7 inputs the acoustic data from the acoustic sensor 202 at the current detection point, the image data from the first camera 203 and the second camera 204, the pressure data from the negative pressure chamber 5, and the identified defect information (if any) into the pre-trained multimodal fusion classification network. This network, by learning from a large amount of blade defect evolution data, can uncover the correlations between data from different dimensions (e.g., although abnormal acoustic attenuation in a certain area has not formed an obvious crack, combined with surface micro-damage in the image, it can be determined that there is a crack initiation trend), and then outputs the potential defect types and expected occurrence times that may occur in that area in the future, providing data support for the preventative maintenance of wind turbine blades.

[0056] 5) After completing the analysis of the current detection point, the processor 7 controls the vacuum pump 6 to release the pressure in the negative pressure chamber 5, so that the detection device can move, but the negative pressure in the negative pressure chamber 5 can ensure that the detection device can be adsorbed. Then, the rudder wheel 3 of the support structure is controlled to drive the device to the next detection point, and the above process is repeated until the detection of the entire blade is completed, and finally a complete detection report containing the defect location, type, size and potential defect prediction is generated.

[0057] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A device for detecting defects in wind turbine blades, characterized in that, include: Base (1); The composite mechanism (2) includes a detection module and a navigation and positioning module mounted on the surface of the base (1); The detection module is used to tap the wind turbine blades and collect acoustic data; the navigation and positioning module is used to acquire images and positions of the wind turbine blades. The adsorption structure includes a negative pressure chamber (5) and a vacuum pump (6). The negative pressure chamber (5) is installed at the center of a surface of the base (1). The vacuum pump (6) is installed on the surface of the base (1) away from the negative pressure chamber (5) and is used to evacuate the negative pressure chamber (5). The support structure includes a steering wheel (3) and a bullseye wheel (4), wherein the steering wheel (3) and the bullseye wheel (4) are distributed on the periphery of the negative pressure chamber (5) and are coplanarly mounted on the same surface of the base (1) with the negative pressure chamber (5); The processor (7) is integrated in the base (1) and is electrically connected to the composite mechanism (2), the vacuum pump (6) and the support structure respectively. The processor (7) is also connected to a memory (8) for storing voiceprint data and images.

2. The device for detecting defects in wind turbine blades according to claim 1, characterized in that, The detection module (2) includes an excitation element (201) and an acoustic fingerprint sensor (202) electrically connected to the processor (7). The excitation element (201) is mounted on the surface of the base (1) and located on one side of the negative pressure chamber (5); Several of the aforementioned acoustic sensors (202) are mounted on the surface of the base (1) and are evenly distributed along the periphery of the exciter (201).

3. The device for detecting defects in wind turbine blades according to claim 2, characterized in that, The navigation and positioning module includes a first camera (203), a second camera (204), and an infrared laser sensor (205). The first camera (203) is mounted on the upper surface of the base (1) to acquire a global image of the wind turbine blades; The second camera (204) is mounted on the lower surface of the base (1) to acquire partial images of the wind turbine blades; Several infrared laser sensors (205) are mounted on the peripheral surface of the base (1) for positioning the detection device; The first camera (203), the second camera (204) and the infrared laser sensor (205) are all connected to the processor (7).

4. The device for detecting defects in wind turbine blades according to claim 2, characterized in that, The actuating element (201) includes a spring (2011), an electromagnet (2012), a hammer (2013), and a movable rod (2014). The hammer (2013) is fixedly connected to a movable rod (2014); The movable rod (2014) is slidably connected to the electromagnet (2012) and passes through the electromagnet (2012). The spring (2011) is sleeved on the surface of the movable rod (2014), located on the side of the electromagnet (2012) away from the hammer (2013), and one end of the spring (2011) abuts against the end flange of the movable rod (2014), and the other end abuts against the electromagnet (2012); When the electromagnet (2012) is energized and magnetized, the hammer (2013) moves toward the compression spring (2011).

5. The device for detecting defects in wind turbine blades according to claim 1, characterized in that, The two steering wheels (3) are symmetrically distributed with the negative pressure chamber (5) as the center; The spatial arrangement of the two steering wheels (3) and the bullseye wheel (4) forms a triangular support structure.

6. The device for detecting defects in wind turbine blades according to claim 1, characterized in that, The steering wheel (3) includes a steering motor (301), a roller (302), a drive motor (303), a base (304), and a support (305); The base (304) is fixed on the base (1); The support (305) is rotatably connected to the base (304); The steering motor (301) is mounted on the surface of the base (304) and is used to drive the support (305) to rotate; The roller (302) is rotatably mounted on the support (305); The drive motor (303) is mounted on the side of the support (305) and is used to drive the roller (302) to rotate; The steering motor (301) and drive motor (303) are both electrically connected to the processor (7).

7. The device for detecting defects in wind turbine blades according to claim 6, characterized in that, The steering motor (301) drives the support (305) to turn via gear transmission or pulley transmission.

8. A detection method for use in the wind turbine blade defect detection device according to any one of claims 1-7, characterized in that, Includes the following steps: The processor (7) plans the movement path on the wind turbine blades based on the position and images collected by the navigation and positioning module; The processor (7) obtains the current position of the detection device on the wind turbine blades through the navigation and positioning module; At the current location, tap the wind turbine blades to collect the generated acoustic data; And, by detecting the structure, images of the wind turbine blades are obtained; And, obtain the pressure value inside the negative pressure chamber (5); The processor (7) determines the defects that have occurred on the wind turbine blades and predicts defects based on the acoustic data, the image of the wind turbine blades and the pressure value in the negative pressure chamber (5).

9. The detection method according to claim 8, characterized in that, Includes the following steps: After acquiring the acoustic data, the image of the fan blade, and the pressure value in the negative pressure chamber (5), the processor (7) timestamps the acoustic data, the image of the fan blade, and the pressure value in the negative pressure chamber (5).

10. The detection method according to claim 8, characterized in that, Based on acoustic signature data, images of the wind turbine blades, and pressure values ​​within the negative pressure chamber (5), the existing defects in the wind turbine blades are determined, and defects are predicted. The steps include: The processor (7) determines the defects that have occurred in the fan blades based on the acoustic data, the image of the fan blades, and the pressure value in the negative pressure chamber (5); The processor (7) inputs the acoustic data, the image of the wind turbine blade, the pressure value in the negative pressure chamber (5) and the defects that have been generated in the wind turbine blade into the multimodal fusion classification network, and predicts the potential defects of the wind turbine blade through cross-modal feature comparison.

Images