An event camera based vertical axis wind turbine blade defect detection system
By using an event-camera-based vertical axis wind turbine blade inspection system, the significance value and area of defects are calculated using event density maps. This solves the problems of inaccurate detection and high labor costs in existing technologies, and enables accurate quantitative detection of defects and prediction of future trends.
Patent Information
- Application Number
- CN202611344084.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-01
- Publication Date
- 2026-11-17
AI Technical Summary
In existing technologies, the detection of defects in vertical axis wind turbine blades suffers from motion blur and image trailing problems. Vibration signal detection makes it difficult to locate defects, and manual inspection cannot achieve real-time online detection, resulting in inaccurate detection and high labor costs.
An event camera-based detection system is adopted. By acquiring the raw event stream data output by the event camera, the event density map of the blade surface within a preset time window is calculated, the defect significance value is calculated and the defect area is quantified. Combined with the safe life prediction module, the system can achieve accurate quantitative detection of defects and prediction of future trends.
It reduces motion blur and ghosting issues, improves the accuracy of defect area identification, provides quantifiable defect area information, and can predict defect development trends, solving the problems of large interference factors, high labor costs, and large errors in existing technologies.
Smart Images

Figure CN123118546A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of wind turbine blade health monitoring technology, and in particular to a vertical axis wind turbine blade defect detection system based on an event camera. Background Technology
[0002] Vertical axis wind turbines are widely used in distributed wind power and urban wind energy utilization due to their compact structure and strong adaptability to wind direction. During operation, the blades of vertical axis wind turbines are subjected to long-term cyclic aerodynamic and centrifugal loads, which can easily lead to structural defects such as cracks and scratches. Timely detection of these structural defects is essential to ensure the normal operation of the blades.
[0003] In existing technologies, blade defect detection is mainly achieved through visual image detection, vibration signal detection, and manual inspection. Visual image detection uses ordinary industrial cameras or high-speed cameras to acquire blade images, and then uses image processing algorithms to identify blade defects. This method has its drawbacks. Specifically, during operation, the blade rotates continuously around its vertical axis, and the blade surface is in a high-speed motion state within the camera's field of view. Traditional frame cameras are limited by exposure time and frame rate, which easily leads to motion blur and image ghosting. Vibration signal detection uses accelerometers installed on the blades to collect wind turbine operating status signals. This detection method is difficult to provide the specific location and area of defects on the blade surface, resulting in inaccurate defect identification. Furthermore, vibration signals are easily affected by wind speed fluctuations and tower vibrations, leading to significant false alarms or missed alarms. Manual inspection requires maintenance personnel to visually observe and use telescopes to collect data after the wind turbine is shut down. This method has problems such as long inspection cycles, high labor costs, and, more importantly, cannot achieve real-time online detection.
[0004] Therefore, there is an urgent need to design a defect detection system for vertical axis wind turbine blades based on event cameras. Summary of the Invention
[0005] Therefore, it is necessary to provide a vertical axis wind turbine blade defect detection system based on an event camera to address the above-mentioned technical problems. This system can solve the technical problems of large interference factors caused by vibration detection of blade defects, high labor costs and large errors in manual detection.
[0006] The technical solution of this invention is as follows: A vertical axis wind turbine blade defect detection system based on an event camera, the system comprising: An event data acquisition module is used to acquire raw event stream data output by an event camera, wherein the event camera is positioned around the blades of a vertical axis wind turbine, and the detection range of the event camera covers the surface of the wind turbine blades when they rotate. The event defect calculation module is used to: calculate the event density map of the blade surface within a preset time window based on the original event stream data, and calculate the defect significance value based on the event density map; The defect area calculation module is used to: calculate the blade defect area based on the area where the defect significance value exceeds a preset significance threshold, wherein a preset time window corresponds to one blade defect area; The safe life prediction module is used to predict the remaining safe life of a blade based on a sequence of blade defect areas composed of multiple consecutive preset time windows.
[0007] Optionally, the raw event stream data includes multiple asynchronous events; The event defect calculation module performs the following steps: Calculates the event density map of the blade surface within a preset time window based on the original event stream data. Step S211: Count the number of asynchronous events falling into the same pixel position within the preset time window; Step S212: Calculate the event density value at the corresponding pixel position based on the number of asynchronous events and the time length of the preset time window, and generate an event density map.
[0008] Optionally, the event density value is calculated based on the following formula:
[0009] in, It is the event density value; It is the length of the preset time window; It is the Dirac function; This indicates whether the i-th asynchronous event falls within the pixel position (x, y); It is an indicator function.
[0010] Optionally, the event defect calculation module performs the following steps: calculating the defect saliency value based on the event density map. Step S221: Obtain the event density value corresponding to each pixel position in the event density map; Step S222: Obtain the normal reference event density value corresponding to each pixel position in the normal reference event density map, wherein the normal reference event density map is preset; Step S223: For any pixel location, calculate the defect significance value corresponding to the pixel location based on the difference between the event density value corresponding to the pixel location and the normal baseline event density value.
[0011] Optionally, the defect saliency value at any pixel location (x, y) at time t is generated based on the following formula:
[0012] in, It is the defect significance value corresponding to the pixel position (x, y) at time t. It is the event density value at pixel position (x, y) in the event density map at time t. It is the normal reference event density value corresponding to the pixel position (x,y) in the normal reference event density map at time t.
[0013] Optionally, the defect area calculation module performs the following steps: Calculates the blade defect area based on the region where the defect significance value exceeds a preset significance threshold. Step S310: Determine the number of pixels whose defect salience value exceeds a preset salience threshold; Step S320: Calculate the defect area of the blade based on the number of pixels and the actual physical area of the blade corresponding to a single pixel.
[0014] Alternatively, the blade defect area can be calculated based on the following formula:
[0015] in, This represents the area of the blade defect. This represents the actual physical area of the blade corresponding to a single pixel. To meet The number of pixels, This is a preset significance threshold.
[0016] Optionally, the safe life prediction module performs the following steps: predicting the remaining safe life of the blade based on a sequence of blade defect areas composed of multiple consecutive preset time windows. Step S410: Fit a linear expansion model of the defect area changing with time based on the defect area sequence of the blade; Step S420: Based on the extended model, the current blade defect area, and the preset defect area safety threshold, predict the remaining safe life of the blade.
[0017] Optionally, the safe life prediction module performs the following steps: predicting the remaining safe life of the blade based on the extended model, the current blade defect area, and a preset defect area safety threshold. Step S421: Determine the defect propagation rate based on the linear propagation model; Step S422: Based on the defect propagation rate, the current blade defect area, and the preset defect area safety threshold, predict the remaining safe life of the blade.
[0018] Alternatively, the remaining safe life of the blade can be predicted based on the following formula:
[0019] in, For the remaining safe life of the blade, To preset a safe threshold for defect area, This represents the current blade defect area. The defect propagation rate is the rate after correction based on the blade operating conditions.
[0020] Optionally, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps described in the event camera-based vertical axis wind turbine blade defect detection system.
[0021] Optionally, a computer-readable storage medium is also provided, on which a computer program is stored, which, when executed by a processor, implements the steps described in the above-described event-camera-based vertical axis wind turbine blade defect detection system.
[0022] The technical effects achieved by this invention are as follows: The aforementioned event camera-based vertical axis wind turbine blade defect detection system, by placing the event camera around the vertical axis wind turbine blade and ensuring its detection range covers the surface of the blade during rotation, can acquire raw event stream data triggered by changes in blade surface brightness during blade operation. Compared to traditional frame-based cameras that acquire complete images at a fixed frame rate, this reduces motion blur and ghosting issues during high-speed rotating blade image acquisition. Based on the raw event stream data output by the event camera, an event density map of the blade surface within a preset time window is calculated. This event density map reflects the frequency of brightness changes at different locations on the blade surface within the corresponding time window, providing a stable data foundation for subsequent defect saliency calculations and avoiding direct... This approach addresses the issues of data dispersion and asynchronous timing caused by processing raw event stream data. It calculates defect significance values based on event density maps, quantifying these anomalies to improve the accuracy of defect area identification. Furthermore, it calculates the blade defect area based on regions where the defect significance value exceeds a preset significance threshold. This transforms defect detection results from simply determining the presence of a defect into quantifiable defect area information. By correlating defect areas across multiple time windows in chronological order, it reflects the expansion trend of defects over time. This approach not only detects the current defect state but also predicts future defect trends, solving the problems of large interference factors, high labor costs, and large errors associated with vibration detection in existing technologies. Attached Figure Description
[0023] Figure 1 This is a flowchart illustrating the execution steps of a vertical axis wind turbine blade defect detection system based on an event camera in one embodiment. Figure 2 This is a structural block diagram of a vertical axis wind turbine blade defect detection system based on an event camera in one embodiment. Detailed Implementation
[0024] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0025] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.
[0026] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0027] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."
[0028] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0029] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0030] The terminal may be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices.
[0031] In one embodiment, such as Figures 1-2 As shown, a vertical axis wind turbine blade defect detection system based on an event camera is provided. The system includes an event data acquisition module, an event defect calculation module, a defect area calculation module, and a safe life prediction module.
[0032] Specifically, the event data acquisition module is used to: acquire raw event stream data output by the event camera, wherein the event camera is set around the blades of the vertical axis wind turbine, and the detection range of the event camera covers the surface of the wind turbine blades when they rotate. In this step, the raw event stream data includes multiple asynchronous events, and each asynchronous event includes at least the pixel coordinates of the event occurrence, the trigger timestamp, and the event polarity.
[0033] The original event stream data is , , It is the first The horizontal and vertical coordinates of each event on the imaging plane of the event camera correspond to the spatial projection position on the blade surface. For the first The event polarity of an asynchronous event; For the first The trigger timestamp of each event.
[0034] The event defect calculation module is used to: calculate the event density map of the blade surface within a preset time window based on the original event stream data, and calculate the defect significance value based on the event density map; The defect area calculation module is used to: calculate the blade defect area based on the region where the defect significance value exceeds a preset significance threshold, wherein a preset time window corresponds to one blade defect area; The safe life prediction module is used to: predict the remaining safe life of the blade based on a sequence of blade defect areas composed of multiple consecutive preset time windows.
[0035] In this embodiment, by placing an event camera around the vertical axis wind turbine blade and ensuring its detection range covers the surface of the blade during rotation, raw event stream data triggered by changes in blade surface brightness can be acquired during blade operation. Compared to traditional frame cameras that acquire complete images at a fixed frame rate, this reduces motion blur and ghosting issues during high-speed rotating blade image acquisition. An event density map of the blade surface within a preset time window is calculated based on the raw event stream data output by the event camera. This event density map reflects the frequency of brightness changes at different locations on the blade surface within the corresponding time window, providing a stable data foundation for subsequent defect saliency calculations and avoiding direct processing of the raw event stream data. This addresses the issues of data dispersion and asynchronous timing caused by vibration detection. Calculating defect significance values based on event density maps and quantifying these anomalies improves the accuracy of defect area identification. Calculating the blade defect area based on areas where the defect significance value exceeds a preset significance threshold transforms defect detection results from simply determining the presence of a defect into quantifiable defect area information. Furthermore, by correlating defect areas across multiple time windows in chronological order, it reflects the expansion trend of defects over time. This not only detects the current defect state but also predicts future defect trends, solving the problems of large interference factors, high labor costs, and large errors associated with vibration detection in existing technologies.
[0036] In one embodiment, the raw event stream data includes multiple asynchronous events; The event defect calculation module performs the following steps: Calculates the event density map of the blade surface within a preset time window based on the original event stream data. Step S211: Count the number of asynchronous events falling into the same pixel position within the preset time window; Step S212: Calculate the event density value at the corresponding pixel position based on the number of asynchronous events and the time length of the preset time window, and generate an event density map.
[0037] In this embodiment, the event defect calculation module uses a preset time window as the statistical unit to filter asynchronous events in the original event stream data, identifying target asynchronous events whose trigger timestamps fall within the preset time window. Then, based on the pixel coordinates of the target asynchronous events, the module counts the number of asynchronous events falling at the same pixel location, and divides the number of asynchronous events corresponding to that pixel location by the duration of the preset time window to obtain the event density value corresponding to that pixel location. After performing the above statistics and calculations on multiple pixel locations within the imaging plane of the event camera, an event density map is generated according to the spatial arrangement of each pixel location. Thus, the asynchronous, discrete original event stream data output by the event camera is converted into an event density map with a spatial distribution relationship, quantifying the brightness change frequency of different locations on the blade surface within the preset time window, facilitating subsequent calculation of defect significance values based on the event density map. Each pixel location in the event density map corresponds to an event density value, which characterizes the event triggering frequency of that pixel location within the preset time window.
[0038] In one embodiment, and in another embodiment, the preset time window is an adaptive event statistics time window determined based on the actual blade rotational speed. Specifically, the event defect calculation module obtains the actual blade angular velocity corresponding to the current preset time window and adjusts the duration of the event statistics time window based on the actual blade angular velocity, so that the angular displacement traversed by the blade within the event statistics time window remains consistent at different rotational speeds.
[0039] For example, the preset angular displacement of the blade within an event statistics time window is set as follows: The current actual angular velocity of the blade is The duration of the event statistics time window is then... Determine according to the following formula:
[0040] in, The duration of the event statistics time window is determined based on the current actual angular velocity of the blade. For the preset angular displacement, This is the current actual angular velocity of the blade.
[0041] When the actual angular velocity of the blade increases, the event defect calculation module decreases the duration of the event statistics time window; when the actual angular velocity of the blade decreases, the event defect calculation module increases the duration of the event statistics time window. This setting ensures that the event density maps generated under different speed conditions correspond to approximately the same range of rotational angular displacement on the blade surface, reducing the impact of blade speed variations on the event quantity statistics results.
[0042] Therefore, when wind speed fluctuations cause changes in the rotational speed of vertical axis wind turbine blades, the event density map can still stably characterize the event triggering frequency within a unit angular displacement range on the blade surface, avoiding the problem of misjudging the significance value of defects due to an increase in the number of events caused by an increase in rotational speed or a decrease in the number of events caused by a decrease in rotational speed.
[0043] In one embodiment, the event density value is calculated based on the following formula:
[0044] in, It is the event density value; It is the length of the preset time window; It is the Dirac function; This indicates whether the i-th asynchronous event falls within the pixel position (x, y); It is an indicator function.
[0045] In this step, when the i-th asynchronous event falls at pixel position (x, y), The value is 1 if it is not 0 otherwise.
[0046] In one embodiment, the event defect calculation module performs the following steps: calculating the defect significance value based on the event density map. Step S221: Obtain the event density value corresponding to each pixel position in the event density map; Step S222: Obtain the normal reference event density value corresponding to each pixel position in the normal reference event density map, wherein the normal reference event density map is preset; Step S223: For any pixel location, calculate the defect significance value corresponding to the pixel location based on the difference between the event density value corresponding to the pixel location and the normal baseline event density value.
[0047] In this embodiment, considering that the vertical axis wind turbine blades rotate continuously around the vertical axis during operation, the same pixel position in the event camera imaging plane may correspond to different actual positions on the blade surface at different times. To ensure that the comparison result between the current event density map and the normal reference event density map can reflect blade surface defects, rather than event density differences caused by changes in the blade rotation position, the event defect calculation module also performs blade rotation phase synchronization and spatial registration processing before calculating the defect significance value.
[0048] Specifically, the event defect calculation module acquires the rotational phase information of the blades corresponding to each preset time window. The rotational phase information includes at least one of the blade number, blade angular position, and blade rotational speed. In this embodiment, the rotational phase information is provided by the wind turbine SCADA system. Based on the rotational phase information, the event defect calculation module groups the raw event stream data collected from the same blade at the same or similar rotational phase into the same comparison group, and calculates the current event density map within the grouped comparison group.
[0049] When establishing the normal baseline event density map, the event defect calculation module also groups the historical event stream data collected in the defect-free state according to the blade number and rotation phase interval, generating normal baseline event density maps for the corresponding blade number and rotation phase interval. This setup ensures that during the current inspection process, the event defect calculation module selects and compares the corresponding normal baseline event density map with the blade number and rotation phase interval corresponding to the current preset time window.
[0050] Furthermore, the event defect calculation module also determines the effective projection area of the blade in the event camera's imaging plane based on the event camera's intrinsic parameters, the installation pose between the event camera and the blade, and the blade contour recognition results, and performs spatial registration on the event density map within the effective projection area. The spatial registration includes aligning the pixel positions corresponding to the blade surface in the current event density map with the pixel positions corresponding to the same blade surface positions in the normal reference event density map. Background areas, tower areas, or non-blade areas outside the effective projection area are not included in the defect significance value calculation.
[0051] This setting ensures that the pixel positions in the current event density map correspond to the same actual positions on the blade surface as the pixel positions in the normal baseline event density map. This reduces the impact of blade rotation and background changes on the calculation results of defect significance values, thereby improving the accuracy of defect area identification and reducing misjudgments caused by normal blade rotation.
[0052] In this embodiment, the normal baseline event density map is preset. The normal baseline event density map is preset when the blade is in a defect-free state or when it is manually confirmed to be in an abnormal state.
[0053] Specifically, when the vertical axis wind turbine is in normal operation, the event camera collects raw event stream data corresponding to multiple preset time windows. The event defect calculation module then generates multiple historical event density maps using the same calculation method as the current detection process. These historical event density maps are then averaged, weighted averaged, or statistically modeled to obtain a normal baseline event density map. Thus, the normal baseline event density map can characterize the event density distribution of the blade under normal rotation, normal illumination, and normal surface conditions. When comparing the current event density map with the normal baseline event density map, the influence of periodic brightness changes caused by normal blade rotation is reduced, thus decreasing misjudgments caused by normal motion.
[0054] After obtaining the event density map corresponding to the current preset time window, the event defect calculation module reads the event density value corresponding to each pixel position in the event density map, and obtains the normal baseline event density value corresponding to the same pixel position in the normal baseline event density map. The normal baseline event density map is used to characterize the normal event triggering frequency at the corresponding pixel position of the blade in a defect-free state.
[0055] For any pixel location, the event defect calculation module calculates the difference between the current event density value corresponding to that pixel location and the normal baseline event density value, and determines the defect saliency value corresponding to that pixel location based on the difference. After performing the above calculation on multiple pixel locations in the event density map, the defect saliency distribution corresponding to the spatial location on the blade surface is obtained.
[0056] The defect saliency value is used to characterize the degree of deviation of the current event density value from the normal baseline event density value. When defects such as cracks, scratches, or peeling exist on the blade surface, the surface reflection characteristics, edge continuity, or texture state of the corresponding area change. The number of events generated by the event camera in this area will increase or decrease compared to the normal baseline state. Therefore, the defect saliency value at the corresponding pixel position increases. By calculating the defect saliency value for each pixel position separately, the defect saliency distribution corresponding to the spatial position on the blade surface can be obtained, thereby transforming the local abnormal changes caused by defects into comparable numerical results and improving the accuracy of defect area identification.
[0057] In another embodiment, before calculating the blade defect area based on the region where the defect significance value exceeds a preset significance threshold, the defect area calculation module also performs multi-rotation cycle consistency verification on the defect candidate region.
[0058] Specifically, when a candidate pixel region with a defect significance value exceeding a preset significance threshold exists within a preset time window, the defect area calculation module maps the candidate pixel region to the same actual location region on the blade surface based on the blade number, rotation phase information, and spatial registration results. Subsequently, within multiple consecutive rotation cycles, it is determined whether the candidate pixel region repeatedly appears in the same location region on the blade surface.
[0059] If the candidate pixel region appears in the same blade surface location area in at least N rotation cycles out of M consecutive rotation cycles, then the candidate pixel region is determined as a valid defect region. If the candidate pixel region appears only in a single rotation cycle, or cannot correspond to the same blade surface location area in different rotation cycles, then the candidate pixel region is determined as a transient interference region and is not included in the blade defect area calculation. Here, M is the number of consecutive rotation cycles, N is a preset consistency threshold, and N≤M.
[0060] For example, M is 5 and N is 3. When a candidate pixel region corresponds to the same blade surface location region in at least 3 out of 5 consecutive rotation cycles, the candidate pixel region is determined to meet the multi-rotation cycle consistency condition and is regarded as a valid defect region; otherwise, the candidate pixel region is rejected as an abnormal region caused by sudden changes in illumination, background interference, raindrops, dust or instantaneous reflection.
[0061] Through the above multi-rotation cycle consistency verification, the characteristic of the periodic rotation of vertical axis wind turbine blades is utilized to distinguish the recurrence of real blade defects at the same blade surface position from instantaneous environmental interference or background disturbances, thereby reducing the defect false detection rate.
[0062] In one embodiment, the defect saliency value at time t for any pixel location (x, y) is generated based on the following formula:
[0063] in, It is the defect significance value corresponding to the pixel position (x, y) at time t. It is the event density value at pixel position (x, y) in the event density map at time t. It is the normal reference event density value corresponding to the pixel position (x,y) in the normal reference event density map at time t.
[0064] In another embodiment, when there are multiple event cameras and the detection ranges of the multiple event cameras overlap, the event defect calculation module also performs multi-event camera consistency verification on the defect candidate regions within the overlapping fields of view.
[0065] Specifically, the event defect calculation module, based on a pre-established spatial registration relationship, determines the corresponding blade surface location regions on the blade surface for each event camera's detected defect candidate region. The spatial registration relationship characterizes the correspondence between pixel positions in the event camera's imaging plane and the location regions on the blade surface. This spatial registration relationship is pre-determined based on the calibration parameters of each event camera, the installation position relationship between each event camera and the blade, and the blade rotation phase information corresponding to the current preset time window. The establishment of the spatial registration relationship is existing technology, and those skilled in the art should know how to establish it; therefore, this embodiment will not elaborate further.
[0066] For the same blade surface location region located within the overlapping field of view of multiple event cameras, the event defect calculation module obtains the defect saliency value corresponding to different event cameras, and determines whether different event cameras have detected defect candidate regions in the blade surface location region.
[0067] When at least two event cameras detect a region on the same blade surface with a defect significance value exceeding a preset significance threshold, the event defect calculation module identifies the region as a multi-view consistent defect region. When only one event camera detects a defect candidate region, while other event cameras with overlapping fields of view do not detect the corresponding anomaly, the event defect calculation module reduces the confidence of the defect candidate region or treats the defect candidate region as a region to be verified and does not directly participate in the blade defect area calculation.
[0068] For example, both the first event camera and the second event camera cover the central region of the blade, and the event defect calculation module maps the defect saliency distributions corresponding to the two cameras to the same blade surface coordinate system. When the first event camera detects a significant anomaly in coordinate region P on the blade surface, and the second event camera also detects a significant anomaly in the same coordinate region P, the coordinate region P is determined to meet the multi-event camera consistency condition and is taken as a valid defect region.
[0069] This setup utilizes repeated observations of the same blade surface area from different perspectives to reduce the probability of false detections caused by camera occlusion or localized reflections during a single event, thereby improving the reliability of defect candidate area confirmation.
[0070] In one embodiment, the event defect calculation module also performs cumulative detection on pixel locations that do not exceed a preset saliency threshold but meet a preset sub-threshold condition, thereby identifying early minor defects on the blade surface.
[0071] Specifically, the event defect calculation module sets a first salience threshold and a second salience threshold, wherein the first salience threshold is less than the second salience threshold, and the second salience threshold is used to determine the defect candidate pixels participating in the calculation of the blade defect area.
[0072] For any blade surface location region, when the defect significance value corresponding to the blade surface location region is greater than the first significance threshold and less than the second significance threshold, the event defect calculation module does not immediately determine the blade surface location region as a valid defect region, but instead performs subthreshold anomaly counting on the blade surface location region.
[0073] If, within multiple consecutive preset time windows, the number of times the same blade surface location area meets the preset subthreshold condition reaches a preset cumulative number threshold, then the event defect calculation module determines the blade surface location area as an early suspected defect area and outputs early defect warning information.
[0074] If the preset subthreshold condition is only occasionally met in the same area of the blade surface, and the preset cumulative number of occurrences is not reached, then no early defect warning information will be output.
[0075] For example, the first significance threshold is 60% of the second significance threshold, and the preset cumulative number threshold is 6 times within 10 consecutive preset time windows. When a certain blade surface location area has more than 6 defect significance values between the first significance threshold and the second significance threshold within 10 consecutive preset time windows, the blade surface location area is determined as an early suspected defect area.
[0076] This configuration allows for the identification of subtle, recurring anomalies without lowering the preset significance threshold used for area calculation. This avoids increasing false positives caused by directly lowering the threshold and enables the early detection of defects such as crack initiation and shallow wear before they develop into obvious area anomalies, thus improving the lead time for blade defect detection.
[0077] In another embodiment, to further improve the accuracy of early warning of minor defects, the event defect calculation module also calculates the early defect risk value corresponding to the blade surface location region based on the difference of the control area under the same operating conditions, the subthreshold anomaly trend and the spatial concentration.
[0078] Specifically, for any blade surface location region to be detected, the event defect calculation module obtains the defect significance value corresponding to the blade surface location region within multiple consecutive rotation cycles or multiple consecutive preset time windows, and determines the blade surface location region as the target region.
[0079] Meanwhile, the event defect calculation module selects a reference area from the same preset time window. The reference area is an area that is in the same or similar operating conditions as the target area.
[0080] The control area includes adjacent blade surface areas located in the same blade partition as the target area and not identified as defect candidate areas.
[0081] For the g-th preset time window, the event defect calculation module calculates the regional statistical value of the defect significance value in the target area and the regional statistical value of the defect significance value in the control area.
[0082] The regional statistical value is the average, median, or preset quantile value of the defect significance value of each pixel position within the corresponding region.
[0083] Subsequently, the event defect calculation module obtains the significance value of the control difference for the target area based on the difference between the regional statistical values of the target area and the regional statistical values of the control area.
[0084] In this embodiment, the significance value of the control difference is calculated as follows:
[0085] in, It is the significance value of the control difference corresponding to the g-th preset time window. It is the statistical value of the significant defect area in the target region within the g-th preset time window. It is the statistical value of the significant defect area in the control area within the g-th preset time window. It is the standard deviation of the significance value of defects within the control area. It is a preset constant to prevent the denominator from being zero.
[0086] The event defect calculation module obtains a control difference significance sequence based on the control difference significance values corresponding to multiple consecutive preset time windows, and performs trend analysis on the control difference significance sequence.
[0087] When the significance of the control difference sequence shows a continuous increasing trend, it indicates that the degree of abnormality of the target region relative to the control region under the same working conditions is gradually increasing. For example, the event defect calculation module performs a linear fit on the significance of the control difference sequence and uses the fit slope as the subthreshold abnormality trend coefficient.
[0088] Further, the event defect calculation module calculates the spatial concentration of subthreshold anomalous pixels within the target area. Specifically, the event defect calculation module identifies pixels with defect significance values greater than a first significance threshold and less than a second significance threshold as subthreshold anomalous pixels, wherein the first significance threshold is less than the second significance threshold, and the second significance threshold is used to determine candidate defect pixels for participating in the blade defect area calculation. The event defect calculation module performs connected component analysis on the subthreshold anomalous pixels within the target area to determine the number of pixels in the largest subthreshold connected region, and determines the spatial concentration based on the ratio of the number of pixels in the largest subthreshold connected region to the total number of subthreshold anomalous pixels within the target area.
[0089] When a target region simultaneously meets the following conditions within multiple consecutive preset time windows, the event defect calculation module identifies the target region as an early high-risk defect region: the significance value of the control difference corresponding to the target region is continuously higher than the preset difference threshold; the subthreshold abnormal trend coefficient is greater than the preset trend threshold; and the spatial concentration is greater than the preset concentration threshold.
[0090] Specifically, in this embodiment, the preset difference threshold, the preset trend threshold, and the preset concentration threshold are predetermined based on historical detection data collected during the calibration phase.
[0091] The historical inspection data includes data collected when the blades are in a defect-free state, as well as early defect sample data confirmed by manual inspection.
[0092] The preset difference threshold is used to determine whether the degree of abnormality of the target region relative to the control region is significant. The preset difference threshold is determined based on the normal fluctuation range of the significance value of the difference in the control region under defect-free conditions. For example, under defect-free conditions, the significance value of the difference in the control region obtained from multiple detections is usually less than 1.0, so the preset difference threshold is set to 1.0. When the significance value of the difference in the control region corresponding to the target region is greater than 1.0 within multiple consecutive preset time windows, it indicates that the target region has a continuous abnormality relative to the control region.
[0093] The preset trend threshold is used to determine whether the degree of anomaly in the target region increases over time. The preset trend threshold can be determined based on the normal fluctuation trend of the control difference significance value over time in a defect-free state. For example, in a defect-free state, the fitting slope of the control difference significance sequence typically does not exceed 0.02, so the preset trend threshold is set to 0.02. When the subthreshold anomaly trend coefficient corresponding to the target region is greater than 0.02, it indicates that the degree of anomaly in the target region has a continuously increasing trend.
[0094] The preset concentration threshold is used to determine whether subthreshold abnormal pixels form a local cluster in space. The preset concentration threshold can be determined based on the spatial concentration of random noise regions in a defect-free state and the spatial concentration of early defect regions confirmed manually. For example, the spatial concentration of random noise regions is typically less than 0.40, while the spatial concentration of early cracks or early coating peeling regions is typically greater than 0.60; therefore, the preset concentration threshold is set to 0.50.
[0095] For example, if a target region has a significance value greater than 1.0 for each of 10 consecutive preset time windows, a subthreshold anomaly trend coefficient of 0.035 (greater than the preset trend threshold of 0.02), and the total number of subthreshold anomaly pixels in the target region is 40, with the largest subthreshold connected region containing 28 pixels, then the spatial concentration is 28 / 40 = 0.70, greater than the preset concentration threshold of 0.50. In this case, the event defect calculation module identifies the target region as an early high-risk defect region.
[0096] If another target area exhibits subthreshold anomalies within a preset time window, but its control difference significance value does not consistently exceed 1.0, or its subthreshold anomaly trend coefficient is less than 0.02, or its spatial concentration is less than 0.50, then that target area will not be identified as an early high-risk defect area. This setting avoids false alarms caused by random noise and transient reflections.
[0097] If the target area only occasionally shows subthreshold abnormalities, or if the target area and the control area show a significant increase simultaneously, then the target area will not be identified as an early high-risk defect area.
[0098] This setup doesn't rely solely on the number of subthreshold anomalies for early warning; instead, it combines the differences between the target area and a control area under similar conditions, the increasing trend of anomalies over time, and the spatial clustering of anomalous pixels. This allows for the differentiation of subtle, localized anomalies such as crack initiation, shallow wear, or early coating peeling from overall changes in illumination and cloud cover, thereby improving the accuracy of early defect warnings without lowering the defect area calculation threshold.
[0099] In one embodiment, the defect area calculation module performs the following steps: Calculates the blade defect area based on the region where the defect significance value exceeds a preset significance threshold. Step S310: Determine the number of pixels whose defect salience value exceeds a preset salience threshold; Step S320: Calculate the defect area of the blade based on the number of pixels and the actual physical area of the blade corresponding to a single pixel.
[0100] In this embodiment, in another embodiment, the defect area calculation module performs connected region filtering on the defect candidate pixels before counting the number of pixels whose defect significance value exceeds a preset significance threshold.
[0101] Specifically, the defect area calculation module identifies pixel positions where the defect salience value exceeds a preset salience threshold as candidate defect pixels and pixel positions where the defect salience value does not exceed the preset salience threshold as non-defect pixels. Then, based on the pixel coordinates of each candidate defect pixel in the event camera's imaging plane, it determines whether the candidate defect pixels satisfy a preset adjacency condition, and merges the candidate defect pixels that satisfy the preset adjacency condition into the same candidate connected region, thereby obtaining one or more candidate connected regions. The preset adjacency condition includes a four-neighbor adjacency condition or an eight-neighbor adjacency condition; the four-neighbor adjacency condition means that two candidate defect pixels are adjacent in the horizontal or vertical direction, and the eight-neighbor adjacency condition means that two candidate defect pixels are adjacent in the horizontal, vertical, or diagonal direction.
[0102] For any candidate connected region, the defect area calculation module determines whether the candidate connected region meets the preset defect region condition based on the number of pixels in the candidate connected region. In this embodiment, the preset defect region condition is specifically: the number of pixels in the candidate connected region is not less than a preset minimum pixel count threshold.
[0103] When the number of pixels in a candidate connected region is less than the preset minimum pixel count threshold, the candidate connected region is identified as a noise region and removed from the defective candidate pixels; when the number of pixels in a candidate connected region is not less than the preset minimum pixel count threshold, the candidate connected region is identified as a valid defective region.
[0104] For example, the preset minimum pixel count threshold is 5 pixels. If a candidate connected region contains 3 candidate defect pixels, the number of pixels in the candidate connected region is less than the preset minimum pixel count threshold, and the candidate connected region is determined as a noise region and is not included in the blade defect area calculation; if another candidate connected region contains 12 candidate defect pixels, the number of pixels in the candidate connected region is not less than the preset minimum pixel count threshold, and the candidate connected region is determined as a valid defect region, and the candidate defect pixels in the valid defect region are included in the blade defect area calculation.
[0105] When calculating the blade defect area, the defect area calculation module only counts the number of candidate defect pixels within the valid defect area, and does not count the number of candidate defect pixels corresponding to the eliminated noise areas. It should be understood that the blade defect area is calculated based on the number of candidate defect pixels within the valid defect area and the actual physical area of the blade corresponding to a single pixel.
[0106] This approach avoids directly including a single or a small number of isolated abnormal pixels in the blade defect area calculation, reducing area calculation errors caused by random triggering of event cameras, instantaneous reflections, or local noise, thereby improving the stability of the blade defect area calculation results.
[0107] In this embodiment, the defect area calculation module determines the pixel positions where the defect significance value exceeds a preset significance threshold as defect candidate pixels, and counts the number of defect candidate pixels within the same preset time window.
[0108] The actual physical area of the blade corresponding to a single pixel is predetermined based on the intrinsic parameters of the event camera, the installation pose between the event camera and the blade, and the projection ratio of the blade surface in the imaging plane. Therefore, the blade defect area corresponding to the preset time window is obtained by converting the number of defect candidate pixels with the actual physical area of the blade corresponding to a single pixel. This setting ensures that the defect detection result is not merely a judgment of whether a defect exists, but further forms quantified area data, which is beneficial for subsequent assessment of defect severity.
[0109] In one embodiment, for example, if, based on camera calibration and installation pose calculations, at the current detection distance and shooting angle, a pixel in the event density map corresponds to 0.5mm horizontally and 0.5mm vertically on the blade surface, then the actual physical area of the blade corresponding to a single pixel is 0.25mm². 2 If, within a preset time window, the number of pixels whose defect significance value exceeds a preset significance threshold is 400, then the corresponding blade defect area for that preset time window is 400 × 0.25 mm. 2 =100mm 2 .
[0110] In another embodiment, when determining the actual physical area of the blade corresponding to a single pixel, the event camera is calibrated beforehand after installation to obtain intrinsic parameters such as focal length and principal point coordinates, and installation pose parameters such as distance and angle of the event camera relative to the blade detection area are obtained. Then, a known-sized calibration area on the blade surface is projected onto the imaging plane of the event camera. Based on the actual size of the calibration area and its corresponding pixel size, the correspondence between the pixel and the actual physical size of the blade is determined. For example, if a calibration area with an actual size of 20 mm × 20 mm is selected on the blade surface, and this calibration area corresponds to 40 pixels × 50 pixels in the imaging plane, then the actual size of a single pixel along the two directions is 20 / 40 = 0.5 mm and 20 / 50 = 0.4 mm, respectively.
[0111] In one embodiment, the blade defect area is calculated based on the following formula:
[0112] in, This represents the area of the blade defect. This represents the actual physical area of the blade corresponding to a single pixel. To meet The number of pixels, This is a preset significance threshold.
[0113] In this embodiment, a preset time window corresponds to a blade defect area, which represents the actual size of a significant abnormal region on the blade surface within that preset time window. The more pixels that satisfy a defect significance value greater than a preset significance threshold, the larger the corresponding blade defect area, indicating a wider range of abnormal regions on the blade surface. By using the actual physical area of the blade corresponding to a single pixel for calculation, the defect area is no longer simply expressed as the number of pixels, but is transformed into a physically meaningful area result, thus facilitating comparison with a preset defect area safety threshold.
[0114] In one embodiment, the safe life prediction module performs the following steps: predicting the remaining safe life of the blade based on a sequence of blade defect areas composed of multiple consecutive preset time windows. Step S410: Fit a linear expansion model of the defect area changing with time based on the defect area sequence of the blade; Step S420: Based on the extended model, the current blade defect area, and the preset defect area safety threshold, predict the remaining safe life of the blade.
[0115] In this embodiment, the safe life prediction module arranges the blade defect areas corresponding to each preset time window according to the chronological order of multiple consecutive preset time windows, thus obtaining a blade defect area sequence. This blade defect area sequence reflects the trend of defect area changes over time.
[0116] The safe life prediction module removes or smooths outliers from the blade defect area sequence and fits a linear expansion model of defect area changes over time based on the processed defect area sequence. This configuration avoids random errors caused by making judgments based solely on defect area within a single time window, ensuring that the life prediction results are based on continuous trends and improving the stability of the prediction results.
[0117] Specifically, in step S410, the extended model is a mathematical model used to characterize the change law of blade defect area over time. Multiple consecutive preset time windows, corresponding to their respective times and blade defect areas, are used as fitting samples for fitting. In this embodiment, the linear expansion law is used as an example. When determining that the defect expansion follows a linear law, a linear model is employed. Where b is the intercept, representing the defect area corresponding to the fitted model at t=0. The defect propagation rate is the change in defect area over time (unit: This represents the average increase in the area of blade defects per unit time, calculated using the following formula:
[0118] Where T is the number of preset time windows; It is the sequence index of the detection data; This represents the detection time corresponding to the g-th preset time window. It is in time The blade defect area calculated within the corresponding preset time window. It is the sum of the product of the total inspection time and the corresponding defect area. It is the sum of all testing times. It is the sum of the areas of all defects in the blades.
[0119] By determining the model parameters of the extended model, the functional relationship between the blade defect area and time is obtained. Based on the fitted functional relationship, the time when the blade defect area reaches the preset safe area threshold is predicted.
[0120] In one embodiment, the safe life prediction module performs the following steps: predicting the remaining safe life of the blade based on the extended model, the current blade defect area, and a preset defect area safety threshold. Step S421: Determine the defect propagation rate based on the linear propagation model; Step S422: Based on the defect propagation rate, the current blade defect area, and the preset defect area safety threshold, predict the remaining safe life of the blade.
[0121] In this embodiment, the defect propagation rate is used to represent the degree of change in the defect area of the blade per unit time.
[0122] When the extended model is a linear model, the defect propagation rate is determined by the slope of the linear model.
[0123] After obtaining the defect propagation rate, the safe life prediction module calculates the defect propagation rate together with the current blade defect area and the preset defect area safety threshold to predict the remaining time before the blade defect area reaches the safety threshold.
[0124] This setting not only identifies the current defect status but also assesses the operational risks that may result from the continued expansion of the defect.
[0125] In another embodiment, the defect propagation rate The adjustment needs to be made based on the real-time operating conditions of the wind turbine. The adjustment formula is as follows:
[0126] in, It is the defect propagation rate after correction for blade operating conditions. The average rated rotational speed of the blades corresponding to the fitted data. To monitor the average actual blade rotation speed within the monitoring period and eliminate the interference of rotation speed fluctuations on defect area statistics, it should be noted that... The value is greater than 0.
[0127] In this embodiment, the blade rated speed corresponding to the fitted data is obtained from the rated operating parameters of the wind turbine. When the same rated speed is used in each preset time window, the rated speed is determined as the average rated speed of the blade corresponding to the fitted data. The average actual speed of the blade within the monitoring period is obtained through the SCADA system of the wind turbine; the SCADA system records the actual speed of the blade within the monitoring period according to a preset sampling interval, and obtains the average actual speed of the blade within the monitoring period by calculating the arithmetic mean of each actual speed.
[0128] In one embodiment, the remaining safe life of the blade is predicted based on the following formula:
[0129] in, For the remaining safe life of the blade, To preset a safe threshold for defect area, This represents the current blade defect area. The defect propagation rate is the rate after correction based on the blade operating conditions.
[0130] In this embodiment, the preset defect area safety threshold represents the maximum allowable defect area of the blade when meeting safe operation requirements. The current blade defect area represents the defect area corresponding to the most recent preset time window. The defect expansion rate, corrected for blade operating conditions, represents the actual expansion speed of the defect area under the current operating state. By dividing the difference between the preset defect area safety threshold and the current blade defect area by the defect expansion rate corrected for blade operating conditions, the estimated remaining time before the blade defect area reaches the safety threshold can be obtained. This prediction result can provide a basis for preventive maintenance, overhaul planning, and downtime risk control of wind turbine blades, thereby improving the safety and operation and maintenance efficiency of vertical axis wind turbines.
[0131] In another embodiment, the alarm rule is: when the defect propagation rate is corrected by the blade operating conditions. Greater than the preset expansion rate When this occurs, an emergency alarm is triggered, and the preset expansion rate is set. The default value is When the remaining safe life of the blade is less than 30 days, a preventive maintenance alarm is triggered.
[0132] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.
[0133] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0134] This application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above method embodiments.
[0135] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.
[0136] This application provides a computer program product that, when run on a mobile terminal, enables the mobile terminal to execute the steps described in the various method embodiments above.
[0137] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-described embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographic device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.
[0138] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0139] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0140] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.
[0141] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0142] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application 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. Such 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 this application, and should all be included within the protection scope of this application.
[0143] One embodiment of this application also provides a computer device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-described methods.
[0144] The computer device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above description is an example of a computer device and does not constitute a limitation on the computer device. It may include more or fewer components than described above, or a combination of certain components, or different components, such as input / output devices, network access devices, etc.
[0145] The processor referred to can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.
[0146] In some embodiments, the memory may be an internal storage unit of the computer device, such as a hard drive or RAM. In other embodiments, the memory may be an external storage device of the computer device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, the memory may include both internal and external storage units of the computer device. The memory is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory can also be used to temporarily store data that has been output or will be output.
[0147] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0148] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A vertical axis wind turbine blade defect detection system based on an event camera, characterized in that, The system includes: An event data acquisition module is used to acquire raw event stream data output by an event camera, wherein the event camera is positioned around the blades of a vertical axis wind turbine, and the detection range of the event camera covers the surface of the wind turbine blades when they rotate. The event defect calculation module is used to: calculate the event density map of the blade surface within a preset time window based on the original event stream data, and calculate the defect significance value based on the event density map; The defect area calculation module is used to: calculate the blade defect area based on the area where the defect significance value exceeds a preset significance threshold, wherein a preset time window corresponds to one blade defect area; The safe life prediction module is used to predict the remaining safe life of a blade based on a sequence of blade defect areas composed of multiple consecutive preset time windows.
2. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 1, characterized in that, The raw event stream data includes multiple asynchronous events; The event defect calculation module performs the following steps: Calculates the event density map of the blade surface within a preset time window based on the original event stream data. Step S211: Count the number of asynchronous events falling into the same pixel position within the preset time window; Step S212: Calculate the event density value at the corresponding pixel position based on the number of asynchronous events and the time length of the preset time window, and generate an event density map.
3. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 2, characterized in that, The event density value is calculated based on the following formula: ; in, It is the event density value; It is the length of the preset time window; It is the Dirac function; This indicates whether the i-th asynchronous event falls within the pixel position (x, y); It is an indicator function.
4. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 1, characterized in that, The event defect calculation module performs the following steps: Calculates the defect significance value based on the event density map. Step S221: Obtain the event density value corresponding to each pixel position in the event density map; Step S222: Obtain the normal reference event density value corresponding to each pixel position in the normal reference event density map, wherein the normal reference event density map is preset; Step S223: For any pixel location, calculate the defect significance value corresponding to the pixel location based on the difference between the event density value corresponding to the pixel location and the normal baseline event density value.
5. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 4, characterized in that, The defect significance value at any pixel location (x, y) at time t is generated based on the following formula: ; in, It is the defect significance value corresponding to the pixel position (x, y) at time t. It is the event density value at pixel position (x, y) in the event density map at time t. It is the normal reference event density value corresponding to the pixel position (x,y) in the normal reference event density map at time t.
6. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 1, characterized in that, The defect area calculation module performs the following steps: Based on the regions where the defect significance value exceeds a preset significance threshold, it calculates the blade defect area. Step S310: Determine the number of pixels whose defect salience value exceeds a preset salience threshold; Step S320: Calculate the defect area of the blade based on the number of pixels and the actual physical area of the blade corresponding to a single pixel.
7. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 6, characterized in that, The blade defect area is calculated using the following formula: ; in, This represents the area of the blade defect. This represents the actual physical area of the blade corresponding to a single pixel. To meet The number of pixels, This is a preset significance threshold.
8. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 1, characterized in that, The safe life prediction module performs the following steps: Based on a sequence of blade defect areas obtained from multiple consecutive preset time windows, it predicts the remaining safe life of the blade. Step S410: Fit a linear expansion model of the defect area changing with time based on the defect area sequence of the blade; Step S420: Based on the extended model, the current blade defect area, and the preset defect area safety threshold, predict the remaining safe life of the blade.
9. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 8, characterized in that, The safe life prediction module performs the following steps: Based on the extended model, the current blade defect area, and the preset defect area safety threshold, it predicts the remaining safe life of the blade. Step S421: Determine the defect propagation rate based on the linear propagation model; Step S422: Based on the defect propagation rate, the current blade defect area, and the preset defect area safety threshold, predict the remaining safe life of the blade.
10. The vertical axis wind turbine blade defect detection system based on an event camera according to claim 9, characterized in that, Predict the remaining safe life of the blades based on the following formula: ; in, For the remaining safe life of the blade, To preset a safe threshold for defect area, This represents the current blade defect area. The defect propagation rate is the rate after correction based on the blade operating conditions.
