A wind power torsion-resistant cable sheath dynamic wear visual detection method

Abstract

The application relates to the technical field of wind turbine state detection, and discloses a wind power torsion-resistant cable sheath dynamic wear visual detection method, wherein the method comprises the following steps: acquiring a plurality of low-resolution image sets and sub-pixel level spatial offset parameters corresponding to each frame; fusing the plurality of low-resolution image sets into a single high-resolution reconstructed image based on the sub-pixel level spatial offset parameters; calculating multi-direction gradient energy of the single high-resolution reconstructed image and a standard embossed texture template respectively, generating a directional texture degradation vector through direction-by-direction channel comparison; calculating an anisotropy index based on the directional texture degradation vector, determining whether there is early directional wear, and extracting a main wear direction angle.

Description

Technical Field

[0001] This invention relates to the field of wind turbine condition monitoring technology, specifically to a visual inspection method for dynamic wear of wind turbine torsion-resistant cable sheaths. Background Technology

[0002] When a wind turbine is yawing, the torsion-resistant cable twists repeatedly inside the tower with the yaw motion, resulting in periodic sliding friction between the cable sheath surface and the guiding device. The sheath surface usually has a regular diamond or grid-like embossed texture, and the height and shape of its ridges can serve as direct indicators of the degree of wear.

[0003] Existing dynamic wear visual inspection methods typically involve fixing a camera inside the tower, acquiring images of the cable sheath surface during yaw, and then comparing the acquired images with a standard template using scalar indicators such as structural similarity to assess the degree of texture degradation.

[0004] This method has two limitations. First, the installation space inside the tower is limited, and the camera installation distance and lens focal length are physically constrained. The spatial resolution of a single frame image is insufficient to distinguish the details of height changes in each direction within the width of a single ridge, and early micro-wear marks are masked by pixel aliasing. Second, the dynamic wear of the cable sheath has significant directional characteristics, namely, the ridges along the relative sliding direction between the cable and the guide device are preferentially worn down, while the ridges in the vertical direction remain intact. Structural similarity, as a scalar indicator, cannot distinguish the differences in texture degradation in each direction. In the early stages, when only a single ridge is slightly worn down, the overall similarity decrease is not significant, leading to the missed detection of directional early wear, and the inability to infer the wear contact direction from texture degradation characteristics. Summary of the Invention

[0005] This invention provides a visual detection method for dynamic wear of wind power torsion-resistant cable sheaths, solving the technical problem in related technologies that early directional wear of wind power torsion-resistant cable sheaths is difficult to effectively identify and quantify under dynamic operating conditions.

[0006] This invention discloses a visual detection method for dynamic wear of wind power torsion-resistant cable sheaths, comprising: acquiring real-time signals from a yaw angle encoder; triggering image acquisition when the cable torsion angle passes through a preset phase anchor point; recording sub-pixel-level spatial offset parameters for each frame; accumulating a set of multiple low-resolution images over multiple yaw cycles; using the sub-pixel-level spatial offset parameters of each frame as displacement priors; fusing the set of multiple low-resolution images into a single high-resolution reconstructed image using an iterative back-projection super-resolution reconstruction algorithm; calculating multi-directional gradient energy for the single high-resolution reconstructed image and a standard embossed texture template, and generating a directional texture degradation vector by comparing each directional channel; calculating an anisotropy index based on the directional texture degradation vector; determining the presence of directional early wear when the anisotropy index exceeds a preset threshold, and extracting the angle corresponding to the directional channel with the largest normalized gradient energy decay rate as the main wear direction angle.

[0007] Furthermore, the subpixel-level spatial offset parameter consists of a lateral displacement component and an axial displacement deviation value. The lateral displacement component is obtained by geometric transformation of the angle deviation value at the acquisition time, and the axial displacement deviation value is caused by the elastic relaxation of the cable. The angle deviation value and the axial displacement deviation value are uniformly converted into displacement in pixels during geometric transformation and then combined to form the subpixel-level spatial offset parameter.

[0008] Furthermore, the reconstruction magnification factor of the iterative back-projection super-resolution reconstruction algorithm is jointly determined by the cumulative number of frames and the spatial coverage of the offset distribution. The spatial coverage is the degree of dispersion of the sub-pixel level spatial offset parameters of each frame in the sub-pixel grid. The value of the reconstruction magnification factor is such that the reconstructed pixel scale is smaller than the width of a single ridge of the embossed texture.

[0009] Furthermore, after obtaining the single-frame high-resolution reconstructed image, a sheath surface perspective unfolding transformation is performed on the single-frame high-resolution reconstructed image based on the cable cylinder geometry prior, mapping the surface region to a planar unfolded image, and subsequent gradient calculations are performed on the planar unfolded image.

[0010] Further, the step of calculating multi-directional gradient energy for the single-frame high-resolution reconstructed image and the standard embossed texture template, and generating a directional texture degradation vector by comparing each directional channel, includes: calculating the Sobel gradient components in the horizontal and vertical directions for the single-frame high-resolution reconstructed image, obtaining the gradient magnitude and gradient direction at each pixel position; dividing the gradient direction into multiple directional channels at preset angular intervals, and statistically analyzing the spatial distribution of the gradient magnitude of the pixels belonging to each directional channel to generate a multi-directional gradient energy atlas; performing the same multi-directional gradient energy calculation on the standard embossed texture template to generate a standard multi-directional gradient energy atlas; and calculating the normalized gradient energy decay rate for each directional channel of the multi-directional gradient energy atlas and the standard multi-directional gradient energy atlas to generate the directional texture degradation vector.

[0011] Furthermore, the preset angle interval is 30 degrees, and the gradient direction ranges from 0 to 180 degrees, corresponding to 6 directional channels.

[0012] Furthermore, the normalized gradient energy decay rate is calculated as follows: for each directional channel, the gradient energy map of that directional channel is summed in space to obtain the energy value of that directional channel. The energy value is then subtracted from the ratio of the energy value of that energy value to the energy value of the corresponding directional channel in the standard multi-directional gradient energy map set. The resulting difference is the normalized gradient energy decay rate of that directional channel. The directional texture degradation vector is composed of the normalized gradient energy decay rates of all directional channels arranged in directional order.

[0013] Furthermore, before calculating the anisotropy index, it is first determined whether the mean of all components in the directional texture degradation vector exceeds a preset minimum energy threshold. When the mean does not exceed the preset minimum energy threshold, it is directly determined that there is no significant wear. When the mean exceeds the preset minimum energy threshold, the anisotropy index is then calculated, wherein the anisotropy index is the ratio of the difference between the maximum and minimum components in the directional texture degradation vector to the mean.

[0014] Furthermore, after determining the existence of directional early wear, the method further includes: taking multiple profile lines at equal intervals on the image along the main wear direction angle; obtaining the peak and valley response values ​​of the ridge position based on the grayscale distribution on each profile line to generate a ridge height profile sequence; statistically analyzing the peak values ​​of each profile in the ridge height profile sequence; fitting a ridge height decay curve based on the peak decay trend to estimate the current remaining height of the texture ridge and the wear progress rate; calculating the expected number of remaining operating cycles for complete wear based on the ratio of the current remaining height of the texture ridge to the wear progress rate; and generating a dynamic wear diagnosis report that includes the main wear direction angle, the current remaining height of the texture ridge, and the expected number of remaining operating cycles for complete wear.

[0015] This invention discloses a dynamic wear visual inspection system for wind power torsion-resistant cable sheaths, used to execute the aforementioned dynamic wear visual inspection method for wind power torsion-resistant cable sheaths. The system includes: a multi-frame image acquisition module, used to acquire real-time signals from a yaw angle encoder, triggering image acquisition when the cable torsion angle passes a preset phase anchor point, recording sub-pixel-level spatial offset parameters for each frame, and accumulating a set of multiple low-resolution images over multiple yaw cycles; a super-resolution reconstruction module, used to use the sub-pixel-level spatial offset parameters of each frame as displacement priors, and merging the set of multiple low-resolution images into a single-frame high-resolution reconstructed image using an iterative back-projection super-resolution reconstruction algorithm; a directional degradation analysis module, used to calculate multi-directional gradient energy for the single-frame high-resolution reconstructed image and a standard embossed texture template, respectively, and generating a directional texture degradation vector by comparing each directional channel; and a wear determination module, used to calculate an anisotropy index based on the directional texture degradation vector, and when the anisotropy index exceeds a preset threshold, determining the presence of early directional wear, and extracting the angle corresponding to the directional channel with the largest normalized gradient energy decay rate as the main wear direction angle.

[0016] The present invention has the following beneficial effects:

[0017] This invention utilizes a yaw angle encoder to trigger data acquisition, accumulating multiple frames of low-resolution images at the same phase anchor point across multiple yaw cycles. Due to differences in encoder triggering time and cable slack, each frame naturally carries a sub-pixel-level spatial offset. Using this sub-pixel-level spatial offset as a displacement prior, an iterative back-projection super-resolution reconstruction algorithm fuses the multiple low-resolution image sets into a single high-resolution reconstructed image. This ensures that the reconstructed pixel scale is smaller than the width of a single ridge, allowing for independent resolution of ridge height variations in each direction. Without replacing the hardware of the image acquisition camera inside the tower, pixel aliasing is eliminated, and the limitation of single-frame physical resolution is overcome.

[0018] This invention elevates texture degradation assessment from scalar structural similarity to a vectorized analysis composed of multi-directional gradient energy decay rates. By independently calculating the normalized gradient energy decay rate for each directional channel and generating a directional texture degradation vector, the anisotropy index can capture ridge wear phenomena occurring only along specific directions. In the early stages of slight ridge wear in only a single direction, the anisotropy index responds significantly due to the difference in normalized gradient energy decay rates between directions, thus avoiding missed detection of early directional wear. Simultaneously, the extraction of the principal wear direction angle provides a quantitative basis for determining the wear contact direction between the cable and the guide device. Attached Figure Description

[0019] Figure 1 This is a flowchart of the dynamic wear visual inspection method for wind power torsion resistant cable sheaths provided in an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the distribution of sub-pixel level spatial offset parameters for each frame provided in an embodiment of the present invention;

[0021] Figure 3 This is a schematic diagram comparing the pixel scale and the number of pixels covered by ridges before and after super-resolution reconstruction according to an embodiment of the present invention;

[0022] Figure 4 This is a schematic diagram showing the comparison of channel gradient energy in different directions provided in an embodiment of the present invention;

[0023] Figure 5 This is a schematic diagram of the normalized gradient energy decay rate of each channel provided in the embodiments of the present invention;

[0024] Figure 6 This is a schematic diagram of the average peak value of the ridge height and the fitted curve for each yaw cycle provided in the embodiments of the present invention;

[0025] Figure 7 This is a schematic diagram of the ridge height attenuation trend and remaining lifetime estimation during each yaw cycle provided in the embodiments of the present invention;

[0026] Figure 8 This is a schematic diagram of the scatter distribution of angular deviation values ​​and axial displacement deviation values ​​provided in an embodiment of the present invention;

[0027] Figure 9 This is a schematic diagram comparing the anisotropy index and the wear determination threshold provided in the embodiments of the present invention. Detailed Implementation

[0028] Example 1

[0029] When a wind turbine yaws, the torsion-resistant cable inside the tower twists repeatedly with the yaw motion. Periodic sliding friction occurs between the cable sheath surface and the guiding device. The sheath surface typically has a regular diamond or grid-like embossed texture, and the height and shape of its ridges can serve as direct indicators of wear. Existing dynamic wear visual inspection methods usually involve fixing a camera inside the tower to acquire images of the cable sheath surface during yaw, and then comparing the acquired images with a standard template using scalar indicators such as structural similarity to assess the degree of texture degradation.

[0030] This dynamic wear visual inspection method has two limitations. First, the installation space inside the tower is limited, and the camera installation distance and lens focal length are physically constrained. The spatial resolution of a single frame image is insufficient to distinguish the details of height changes in each direction within the width of a single ridge, and early micro-wear marks are masked by pixel aliasing. Second, the dynamic wear of the cable sheath has significant directional characteristics, namely, the ridges along the relative sliding direction between the cable and the guide device are preferentially worn down, while the ridges in the vertical direction remain intact. Structural similarity, as a scalar indicator, cannot distinguish the differences in texture degradation in each direction. In the early stage, when only a single ridge is slightly worn down, the overall similarity decrease is not significant, directional early wear is missed, and the wear contact direction cannot be inferred from the texture degradation characteristics.

[0031] In this embodiment, an image acquisition camera is fixedly installed inside the tower, facing the surface of the cable sheath. The yaw system is equipped with a yaw angle encoder for real-time output of the yaw angle signal.

[0032] A visual inspection method for dynamic wear of wind power torsion-resistant cable sheath according to an embodiment of the present invention, such as... Figure 1 As shown, it includes the following steps:

[0033] Step 1: Obtain a set of multiple low-resolution images and the sub-pixel level spatial offset parameters corresponding to each frame.

[0034] The real-time signal from the yaw angle encoder is acquired to calculate the current torsion angle of the cable. When the torsion angle passes through the preset phase anchor point, the image acquisition camera is triggered to capture a frame of the sheath surface image. Simultaneously, the angular deviation value and axial displacement deviation value at this acquisition moment are recorded, generating sub-pixel-level spatial offset parameters corresponding to that frame image. Multiple low-resolution images at the same phase anchor point are accumulated over multiple yaw cycles to form a multi-frame low-resolution image set.

[0035] It should be noted that the aforementioned angular deviation and axial displacement deviation values ​​are caused by encoder trigger time differences and cable elastic relaxation, respectively. During different yaw cycles, the actual spatial position of the cable when passing the same phase anchor point exhibits a slight random offset. This offset is within the sub-pixel range, causing slight differences in the sampling position of the sheath surface texture in each frame. The sub-pixel level spatial offset parameter corresponding to each frame consists of the lateral displacement component obtained through geometric transformation of the angular deviation value and the axial displacement deviation value. The unit of the angular deviation value is angle, and the unit of the axial displacement deviation value is length. Both are uniformly converted to pixel-level displacement during geometric transformation to eliminate dimensional differences before being combined into the sub-pixel level spatial offset parameter.

[0036] Step 2: Based on subpixel-level spatial offset parameters, the set of multiple low-resolution images is fused into a single high-resolution reconstructed image.

[0037] The subpixel-level spatial offset parameters corresponding to each frame are used as displacement priors and input into the iterative backprojection super-resolution reconstruction algorithm. The iterative backprojection super-resolution reconstruction algorithm takes a set of multiple low-resolution images and the subpixel-level spatial offset parameters of each frame as input, and outputs a single high-resolution reconstructed image.

[0038] It should be noted that the reconstruction magnification is determined by both the cumulative number of frames and the spatial coverage of the offset distribution. Spatial coverage refers to the degree of dispersion of the subpixel-level spatial offset parameters in the subpixel grid for each frame. When the cumulative number of frames is sufficient and the offset distribution is uniformly dispersed in the subpixel grid, a higher magnification can be supported. The reconstruction magnification should be chosen such that the reconstructed pixel scale is smaller than the width of a single ridge of the embossed texture, so that the details of ridge height variations in each direction can be independently resolved.

[0039] In this embodiment, to eliminate the bending distortion introduced by the cylindrical surface of the cable sheath in the image, after obtaining a single-frame high-resolution reconstructed image, a perspective unfolding transformation of the sheath surface is performed on the single-frame high-resolution reconstructed image. Based on the geometric prior of the cable cylinder, the curved surface region is mapped to a planar unfolded image. Gradient calculations in subsequent steps are all performed on the planar unfolded image to ensure that the gradient response in different regions has a consistent spatial metric.

[0040] Step 3: Calculate the multi-directional gradient energy for the single-frame high-resolution reconstructed image and the standard embossed texture template respectively, and generate a directional texture degradation vector by comparing the channels in each direction.

[0041] Step 3 may specifically include:

[0042] Step 301: Calculate the Sobel gradient components in the horizontal and vertical directions for a single frame of high-resolution reconstructed image, and obtain the gradient magnitude and gradient direction at each pixel position.

[0043] Step 302: Divide the gradient direction into multiple directional channels according to a preset angular interval. For each directional channel, statistically analyze the spatial distribution of the gradient magnitude of the pixels belonging to that directional channel to generate a multi-directional gradient energy atlas.

[0044] It should be noted that the preset angle interval is 30 degrees. The gradient direction ranges from 0 to 180 degrees, corresponding to 6 directional channels. The gradient energy map corresponding to each directional channel records the response intensity distribution of the texture edge within that angle range.

[0045] Step 303: Perform the same multi-directional gradient energy calculation as in steps 301 and 302 on the standard embossed texture template to generate a standard multi-directional gradient energy map as a reference.

[0046] Step 304: Calculate the normalized gradient energy decay rate for each directional channel of the multi-directional gradient energy atlas and the standard multi-directional gradient energy atlas to generate a directional texture degradation vector.

[0047] It should be noted that the normalized gradient energy decay rate is calculated as follows: for the th For each directional channel, the energy value is obtained by summing the gradient energy maps of that directional channel in space. The energy value of the corresponding directional channel in the standard multi-directional gradient energy map set is denoted as... Then the normalized gradient energy decay rate of the channel in that direction for:

[0048]

[0049] in, This refers to the index number of the directional channel. For single-frame high-resolution image reconstruction in the first Gradient energy values ​​of each directional channel. For the standard embossed texture template in the first The gradient energy values ​​of each directional channel are both of the same dimension, representing the spatial accumulation of gradient magnitude, and their ratio is... It is a dimensionless number. The value ranges from 0 to 1, with larger values ​​indicating more severe texture ridge degradation in that direction. The directional texture degradation vector is calculated from the normalized gradient energy decay rate of all directional channels. The components are arranged in directional order, with each component representing the degree of preservation of texture ridges in the corresponding direction.

[0050] Step 4: Calculate the anisotropy index based on the directional texture degradation vector, determine whether there is directional early wear, and extract the main wear direction angle.

[0051] Statistical analysis was performed on each component of the directional texture degradation vector to calculate the anisotropy index. Defined as:

[0052]

[0053] in, The maximum normalized gradient energy decay rate in the directional texture degradation vector. To minimize the normalized gradient energy decay rate, This represents the mean of the normalized gradient energy decay rate for all directional channels. , and Both are dimensionless normalized gradient energy decay rates, with the same dimensions. It is a dimensionless number.

[0054] It should be noted that when When the anisotropy index approaches zero, it indicates that the normalized gradient energy decay rate of each channel is close to zero, meaning the overall texture remains intact. At this point, the anisotropy index... The calculation results are not discriminative. Therefore, when performing anisotropy index calculations... Before calculation, it should be determined first. Does it exceed the preset minimum energy threshold? ,in It is a positive number greater than zero. When When there is no significant wear, it is directly determined that there is no wear at present, and subsequent anisotropy determination is skipped; when Then calculate the anisotropy index. And compare it with a preset threshold.

[0055] When the anisotropy index When the threshold is exceeded, directional early wear is determined to exist. At this point, the angle corresponding to the direction channel with the highest normalized gradient energy attenuation rate is extracted as the principal wear direction angle. The principal wear direction angle indicates the direction in which the texture ridges are most severely worn down, i.e., the main direction of relative sliding contact between the cable and the guide device. When the anisotropy index... If the preset threshold is not exceeded, it is determined that there is no significant directional wear at present.

[0056] In this embodiment of the application, to further quantify the degree of wear progression, after determining the existence of directional early wear, the following steps are also included:

[0057] Step 401: Extract the texture ridge height profile sequence from the unfolded image along the main wear direction angle. Take multiple profile lines at equal intervals along the main wear direction angle, and obtain the peak and valley response values ​​of the ridge position based on the gray-level distribution on each profile line to generate the ridge height profile sequence.

[0058] Step 402: Statistically analyze the peak values ​​of each profile in the ridge height profile sequence, fit the ridge height decay curve based on the peak decay trend, and estimate the current remaining height of the texture ridge and the wear progress rate.

[0059] It should be noted that the unit of the wear progression rate mentioned above is the ridge height reduction per yaw cycle. The ridge height reduction curve uses the number of yaw cycles as the independent variable and the mean peak value of the ridge height profile sequence corresponding to each cycle as the dependent variable. The slope is obtained by linear least squares fitting, and this slope is the wear progression rate.

[0060] Step 403: Based on the current remaining height of the texture ridge and the wear progress rate, calculate the estimated number of remaining operating cycles to complete the wear, and generate a dynamic wear diagnosis report that includes the main wear direction angle, the current remaining height of the texture ridge, and the estimated number of remaining operating cycles to complete the wear.

[0061] It should be noted that the remaining number of operating cycles is expected to be fully ground. The calculation method is as follows:

[0062]

[0063] in, This represents the remaining height of the current texture ridge, in length. The wear rate is expressed as the ridge height reduction per yaw cycle. The remaining number of operating cycles to be expected to be completely worn down represents the number of remaining yaw cycles required for the texture ridges to be completely worn down at the current rate of wear progression.

[0064] This implementation utilizes a yaw angle encoder to trigger data acquisition, accumulating multiple frames of low-resolution images at the same phase anchor point across multiple yaw cycles. Due to differences in encoder triggering time and cable slack, each frame naturally carries a sub-pixel-level spatial offset. Using this sub-pixel-level spatial offset as a displacement prior, an iterative back-projection super-resolution reconstruction algorithm fuses the multiple low-resolution image sets into a single high-resolution reconstructed image. This ensures that the reconstructed pixel scale is smaller than the width of a single ridge, allowing for independent resolution of ridge height variations in each direction. Therefore, without replacing the hardware of the image acquisition camera inside the tower, pixel aliasing is eliminated, and the limitation of single-frame physical resolution is overcome.

[0065] Building upon this foundation, this implementation elevates texture degradation assessment from scalar structural similarity to a vectorized analysis composed of multi-directional gradient energy decay rates. By independently calculating the normalized gradient energy decay rate for each directional channel and generating directional texture degradation vectors, the anisotropy index can capture ridge wear phenomena occurring only along specific directions. In the early stages, when only a single direction of ridge wear is slight, the overall scalar index shows little change, while the anisotropy index responds significantly due to the differences in normalized gradient energy decay rates between directions, thus avoiding missed detections of early directional wear. Simultaneously, the extraction of the principal wear direction angle provides a quantitative basis for determining the wear contact direction between the cable and the guide device.

[0066] The following is an example of an application of the present invention, such as... Figure 2-9 As shown, the implementation process is as follows:

[0067] In 20XX, an offshore wind farm conducted periodic wear monitoring of the torsion-resistant cable sheath inside the tower. The wind turbine model is WTG-A, with a rated power of 5MW. The yaw system is equipped with an incremental encoder with a resolution of 0.01 degrees. An industrial camera is fixedly installed inside the tower. Due to space constraints, the lens focal length is limited, resulting in a single-frame image resolution of 640×480 pixels, corresponding to a pixel size of approximately 0.18mm on the sheath surface. The sheath surface has an embossed diamond grid texture, with a single ridge width of approximately 0.45mm and an initial ridge height of approximately 0.30mm. In one monitoring task, the system selected a yaw angle of 135 degrees as the phase anchor point, accumulating 12 low-resolution images over 12 consecutive yaw cycles. Subsequent subpixel-level spatial offset parameters were recorded for each frame, followed by super-resolution reconstruction and directional wear analysis.

[0068] In step 1, the yaw angle encoder outputs the yaw angle signal in real time. When the cable torsion angle passes the 135-degree phase anchor point, the system triggers the camera to acquire a frame of sheath surface image, simultaneously recording the angle deviation and axial displacement deviation values ​​for that frame. The angle deviation is caused by the difference in encoder trigger time, while the axial displacement deviation is caused by the cable's elastic relaxation. After geometric transformation, both are uniformly converted into lateral and longitudinal displacement components in pixels, combined to form the sub-pixel-level spatial offset parameters for that frame. After 12 yaw cycles, a set of 12 low-resolution images and the corresponding sub-pixel-level spatial offset parameter set is formed.

[0069] Table 1. Subpixel-level spatial offset parameters of each frame of low-resolution image

[0070]

[0071] In step 2, the subpixel-level spatial offset parameters of each frame in Table 1 are used as displacement priors and input into the iterative back-projection super-resolution reconstruction algorithm. After fusing 12 low-resolution images (640×480 pixels, pixel scale 0.18mm), a single high-resolution reconstructed image is output. The offsets of the 12 frames are relatively evenly distributed in the subpixel grid, and the spatial coverage meets the requirement of a 3x magnification. The reconstructed image resolution is 1920×1440 pixels, corresponding to a pixel scale of approximately 0.06mm, which is smaller than the width of a single ridge (0.45mm). Details of ridge height changes in each direction can be independently distinguished. Subsequently, a sheath surface perspective unfolding transformation is performed on the high-resolution reconstructed image. Based on the cable cylinder geometry prior, the curved surface region is mapped to a planar unfolded image for subsequent gradient calculations.

[0072] Table 2 Comparison of image parameters before and after super-resolution reconstruction

[0073]

[0074] In step 3, multi-directional gradient energy calculations are performed on the unfolded planar image and the standard embossed texture template, respectively. Step 301 calculates the horizontal and vertical Sobel gradient components for the unfolded planar image, obtaining the gradient magnitude and direction at each pixel location. Step 302 divides the gradient direction (0 to 180 degrees) into 6 directional channels at 30-degree intervals (Channel 1: 0 to 30 degrees, Channel 2: 30 to 60 degrees, Channel 3: 60 to 90 degrees, Channel 4: 90 to 120 degrees, Channel 5: 120 to 150 degrees, Channel 6: 150 to 180 degrees), statistically analyzing the spatial distribution of the gradient magnitude of the assigned pixels for each channel, generating a multi-directional gradient energy atlas. Step 303 performs the same calculation on the standard embossed texture template, generating a standard multi-directional gradient energy atlas. Step 304 follows the formula... The normalized gradient energy decay rate is calculated for each directional channel to generate a directional texture degradation vector.

[0075] In this test, the ridges on the sheath surface along the yaw sliding direction (corresponding to channel 2, i.e., 30 to 60 degrees) were significantly worn down, while the ridges in the vertical direction (corresponding to channel 5, i.e., 120 to 150 degrees) remained intact.

[0076] Table 3. Channel gradient energy and normalized gradient energy decay rate in each direction

[0077]

[0078] Directional texture degradation vector is .

[0079] In step 4, first calculate

[0080] Exceeding the preset minimum energy threshold Then proceed to the anisotropy index calculation. (Channel 2) (Channel 5), Substitute into the formula:

[0081]

[0082] The preset threshold is 1.5. If the threshold is exceeded, directional early wear is determined to exist. The main wear direction angle is 45 degrees (center angle in the 30 to 60 degree range) corresponding to channel 2, indicating that the main direction of relative sliding contact between the cable and the guide device is 45 degrees.

[0083] Then, steps 401 to 403 are executed. Step 401 involves taking eight equally spaced profile lines on the unfolded image along a 45-degree direction, and obtaining the peak and valley response values ​​of the ridge position based on the gray-level distribution on each profile line to generate a ridge height profile sequence. Step 402 involves performing linear least squares fitting on the mean peak value of the ridge height profile sequence for each yaw cycle to obtain the ridge height decay curve.

[0084] Table 4. Average peak values ​​of ridge height profiles for each yaw cycle

[0085]

[0086] The wear progression rate is obtained from the fitted slope. mm / cycle. In step 403, the remaining height of the current texture ridge. mm, substituting into the formula:

[0087]

[0088] The system generates a dynamic wear diagnosis report, which includes: the main wear direction angle is 45 degrees, the current remaining height of the texture ridge is 0.23 mm, and the estimated number of remaining operating cycles to be completely worn down is approximately 11 to 12 yaw cycles.

[0089] Throughout the detection process, data originates from the yaw encoder signal, undergoes sub-pixel-level spatial offset parameter extraction, carries prior spatial positions from each frame, and is then used for super-resolution reconstruction, outputting a high-resolution reconstructed image with a pixel scale smaller than the ridge width. This high-resolution reconstructed image is then subjected to surface unfolding and multi-directional gradient energy analysis, with channel-by-channel attenuation rates forming directional texture degradation vectors. The statistical characteristics of each vector component drive anisotropy index determination, identifying the significantly high attenuation rate in channel 2 and locating the main wear direction. Finally, the ridge height profile sequence extracted along the main wear direction supports wear progression rate fitting and remaining life estimation, forming a diagnostic conclusion encompassing direction, degree, and trend. The output data from each step serves as direct input for the next step, ensuring a complete and coherent data flow logic.

[0090] The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A method for visually inspecting dynamic wear of the sheath of a wind power torsion-resistant cable, characterized in that, include: The real-time signal of the yaw angle encoder is acquired, and image acquisition is triggered when the cable torsion angle passes through the preset phase anchor point. Sub-pixel level spatial offset parameters of each frame are recorded, and a set of multiple low-resolution images is accumulated over multiple yaw cycles. Using the subpixel-level spatial offset parameters of each frame as displacement priors, the set of multiple low-resolution images is fused into a single high-resolution reconstructed image through an iterative back-projection super-resolution reconstruction algorithm. Multi-directional gradient energy is calculated for the single-frame high-resolution reconstructed image and the standard embossed texture template, and directional texture degradation vector is generated by comparing the channels in each direction. An anisotropy index is calculated based on the directional texture degradation vector. When the anisotropy index exceeds a preset threshold, it is determined that there is directional early wear, and the angle corresponding to the directional channel with the largest normalized gradient energy decay rate is extracted as the main wear direction angle.

2. The method for visual inspection of dynamic wear of wind power torsion-resistant cable sheaths according to claim 1, characterized in that, The subpixel-level spatial offset parameter consists of a lateral displacement component and an axial displacement deviation value. The lateral displacement component is obtained by geometric transformation of the angle deviation value at the acquisition time, and the axial displacement deviation value is caused by the elastic relaxation of the cable. The angle deviation value and the axial displacement deviation value are uniformly converted into displacement in pixels during geometric transformation and then combined to form the subpixel-level spatial offset parameter.

3. The method for visually inspecting dynamic wear of wind power torsion-resistant cable sheaths according to claim 1, characterized in that, The reconstruction magnification factor of the iterative back-projection super-resolution reconstruction algorithm is determined by the cumulative number of frames and the spatial coverage of the offset distribution. The spatial coverage is the degree of dispersion of the sub-pixel level spatial offset parameters of each frame in the sub-pixel grid. The value of the reconstruction magnification factor is such that the reconstructed pixel scale is smaller than the width of a single ridge of the embossed texture.

4. The method for visually inspecting dynamic wear of wind power torsion-resistant cable sheaths according to claim 1, characterized in that, After obtaining the single-frame high-resolution reconstructed image, a sheath surface perspective unfolding transformation is performed on the single-frame high-resolution reconstructed image based on the cable cylinder geometry prior, mapping the surface region into a planar unfolded image, and subsequent gradient calculations are performed on the planar unfolded image.

5. The method for visual inspection of dynamic wear of wind power torsion-resistant cable sheaths according to claim 1, characterized in that, The step of calculating multi-directional gradient energy for the single-frame high-resolution reconstructed image and the standard embossed texture template, and generating a directional texture degradation vector by comparing channels in each direction, includes: The Sobel gradient components in the horizontal and vertical directions are calculated for the single-frame high-resolution reconstructed image to obtain the gradient magnitude and gradient direction at each pixel position. The gradient direction is divided into multiple directional channels according to a preset angle interval. The spatial distribution of the gradient magnitude of the pixels belonging to each directional channel is statistically analyzed to generate a multi-directional gradient energy atlas. Perform the same multi-directional gradient energy calculation on the standard embossed texture template to generate a standard multi-directional gradient energy atlas; The normalized gradient energy decay rate is calculated for each directional channel of the multi-directional gradient energy atlas and the standard multi-directional gradient energy atlas to generate the directional texture degradation vector.

6. The method for visual inspection of dynamic wear of wind power torsion-resistant cable sheaths according to claim 5, characterized in that, The preset angle interval is 30 degrees, and the gradient direction ranges from 0 degrees to 180 degrees, corresponding to 6 directional channels.

7. The method for visually inspecting dynamic wear of wind power torsion-resistant cable sheaths according to claim 5, characterized in that, The normalized gradient energy decay rate is calculated as follows: for each directional channel, the gradient energy map of that directional channel is summed in space to obtain the energy value of that directional channel. The energy value is then subtracted from the ratio of the energy value of that energy value to the energy value of the corresponding directional channel in the standard multi-directional gradient energy map set. The difference obtained is the normalized gradient energy decay rate of that directional channel. The directional texture degradation vector is composed of the normalized gradient energy decay rates of all directional channels arranged in directional order.

8. The method for visual inspection of dynamic wear of wind power torsion-resistant cable sheaths according to claim 1, characterized in that, Before calculating the anisotropy index, it is first determined whether the mean of all components in the directional texture degradation vector exceeds a preset minimum energy threshold. If the mean does not exceed the preset minimum energy threshold, it is directly determined that there is no significant wear. If the mean exceeds the preset minimum energy threshold, the anisotropy index is then calculated, wherein the anisotropy index is the ratio of the difference between the maximum and minimum components in the directional texture degradation vector to the mean.

9. The method for visual inspection of dynamic wear of wind power torsion-resistant cable sheaths according to claim 1, characterized in that, After determining the presence of directional early wear, the following is also included: Multiple profile lines are taken at equal intervals on the image along the main wear direction angle. The peak and valley response values ​​of the ridge position are obtained based on the gray-scale distribution on each profile line to generate a ridge height profile sequence. The peak values ​​of each profile in the ridge height profile sequence are statistically analyzed, and the ridge height decay curve is fitted based on the peak decay trend to estimate the current remaining height of the texture ridge and the wear progress rate. Based on the ratio of the current remaining height of the texture ridge to the wear progress rate, the estimated number of remaining operating cycles for complete wear is calculated, and a dynamic wear diagnosis report is generated that includes the main wear direction angle, the current remaining height of the texture ridge, and the estimated number of remaining operating cycles for complete wear.

10. A visual inspection system for dynamic wear of wind power torsion-resistant cable sheaths, used to perform the visual inspection method for dynamic wear of wind power torsion-resistant cable sheaths as described in any one of claims 1 to 9, characterized in that, include: The multi-frame image acquisition module is used to acquire the real-time signal of the yaw angle encoder. It triggers image acquisition when the cable torsion angle passes through the preset phase anchor point, records the sub-pixel level spatial offset parameters of each frame, and accumulates a set of multiple low-resolution images over multiple yaw cycles. The super-resolution reconstruction module is used to use the sub-pixel level spatial offset parameters of each frame as displacement priors, and to fuse the set of multiple low-resolution images into a single high-resolution reconstructed image through an iterative back-projection super-resolution reconstruction algorithm. The directional degradation analysis module is used to calculate multi-directional gradient energy for the single-frame high-resolution reconstructed image and the standard embossed texture template, and generate directional texture degradation vectors by comparing channels in each direction. The wear determination module is used to calculate the anisotropy index based on the directional texture degradation vector. When the anisotropy index exceeds a preset threshold, it is determined that there is directional early wear, and the angle corresponding to the directional channel with the largest normalized gradient energy decay rate is extracted as the main wear direction angle.

Images