A high-speed rail catenary support device cantilever system image segmentation method and system

Abstract

The application relates to the technical field of image segmentation, and provides a high-speed rail overhead line support device wrist arm system image segmentation method and system, which solves the problems of inability to detect internal defects of metal components and low defect recognition precision under high-frequency vibration noise interference. The method comprises the following steps: scanning a metal component of a wrist arm system to generate a three-dimensional image, and collecting stress wave signals of the surface of the metal component under the action of dynamic load; extracting frequency spectrum features of the stress wave signals based on a cyclic load parameter of the metal component, combining geometric structure parameters of bolt connections of the metal component to determine energy distribution characteristics of the bolt connections; fusing the three-dimensional image, the energy distribution characteristics, a temperature rise abnormal area and an electric field intensity distribution to generate a multi-source data set, and combining the cyclic load parameter to generate segmentation results of fatigue cracks and insulation aging defects of the metal component. The application realizes high-precision collaborative segmentation of metal fatigue cracks and insulation aging defects of the wrist arm system.

Description

Technical Field

[0001] This invention belongs to the field of image segmentation technology, and specifically relates to an image segmentation method and system for the cantilever system of a high-speed railway catenary support device. Background Technology

[0002] The cantilever system of the high-speed railway catenary support device is exposed to a complex dynamic load environment for a long time. Its metal components are prone to microcracks due to fatigue stress, while the insulation material is prone to aging and cracking under electrothermal coupling. To ensure the reliability and safety of the high-speed railway catenary, high-precision detection of complex defects in the cantilever system is required.

[0003] Existing technologies for defect detection in high-speed rail cantilever systems include defect localization methods based on a combination of laser scanning and thermal imaging. These methods acquire surface topography data of the metal components through laser scanning and then use image processing algorithms to segment and locate defect areas within the data. However, these methods have certain limitations, such as the inability to detect internal defects in metal components through laser scanning alone, and low accuracy in defect identification under high-frequency vibration and noise interference. Summary of the Invention

[0004] This application provides an image segmentation method and system for the cantilever system of a high-speed railway catenary support device, which solves the problems in the prior art such as the inability to detect defects inside metal parts and the low accuracy of defect identification under high-frequency vibration and noise interference.

[0005] In a first aspect, this application provides an image segmentation method for a cantilever system of a high-speed railway catenary support device, including:

[0006] The metal components of the carpal tunnel system are scanned to generate three-dimensional images. Stress wave signals of the metal components under dynamic loads are collected. The electric field intensity distribution of abnormal temperature rise areas and aging areas on the surface of the metal components is detected. The aging areas characterize the aging defect areas of non-metallic components connected to or wrapped with the metal components in the carpal tunnel system.

[0007] Based on the cyclic load parameters of the metal component, the spectral characteristics of the stress wave signal are extracted, and combined with the geometric structural parameters of the bolt connection of the metal component, the energy distribution characteristics of the bolt connection in the frequency dimension are determined.

[0008] The three-dimensional image, the energy distribution characteristics, the abnormal temperature rise region, and the electric field intensity distribution are fused to generate a multi-source data set. Combined with the cyclic load parameters, the segmentation results of fatigue cracks and insulation aging defects of the metal component are generated.

[0009] Optionally, the three-dimensional image, the energy distribution characteristics, the temperature rise anomaly region, and the electric field intensity distribution are fused to generate a multi-source data set. Combined with the cyclic load parameters, segmentation results for fatigue cracks and insulation aging defects in the metal component are generated, including:

[0010] Based on the thread pitch and hole diameter in the geometric structural parameters, the three-dimensional image is divided into a multi-scale spatial mesh to obtain multiple mesh units;

[0011] Based on the trend of the frequency domain energy amplitude changing with time in the energy distribution characteristics, phase compensation processing is performed on the temperature gradient data in the temperature rise anomaly region to generate first temperature distribution data corresponding to the spatial coordinates of the grid cell.

[0012] The aging region in the first temperature distribution data is subjected to boundary topology correction processing to generate the second temperature distribution data.

[0013] Based on the correspondence between stress amplitude and cycle number in the cyclic load parameters, dynamic weight allocation is performed on the crack depth distribution and frequency domain energy amplitude within the grid cell to generate target risk level information;

[0014] Based on the gradient direction consistent region in the second temperature distribution data and the target risk level information, combined with the amplitude attenuation characteristics of the electric field intensity distribution, the segmentation results of fatigue cracks and insulation aging defects of the metal component are generated respectively.

[0015] Optionally, boundary topology correction processing is performed on the aging region in the first temperature distribution data to generate second temperature distribution data, including:

[0016] Based on the thread profile at the bolt connection, the gradient direction angle of the electric field intensity distribution is transformed to generate a first direction angle distribution;

[0017] Extract the fluctuation amplitude of the gradient direction angle corresponding to the aging region in the first temperature distribution data within the preset dynamic load frequency band, and combine it with the crack curvature of the thread profile to generate a second direction angle distribution;

[0018] Based on the thread pitch and helix angle direction at the bolt connection, multi-cycle phase matching is performed on the difference between the gradient direction angles of the first and second direction angle distributions to obtain the difference distribution;

[0019] For regions in the difference distribution where the gradient direction angle difference exceeds a preset adaptive threshold, a fracture boundary connection process is performed to generate second temperature distribution data.

[0020] Optionally, regions in the difference distribution where the gradient direction angle difference exceeds a preset adaptive threshold are subjected to fracture boundary connection processing to generate second temperature distribution data, including:

[0021] Based on the helix angle direction and crack curvature at the bolt connection, a helical path template corresponding to the number of helical turns of the thread profile is generated. At the same time, the difference distribution is meshed to generate the target mesh.

[0022] Target break points are selected from the target grid;

[0023] Based on the spiral path template, the target fracture point is extended by spiral path to generate a first spiral path and a second radius of curvature change curve.

[0024] Based on the dynamic deviation between the preset first radius of curvature change curve and the second radius of curvature change curve, the first helical path is subjected to curvature correction processing to generate a second helical path. Combined with the helical cycle number of the thread profile, the second helical path is subjected to phase compensation processing to obtain a third helical path.

[0025] Based on the gradient decay rate in the amplitude decay characteristics, the curvature smoothing process is applied to the third spiral path to generate second temperature distribution data.

[0026] Optionally, based on the dynamic deviation between the preset first radius of curvature change curve and the second radius of curvature change curve, the first helical path is subjected to curvature correction processing to generate a second helical path. Then, combined with the helical cycle number of the thread profile, the second helical path is subjected to phase compensation processing to obtain a third helical path, including:

[0027] Based on the number of helical turns of the thread profile at the bolt connection, the preset first curvature radius variation curve is divided into multiple first curvature segments;

[0028] Based on the number of helical turns of the thread profile at the bolted connection, multiple second curvature segments are extracted from the second curvature radius variation curve;

[0029] Calculate the deviation between the second curvature segment and the corresponding first curvature segment of each thread turn to obtain the nominal deviation value. When the nominal deviation value exceeds the preset adaptive threshold, correct the second curvature radius change curve to obtain the third curvature radius change curve. Combine the helix angle direction to obtain the second helical path.

[0030] Based on the number of helical cycles of the thread profile and the direction of the helical helix angle, the phase angle offset of the second helical path is calculated. Based on the phase angle offset, phase synchronization compensation is performed on the node spacing in the second helical path to obtain the third helical path.

[0031] Optionally, the spectral characteristics of the stress wave signal are extracted based on the cyclic load parameters of the metal component, and the energy distribution characteristics of the bolted connection in the frequency dimension are determined by combining the geometric structural parameters of the bolted connection of the metal component, including:

[0032] Based on the correspondence between stress amplitude and number of cycles in the cyclic load parameters of the metal component, the time-domain waveform of the stress wave signal is divided into multiple vibration period segments;

[0033] Based on the thread pitch at the bolted connection, a multi-channel filter bank is constructed with the helix angle direction consistent with that at the bolted connection.

[0034] The vibration period segment is input into the multi-channel filter bank for frequency domain energy allocation to extract the target frequency band energy amplitude in each vibration period segment and generate an initial frequency domain energy distribution.

[0035] Based on the crack curvature, the initial frequency domain energy distribution is corrected by curvature weighting to obtain the corrected frequency domain energy distribution. Based on the corrected frequency domain energy distribution, the energy distribution characteristics of the bolt connection in the frequency dimension are generated.

[0036] Optionally, the vibration period segment is input into the multi-channel filter bank for frequency domain energy allocation to extract the target frequency band energy amplitude in each vibration period segment and generate an initial frequency domain energy distribution, including:

[0037] The center frequency distribution of the multi-channel filter bank is determined based on the thread pitch and hole diameter at the bolt connection.

[0038] Based on the center frequency distribution, the time-domain waveform of each vibration period segment is converted into a frequency-domain energy distribution;

[0039] Based on the frequency domain energy distribution, the target frequency band energy amplitude corresponding to the center frequency distribution is extracted to generate an initial frequency domain energy distribution.

[0040] Secondly, this application provides an image segmentation system for the cantilever system of a high-speed railway catenary support device, comprising:

[0041] The acquisition module is used to scan the metal components of the carpal tunnel system, generate three-dimensional images, acquire stress wave signals of the metal components under dynamic load, and detect the electric field intensity distribution of abnormal temperature rise areas and aging areas on the surface of the metal components. The aging areas characterize the aging defect areas of non-metallic components connected to or wrapped with the metal components in the carpal tunnel system.

[0042] The determination module is used to extract the spectral characteristics of the stress wave signal based on the cyclic load parameters of the metal component, and combine them with the geometric structural parameters of the bolt connection of the metal component to determine the energy distribution characteristics of the bolt connection in the frequency dimension.

[0043] The fusion module is used to fuse the three-dimensional image, the energy distribution characteristics, the temperature rise anomaly region, and the electric field intensity distribution to generate a multi-source data set. Combined with the cyclic load parameters, it generates the segmentation results of fatigue cracks and insulation aging defects of the metal component.

[0044] Thirdly, this application provides a computing device, including a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the image segmentation method for the cantilever system of a high-speed rail catenary support device as described in any of the first aspects.

[0045] Fourthly, this application provides a computer storage medium storing computer program instructions thereon, which, when executed by a processor, implement the image segmentation method for the cantilever system of a high-speed rail contact network support device as described in any one of the first aspects.

[0046] The beneficial effects of this application are:

[0047] This application provides an image segmentation method for the cantilever system of a high-speed railway catenary support device. By scanning the metal components of the cantilever system and generating three-dimensional images, combined with stress wave signal analysis under dynamic load and detection of electric field intensity distribution in abnormal surface temperature rise areas and aging areas, it can identify aging and fatigue defects inside and on the surface of the metal components. Based on the cyclic load parameters of the metal components, the spectral characteristics of the stress wave signal are extracted, and considering the geometric structural parameters of the bolt connections, the energy distribution characteristics are determined, achieving precise location of potential fault points. The application of multi-data fusion technology integrates the analysis of multiple information sources, further improving the accuracy and reliability of the segmentation results, and helping to promptly detect and repair fatigue cracks and insulation aging problems in the cantilever system of the high-speed railway catenary support device.

[0048] Furthermore, the three-dimensional image is divided into multi-scale spatial meshes based on the thread pitch and hole diameter. Then, phase compensation processing is performed on the temperature gradient data to obtain the first temperature distribution data, and boundary topology correction is performed on the aging area to obtain the second temperature distribution data. Then, dynamic weight allocation is performed on the crack depth distribution based on the relationship between stress amplitude and cycle number. Finally, the segmentation results of fatigue cracks and insulation aging defects are generated by combining the amplitude attenuation characteristics of electric field intensity distribution. This improves the identification accuracy of composite defects in the cantilever system of the high-speed railway catenary support device under complex dynamic load environment, overcomes the problem that traditional methods can only detect surface morphology and cannot assess internal conditions, and enhances the ability to identify defects under high-frequency vibration and noise interference. This improves the efficiency and safety of maintenance work and ensures the stable operation of the high-speed railway catenary. Attached Figure Description

[0049] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0050] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 A flowchart illustrating an image segmentation method for a cantilever system of a high-speed railway catenary support device, provided in this application embodiment;

[0052] Figure 2 A schematic diagram of the structure of an image segmentation system for a cantilever system of a high-speed rail catenary support device provided in this application embodiment;

[0053] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0054] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0055] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] Figure 1 This application provides a flowchart of an image segmentation method for a cantilever system of a high-speed railway catenary support device, as shown in the embodiments of this application. Figure 1 As shown, the method includes:

[0058] To address the problems of existing solutions that rely solely on surface topography data, leading to an inability to identify internal metal defects and insufficient defect segmentation accuracy under high-frequency vibration and noise interference, this application provides an image segmentation method for the cantilever system of a high-speed railway catenary support device. The core idea of ​​this method is to simultaneously acquire the three-dimensional structural information of the metal components of the cantilever system, the stress wave response under dynamic load, the distribution of abnormal temperature rise areas, and the electric field intensity characteristics of aging areas. Combined with cyclic load parameters and the geometric structure of bolt connections, frequency domain energy distribution characteristics are extracted. These multi-modal information are then fused to construct a unified multi-source data representation, thereby achieving collaborative identification and accurate segmentation of fatigue cracks and insulation aging defects without relying on a single detection method. Based on this, this application provides an image segmentation method for the cantilever system of a high-speed railway catenary support device, such as... Figure 1 ,include:

[0059] Step 101: Scan the metal components of the carpal tunnel system to generate a three-dimensional image, collect the stress wave signal of the metal components under dynamic load, and detect the electric field intensity distribution of the abnormal temperature rise area and aging area on the surface of the metal components. The aging area represents the aging defect area of ​​the non-metallic components connected to or wrapped by the metal components in the carpal tunnel system.

[0060] In this step, dynamic load action refers to the periodic mechanical impact and stress changes on the metal components of the cantilever system caused by train vibration, wind load, and electrothermal coupling during high-speed rail operation.

[0061] Stress wave signal refers to the transient elastic wave signal generated by crack propagation or stress concentration in a metal component under dynamic load, which is collected by an acoustic emission sensor.

[0062] An abnormal temperature rise area refers to a region on the surface of a metal component that shows a significant change in temperature gradient due to partial discharge or friction, as detected by a high-resolution thermal imaging device.

[0063] Electric field intensity distribution refers to the non-uniform variation of electric field intensity around the aging area detected by an electric field gradient analysis device.

[0064] In this embodiment, an ultrasonic phased array probe is first used to emit a sound beam at a preset angle and receive the reflected signal. Then, a three-dimensional image of the metal component is generated by synthetic aperture focusing technology. During the process of dynamic load caused by the passing of the train, the original stress wave signal generated by the metal component is collected in real time by an acoustic emission sensor. In order to avoid signal distortion caused by environmental interference, the original stress wave signal is further bandpass filtered to effectively remove environmental noise and finally obtain the stress wave signal.

[0065] Secondly, an infrared thermal imager is used to capture the surface temperature field data of the metal parts at a rate of 30 frames per second. The temperature field data is then preprocessed to remove environmental thermal noise interference, resulting in denoised temperature field data. Based on the denoised temperature field data, the surface temperature gradient is calculated, and areas where the temperature gradient exceeds a preset temperature change threshold are selected as abnormal temperature rise areas on the surface of the metal parts. Simultaneously, an electric field gradient sensor is used to scan the insulating material connected to the metal parts in a gridded pattern, thereby generating the electric field intensity distribution of the aging area.

[0066] Step 102: Extract the spectral characteristics of the stress wave signal based on the cyclic load parameters of the metal component, and combine them with the geometric structural parameters of the bolt connection of the metal component to determine the energy distribution characteristics of the bolt connection in the frequency dimension.

[0067] In this step, the cyclic load parameter refers to the fatigue characteristic parameters of metal components defined based on industry standards. These parameters include core parameters such as stress amplitude and number of cycles.

[0068] Spectral characteristics refer to the frequency band energy distribution characteristics extracted by frequency domain analysis of stress wave signals.

[0069] Geometric parameters refer to the design parameters of bolted connections, including thread pitch, hole diameter, and helix angle direction.

[0070] Energy distribution characteristics refer to the energy distribution in the frequency dimension at the bolted connection, which is generated after frequency domain energy allocation of the vibration period segment through a multi-channel filter bank.

[0071] In this embodiment of the application, step 102 specifically includes the following steps:

[0072] Step 201: Based on the correspondence between stress amplitude and number of cycles in the cyclic load parameters of the metal component, the time-domain waveform of the stress wave signal is divided into multiple vibration period segments.

[0073] In this step, the vibration period segment refers to the time period corresponding to a single dynamic load action, obtained by dividing the time-domain waveform of the stress wave signal according to the correspondence between the stress amplitude and the number of cycles in the cyclic load parameters.

[0074] In this embodiment, the vibration period of the stress wave signal is first determined by the correspondence between the stress amplitude and the number of cycles in the cyclic load parameters; then, the time point when the stress amplitude reaches its peak value in the time domain waveform of the stress wave signal is extracted, and the time domain waveform is segmented according to the determined vibration period duration, based on this time point as the starting reference, to finally obtain multiple vibration period segments.

[0075] For example, when the relationship between stress amplitude and number of cycles in the cyclic load parameters shows that the duration of a single dynamic load is 20ms, the time-domain waveform is divided every 20ms starting from the peak time of the stress amplitude to generate multiple vibration period segments.

[0076] Step 202: Based on the thread pitch at the bolt connection, construct a multi-channel filter bank that is consistent with the helix angle direction at the bolt connection.

[0077] In this step, the multi-channel filter bank refers to a filter device that includes multiple sub-filters, designed based on the thread pitch and helix angle direction at the bolt connection, with the center frequency of each sub-filter matching the characteristic frequency of the stress wave at the root of the thread.

[0078] In this embodiment, the characteristic frequency of stress wave propagation at the root of the thread is first calculated based on the thread pitch at the bolt connection. The characteristic frequency is obtained by calculating the stress wave velocity and twice the thread pitch. The stress wave velocity refers to the elastic wave propagation speed of the metal component material, which is determined by the elastic modulus of the metal component material. The thread pitch refers to the axial distance between adjacent threads at the bolt connection, which directly affects the propagation path length of the stress wave at the root of the thread. Then, combined with the helix angle direction at the bolt connection, the frequency band distribution angle of each sub-filter is adjusted so that the frequency band coverage direction is consistent with the helix path of the thread. Finally, the center frequency of each sub-filter is set according to the characteristic frequency, the frequency band is divided according to the preset bandwidth, and a multi-channel filter bank is constructed.

[0079] For example, a characteristic frequency of approximately 3.3 kHz corresponds to a thread pitch of 1.5 mm, and the helix angle is 15°. In this case, a multi-channel filter bank can be designed to cover the 2-5 kHz frequency band, with the frequency distribution angle of each sub-filter consistent with the 15° helix angle.

[0080] Step 203: Input the vibration period segment into the multi-channel filter bank for frequency domain energy allocation to extract the target frequency band energy amplitude in each vibration period segment and generate an initial frequency domain energy distribution.

[0081] In this step, the target frequency band energy amplitude refers to the energy amplitude corresponding to the characteristic frequency of the stress concentration area at the bolt connection after the vibration period segment is filtered by a multi-channel filter bank.

[0082] The initial frequency domain energy distribution refers to the original distribution data that reflects the degree of energy concentration at different frequencies at the bolt connection after integrating the target frequency band energy amplitudes of all vibration period segments.

[0083] In this embodiment of the application, step 203 specifically includes the following steps:

[0084] Step 211: Determine the center frequency distribution of the multi-channel filter bank based on the thread pitch and hole diameter at the bolt connection.

[0085] In this step, the aperture size refers to the inner diameter of the bolt connection. Its size determines the length of the stress wave reflection path at the aperture edge, which in turn affects the frequency characteristics of the reflected wave.

[0086] The center frequency distribution refers to the set of core frequencies of each sub-filter in a multi-channel filter bank, which matches the characteristic frequency of stress wave propagation in stress concentration areas such as the root of the thread and the edge of the hole.

[0087] In this embodiment, firstly, based on the propagation characteristic frequency of the stress wave at the root of the thread, and combined with the aperture size, the reflection characteristic frequency of the stress wave at the aperture edge is calculated. The reflection characteristic frequency is obtained by calculating the quotient of the stress wave velocity and twice the aperture size. Then, using these two types of characteristic frequencies as core references, frequency intervals are divided according to preset intervals, with each frequency interval corresponding to the center frequency of a sub-filter. Finally, the center frequencies of all sub-filters are integrated to generate the center frequency distribution of the multi-channel filter bank. For example, a thread pitch of 1.5mm corresponds to a characteristic frequency of 3.3kHz, and an aperture size of 17mm corresponds to a reflection characteristic frequency of 2.8kHz. Therefore, the center frequencies are set to include 2.8kHz, 2.9kHz, ..., 3.3kHz, ..., 5.0kHz, forming a center frequency distribution.

[0088] Step 212: Based on the center frequency distribution, convert the time-domain waveform of each vibration period segment into a frequency-domain energy distribution.

[0089] In this step, frequency domain energy distribution refers to the energy amplitude distribution data corresponding to each frequency component after the time domain waveform of the vibration period is converted to the frequency domain.

[0090] In this embodiment, the time-domain waveform of each vibration period segment is first windowed using a Hanning window to avoid spectral leakage; then, the windowed time-domain waveform is subjected to a Fourier transform to convert it into a frequency-domain signal; finally, the energy amplitude of each frequency point in the frequency-domain signal is extracted and integrated according to the correspondence between frequency and energy amplitude to generate the frequency-domain energy distribution of each vibration period segment.

[0091] Step 213: Based on the frequency domain energy distribution, extract the target frequency band energy amplitude corresponding to the center frequency distribution to generate an initial frequency domain energy distribution.

[0092] In this step, the target frequency band energy amplitude refers to the energy amplitude within the frequency range corresponding to the center frequency distribution of the multi-channel filter bank in the frequency domain energy distribution.

[0093] The initial frequency domain energy distribution refers to the original energy distribution data obtained by integrating the target frequency band energy amplitudes of all vibration period segments according to the dimensions of frequency and spatial location.

[0094] In this embodiment, firstly, the energy amplitude within the frequency range corresponding to the center frequency distribution is extracted from the frequency domain energy distribution of each vibration cycle segment as the target frequency band energy amplitude of that vibration cycle segment; then, according to the dynamic load action time corresponding to each vibration cycle segment, the target frequency band energy amplitude is mapped to the spatial position of the bolt connection; finally, all mapped energy amplitude data are integrated to generate the initial frequency domain energy distribution.

[0095] Step 204: Based on the crack curvature, the initial frequency domain energy distribution is corrected by curvature weight to obtain the corrected frequency domain energy distribution. Based on the corrected frequency domain energy distribution, the energy distribution characteristics of the bolt connection in the frequency dimension are generated.

[0096] In this step, crack curvature refers to the degree of bending of the crack surface at the bolt connection, expressed as the radius of curvature; corrected frequency domain energy distribution refers to the energy distribution data obtained after correcting the initial frequency domain energy distribution by curvature weight; energy distribution characteristics refer to data reflecting the correlation between the energy concentration intensity and spatial location at different frequencies at the bolt connection.

[0097] In this embodiment, firstly, the crack curvature data at the bolt connection is extracted from the three-dimensional image generated by the ultrasonic phased array to determine the crack curvature radius at each spatial location; then, a curvature weighting rule is set: the smaller the crack curvature radius, i.e. the more severe the stress concentration, the larger the energy weighting coefficient at the corresponding spatial location; then, according to this rule, a weighting coefficient is assigned to the energy amplitude at each spatial location in the initial frequency domain energy distribution, and the energy amplitude is multiplied by the corresponding weighting coefficient to obtain the corrected energy amplitude; finally, all corrected energy amplitudes are integrated to generate the corrected frequency domain energy distribution, thereby generating the energy distribution characteristics of the bolt connection in the frequency dimension.

[0098] For example, the weighting coefficient for the region with a crack curvature radius of 0.2 mm is set to 1.5, and the weighting coefficient for the region with a curvature radius of 0.5 mm is set to 0.8. The initial energy amplitude is multiplied by the corresponding coefficient and then integrated to obtain the corrected frequency domain energy distribution.

[0099] The embodiments of this application can avoid feature distortion caused by cross-cycle signal aliasing; solve the problem of missed detection of signals in stress concentration areas caused by orientation mismatch in traditional filter banks; and highlight the energy signal of the real stress concentration area and reduce the background noise interference in non-stress concentration areas based on the weight correction mechanism of crack curvature, so as to provide accurate frequency domain data support for subsequent fatigue crack segmentation.

[0100] Step 103: Perform multi-data fusion on the three-dimensional image, the energy distribution characteristics, the temperature rise anomaly region, and the electric field intensity distribution to generate a multi-source data set. Combine the cyclic load parameters to generate the segmentation results of fatigue cracks and insulation aging defects of the metal component.

[0101] In this embodiment of the application, step 103 specifically includes the following steps:

[0102] Step 301: Based on the thread pitch and hole diameter in the geometric structure parameters, perform multi-scale spatial meshing on the three-dimensional image to obtain multiple mesh units;

[0103] In this embodiment, the mesh cell size is first defined according to the thread pitch parameter to ensure that each mesh cell covers one thread cycle; then, a finer mesh is used in the stress concentration area to improve the crack detection accuracy; and finally, the pixel coordinates of the three-dimensional image are aligned with the mesh cells to generate meshed data with spatial resolution adaptation.

[0104] Step 302: Based on the trend of the frequency domain energy amplitude changing with time in the energy distribution characteristics, perform phase compensation processing on the temperature gradient data in the temperature rise anomaly region to generate first temperature distribution data corresponding to the spatial coordinates of the grid cell;

[0105] In this step, the frequency domain energy amplitude refers to the specific frequency band energy amplitude extracted from the frequency domain energy distribution that matches the characteristic frequency of the stress wave propagation at the root of the thread.

[0106] Temperature gradient data refers to the spatial distribution of temperature gradient in areas of abnormal temperature rise, and phase compensation is needed to eliminate time domain offset.

[0107] Phase compensation refers to interpolating temperature gradient data in the time domain according to a preset dynamic load frequency band to eliminate temperature field shifts caused by vibration.

[0108] In this embodiment, the time-domain windows of the temperature gradient data are first divided according to the main frequency of the dynamic load, with each window corresponding to a vibration cycle; then, the timestamps of the frequency-divided temperature gradient data are aligned with the phase of the stress wave signal using an interpolation algorithm to compensate for the temperature field shift caused by vibration; finally, the compensated temperature gradient data is mapped to the spatial coordinates of the grid cells to generate the first temperature distribution data.

[0109] Step 303: Perform boundary topology correction processing on the aging region in the first temperature distribution data to generate the second temperature distribution data;

[0110] In this step, the difference refers to the angular difference between the electric field distortion gradient and the temperature gradient in the spatial direction, which is used to identify the boundary of the insulation aging region.

[0111] Boundary topology correction refers to correcting the boundary of the insulation aging region based on the gradient direction angle difference to ensure that the boundary curvature is continuous and consistent with the stress release path.

[0112] In this embodiment of the application, step 303 specifically includes the following steps:

[0113] Step 311: Based on the thread profile at the bolt connection, perform coordinate transformation on the gradient direction angle of the electric field intensity distribution to generate a first direction angle distribution;

[0114] In this step, the thread profile refers to the geometry of the thread at the bolt connection.

[0115] The first directional angular distribution refers to the angular distribution data of the electric field intensity gradient direction in a local coordinate system that is consistent with the helical direction of the thread after coordinate transformation. This data has eliminated the interference of the thread geometry on the electric field distortion direction.

[0116] In this embodiment, based on the thread profile at the bolt connection, the gradient direction angle of the electric field intensity distribution is subjected to coordinate transformation. Specifically, an affine transformation is used to convert the gradient direction angle in the global coordinate system into a local coordinate system consistent with the helical direction of the thread, generating a first direction angle distribution. For example, the electric field gradient direction angle is projected along the helix angle direction to eliminate the influence of the thread geometry on the electric field distortion direction.

[0117] Step 312: Extract the fluctuation amplitude of the gradient direction angle of the temperature gradient data within the preset dynamic load frequency band, and combine it with the crack curvature of the thread profile to generate a second direction angle distribution;

[0118] In this step, crack curvature refers to the curvature characteristics of the crack propagation path at the root of the bolt thread, which reflects the degree of stress concentration.

[0119] The second directional angle distribution refers to the temperature gradient directional angle distribution data that matches the crack propagation path after curvature weight correction. The temperature gradient directional angle fluctuation characteristics in the high stress concentration area have been amplified in this data.

[0120] In this embodiment, the temperature gradient direction angle fluctuation amplitude within a preset dynamic load frequency band is first extracted. Then, the temperature gradient direction angle is corrected by curvature weighting in combination with the crack curvature of the thread profile. One correction method is to amplify the fluctuation amplitude in the high curvature region, and finally generate a second direction angle distribution that matches the crack propagation path.

[0121] Step 313: Based on the thread pitch and helix angle direction at the bolt connection, perform multi-cycle phase matching on the difference between the gradient direction angles of the first direction angle distribution and the second direction angle distribution to obtain the difference distribution;

[0122] In this embodiment, the difference between the first directional angular distribution and the second directional angular distribution is segmented and matched according to the number of helical cycles based on the thread pitch and helix angle direction. The average difference within each matched phase interval is calculated using a sliding window algorithm to generate a difference distribution, thereby suppressing non-periodic noise interference.

[0123] Step 314: Perform fracture boundary connection processing on the regions in the difference distribution where the gradient direction angle difference exceeds a preset adaptive threshold to generate preliminary corrected temperature distribution data;

[0124] In this step, the preset adaptive threshold refers to the judgment threshold set based on the dynamic load frequency band and thread geometry parameters. It is used to distinguish between normal angle differences and abnormal angle differences caused by defects. Different dynamic load frequency bands correspond to different threshold standards.

[0125] In this embodiment of the application, step 314 specifically includes the following steps:

[0126] Step 321: Based on the helix angle direction and crack curvature at the bolt connection, generate a helical path template corresponding to the number of helical turns of the thread profile, and simultaneously perform meshing on the difference distribution to generate the target mesh;

[0127] In this step, the direction of the helix angle refers to the helical extension direction of the bolt thread, which determines the geometric orientation of the helical path template.

[0128] A spiral path template refers to a reference path generated based on the number of thread turns and the helix angle, used to guide the direction of crack propagation.

[0129] The target mesh refers to a multi-level mesh divided according to the thread pitch, with the axis of the mesh unit aligned with the helix angle, used to locate the break point.

[0130] In this embodiment, a helical path axis is first generated based on the number of thread turns and the direction of the helix angle. The curvature of this axis is consistent with the crack curvature. Then, a helical path template is generated based on the helical path axis. Next, the difference distribution is divided into multiple levels according to the thread pitch, with each level corresponding to one thread turn. The mesh cell size is set to decrease with each level to generate the target mesh.

[0131] Step 322: Select target break points from the target mesh;

[0132] In this step, the cumulative distribution of vibration dominant frequency energy amplitude refers to the cumulative result of the energy amplitude of each frequency within the preset dynamic load frequency band, which is used to screen for effective fracture points.

[0133] The target fracture point refers to the set of fracture point coordinates after filtering by the energy threshold, which represents the location of potential cracks.

[0134] In this embodiment, the energy amplitude within each fracture point detection grid cell is first accumulated in the time domain according to the dominant vibration frequency to generate a cumulative distribution of the dominant vibration frequency energy amplitude; then, based on this distribution, an energy threshold corresponding to a preset dynamic load frequency band is set, and target fracture points exceeding the energy threshold are retained.

[0135] Step 323: Extend the spiral path of the target fracture point based on the spiral path template to generate a first spiral path and a second radius of curvature change curve;

[0136] In this step, the second radius of curvature variation curve refers to the rate distribution of the radius of curvature of the helical extension path as a function of the number of thread turns.

[0137] In this embodiment, the target fracture point is first projected onto the helical path template and extended along the helical helix angle direction according to the thread pitch to generate a first helical path covering multiple turns of thread; then the rate of change of the radius of curvature of each turn of thread on the helical extension path is calculated to generate a second radius of curvature change curve.

[0138] Step 324: Based on the dynamic deviation between the preset first radius of curvature change curve and the second radius of curvature change curve, perform curvature correction processing on the first helical path to generate a second helical path. Combine the helical cycle number of the thread profile, perform phase compensation processing on the second helical path to obtain a third helical path.

[0139] In this step, the first curvature radius change curve refers to the theoretical curvature change reference curve based on industry standards or experimental data.

[0140] In this embodiment of the application, step 324 specifically includes the following steps:

[0141] Step 331: Based on the number of helical turns of the thread profile at the bolt connection, divide the preset first curvature radius variation curve into multiple first curvature segments;

[0142] In this step, the first curvature segment characterizes the curvature decrease pattern of each thread turn under undisturbed conditions.

[0143] In this embodiment of the application, the first curvature radius change curve is divided into multiple first curvature segments according to the actual number of turns of the bolt thread, and each first curvature segment defines the nominal curvature radius change rate of that turn of the thread.

[0144] Step 332: Based on the number of helical turns of the thread profile at the bolt connection, extract multiple second curvature segments from the second curvature radius variation curve;

[0145] In this step, the second curvature segment refers to the measured crack path curvature change data generated by spiral path extension, which reflects the actual propagation path under dynamic load.

[0146] In this embodiment, the corresponding second curvature segments are extracted one by one from the actual second curvature radius change curve according to the number of spiral turns. For example, the curvature radius decreases from 0.2mm to 0.19mm in the first turn and decreases to 0.18mm in the second turn.

[0147] Step 333: Calculate the deviation between the second curvature segment and the corresponding first curvature segment of each thread turn to obtain the deviation nominal value. When the deviation nominal value exceeds the preset adaptive threshold, correct the second curvature radius change curve to obtain the third curvature radius change curve. Combined with the helix angle direction, obtain the second helical path.

[0148] In this step, deviation from nominal value refers to the difference between the actual rate of change of radius of curvature and the nominal value, reflecting the influence of dynamic load on crack path.

[0149] In this embodiment, the deviation between the second curvature segment and the first curvature segment in each revolution is first calculated. If the deviation exceeds the current preset adaptive threshold, the curvature radius is forcibly corrected according to the nominal value to obtain the third curvature radius change curve. Then, combined with the helical rise angle direction, the second helical path is regenerated.

[0150] Step 334: Calculate the phase angle offset of the second helical path based on the number of helical cycles of the thread profile and the direction of the helical helix angle. Based on the phase angle offset, perform phase synchronization compensation on the node spacing in the second helical path to obtain the third helical path.

[0151] In this step, the phase angle offset refers to the difference in phase angle between the second spiral path node and the theoretical path, caused by path deformation due to dynamic loads.

[0152] In this embodiment, the phase angle offset of the corrected path is first calculated based on the number of helical cycles and the direction of the helical helix angle. Based on this, the node spacing in the second helical path is adjusted by multiplying the thread pitch and the number of cycles. After eliminating the phase offset, the third helical path is obtained.

[0153] Step 325: Based on the gradient decay rate in the amplitude decay characteristics, perform curvature smoothing on the third spiral path to generate preliminary corrected temperature distribution data.

[0154] In this step, curvature smoothing refers to eliminating local abrupt changes in the curvature of the path to make it conform to the continuity of the stress relief path.

[0155] In this embodiment, the path nodes in the third spiral path are first assigned weights according to the gradient decay rate. For example, one method is to reduce the weight of high decay regions. Then, the curvature of the weighted path is smoothed by the sliding window averaging method. Finally, preliminary corrected temperature distribution data matching the dynamic load characteristics of the high-speed rail cantilever system is generated to suppress abrupt changes caused by noise.

[0156] Step 315: Perform a coupling consistency verification of boundary curvature and gradient decay rate on the initially corrected temperature distribution data to generate second temperature distribution data.

[0157] In this step, the gradient decay rate refers to the decay characteristic of the electric field gradient magnitude with distance, which is used to verify the physical rationality of the defect boundary.

[0158] In this embodiment, the boundary curvature continuity of the initially corrected temperature distribution data is first checked, and then abnormal segments with abrupt curvature changes are removed. Simultaneously, the spatial consistency between the temperature gradient boundary and the electric field distortion region is verified by combining the electric field gradient decay rate, ultimately generating the second temperature distribution data.

[0159] Step 304: Based on the correspondence between stress amplitude and number of cycles in the cyclic load parameters, perform dynamic weight allocation processing on the crack depth distribution and frequency domain energy amplitude within the grid cell to generate target risk level information;

[0160] In this step, the target risk level information refers to a quantitative level that characterizes the degree of crack propagation risk within a grid cell, generated by combining crack depth distribution, frequency domain energy amplitude, and cyclic load parameters.

[0161] In this embodiment, the fatigue curve of the metal component in the industry standard is first obtained. The fatigue curve clarifies the relationship between the degree of fatigue damage of the metal component under different combinations of stress amplitude and number of cycles. Based on this fatigue curve, a three-dimensional mapping relationship table is established between the stress amplitude, number of cycles and the weighting coefficient of crack propagation risk in the cyclic load parameters. Specifically, when the stress amplitude is larger and the number of cycles is closer to the limit number of cycles in the fatigue curve, the corresponding weighting coefficient of crack propagation risk is larger, and vice versa.

[0162] Subsequently, for each grid cell, crack depth distribution data and frequency domain energy amplitude data are extracted from that grid cell. Basic weight coefficients are assigned to these two types of data. These basic weight coefficients are determined based on the contribution of the two types of data to the crack propagation risk assessment. Two basic weighted values ​​are obtained by multiplying the crack depth distribution data with the corresponding basic weight coefficient and the frequency domain energy amplitude data with the corresponding basic weight coefficient. These two basic weighted values ​​are then summed to obtain the comprehensive basic weighted value of the grid cell.

[0163] Finally, based on the stress amplitude and number of cycles in the cyclic load parameters corresponding to the grid cell, the weight coefficient of the corresponding crack propagation risk is obtained from the above three-dimensional mapping table. The comprehensive basic weighted value is multiplied by the weight coefficient of the crack propagation risk to obtain the crack propagation risk quantification value of each grid cell. Then, according to the preset risk quantification value interval division rules, the risk quantification value of each grid cell is mapped to a specific risk level, and finally the target risk level corresponding to each grid cell is generated.

[0164] Step 305: Based on the gradient direction consistent region in the second temperature distribution data and the target risk level information, and combined with the amplitude attenuation characteristics of the electric field intensity distribution, generate the segmentation results of fatigue cracks and insulation aging defects of the metal component.

[0165] In this step, the region with consistent gradient direction refers to the region in the second temperature distribution data where the electric field gradient and temperature gradient are in the same direction, characterizing the physical correlation between insulation aging and metal cracking.

[0166] The amplitude decay characteristic refers to the decay law of the electric field gradient amplitude with distance. It is used to filter low confidence areas and is determined by the exponential decay model of the electric field gradient amplitude with distance.

[0167] In this embodiment, the target risk level and the gradient direction consistency region in the second temperature distribution data are first aligned with the geometric structural parameters of the bolt connection using a multi-scale grid to ensure that their spatial coordinates are consistent. Then, in the overlapping region, the weight of the crack propagation risk coefficient is multiplied with the confidence level of the gradient direction consistency intensity to generate a fusion weight map.

[0168] Next, based on the exponential decay characteristic of the electric field intensity distribution amplitude with distance, amplitude decay weights are applied to the regions in the fusion weight map to remove suspected noise regions with low decay weights, resulting in an optimized fusion weight map. Then, regions with high crack risk coefficients and matching crack depth distribution in the 3D image are extracted from the optimized fusion weight map to generate metal crack boundaries. Finally, the parts in the gradient direction consistency region that match the amplitude decay characteristic of the electric field intensity distribution are extracted to generate insulation aging defect boundaries.

[0169] This application's embodiments ensure the adaptability of detection accuracy in different regions through multi-scale meshing based on geometric parameters. Phase compensation combined with energy distribution characteristics eliminates temperature field shifts caused by dynamic loads. Defect boundary differences are accurately identified through direction angle conversion of thread profile constraints and multi-cycle phase matching. A continuous and reliable defect path is formed through helical path extension, curvature correction, and phase compensation. The reliability of temperature distribution data is improved through coupling verification of boundary curvature and gradient decay rate. Finally, by combining dynamic weight allocation of cyclic load parameters and multi-dimensional data fusion, accurate differentiation and segmentation of the two types of defects are achieved, solving the problem of easy omission and misjudgment in single data detection under dynamic load conditions, and improving the accuracy and reliability of defect detection in high-speed rail cantilever systems.

[0170] Figure 2 This application provides a schematic diagram of the structure of an image segmentation system for a high-speed rail overhead contact line support device cantilever system, as shown in the embodiment of this application. Figure 2 As shown, the system includes:

[0171] The acquisition module 21 is used to scan the metal components of the carpal tunnel system, generate three-dimensional images, acquire stress wave signals of the metal components under dynamic load, and detect the electric field intensity distribution of abnormal temperature rise areas and aging areas on the surface of the metal components. The aging areas characterize the aging defect areas of non-metallic components connected to or wrapped with the metal components in the carpal tunnel system.

[0172] The determination module 22 is used to extract the spectral characteristics of the stress wave signal based on the cyclic load parameters of the metal component, and combine them with the geometric structural parameters of the bolt connection of the metal component to determine the energy distribution characteristics of the bolt connection in the frequency dimension.

[0173] The fusion module 23 is used to fuse the three-dimensional image, the energy distribution characteristics, the temperature rise anomaly area, and the electric field intensity distribution to generate a multi-source data set. Combined with the cyclic load parameters, it generates the segmentation results of fatigue cracks and insulation aging defects of the metal component.

[0174] Figure 2 The image segmentation system of the cantilever system of the high-speed railway catenary support device can perform... Figure 1 The implementation principle and technical effects of the image segmentation method for the cantilever system of a high-speed railway catenary support device described in the above embodiment will not be repeated here. The specific operation methods of each module and unit in the image segmentation system of the cantilever system of a high-speed railway catenary support device in the above embodiment have been described in detail in the embodiments of the relevant method, and will not be elaborated upon here.

[0175] In one possible design, Figure 2 The image segmentation system of the cantilever system of a high-speed railway catenary support device shown in the embodiment can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0176] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0177] The processing component 32 is used to: scan the metal components of the carpal tunnel system to generate a three-dimensional image; acquire stress wave signals of the metal components under dynamic load; detect the electric field intensity distribution of abnormal temperature rise areas and aging areas on the surface of the metal components, wherein the aging areas characterize the aging defect areas of non-metallic components connected to or wrapped by the metal components in the carpal tunnel system; extract the spectral features of the stress wave signals based on the cyclic load parameters of the metal components; determine the energy distribution characteristics of the bolted connections in the frequency dimension by combining the geometric structural parameters of the bolted connections of the metal components; fuse the three-dimensional image, the energy distribution characteristics, the abnormal temperature rise areas, and the electric field intensity distribution to generate a multi-source data set; and generate the segmentation results of fatigue cracks and insulation aging defects of the metal components by combining the cyclic load parameters.

[0178] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits, digital signal processors, digital signal processing devices, programmable logic devices, field-programmable gate arrays, controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0179] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk, or optical disk.

[0180] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0181] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0182] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0183] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0184] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The illustrated embodiment presents an image segmentation method for the cantilever system of a high-speed railway catenary support device.

[0185] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0186] The device embodiments described above are merely illustrative. 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 modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0187] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0188] Finally, it should be noted that the above 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.

Claims

1. A method for image segmentation of a cantilever system for a high-speed railway overhead contact line support device, characterized in that, include: The metal components of the carpal tunnel system are scanned to generate three-dimensional images. Stress wave signals of the metal components under dynamic loads are collected. The electric field intensity distribution of abnormal temperature rise areas and aging areas on the surface of the metal components is detected. The aging areas characterize the aging defect areas of non-metallic components connected to or wrapped with the metal components in the carpal tunnel system. Based on the cyclic load parameters of the metal component, the spectral characteristics of the stress wave signal are extracted, and combined with the geometric structural parameters of the bolt connection of the metal component, the energy distribution characteristics of the bolt connection in the frequency dimension are determined. The three-dimensional image, the energy distribution characteristics, the abnormal temperature rise region, and the electric field intensity distribution are fused to generate a multi-source data set. Combined with the cyclic load parameters, the segmentation results of fatigue cracks and insulation aging defects of the metal component are generated.

2. The image segmentation method for the cantilever system of the high-speed railway catenary support device according to claim 1, characterized in that, The three-dimensional image, the energy distribution characteristics, the temperature rise anomaly region, and the electric field intensity distribution are fused to generate a multi-source data set. Combined with the cyclic load parameters, segmentation results for fatigue cracks and insulation aging defects in the metal component are generated, including: Based on the thread pitch and hole diameter in the geometric structural parameters, the three-dimensional image is divided into a multi-scale spatial mesh to obtain multiple mesh units; Based on the trend of the frequency domain energy amplitude changing with time in the energy distribution characteristics, phase compensation processing is performed on the temperature gradient data in the temperature rise anomaly region to generate first temperature distribution data corresponding to the spatial coordinates of the grid cell. The aging region in the first temperature distribution data is subjected to boundary topology correction processing to generate the second temperature distribution data. Based on the correspondence between stress amplitude and cycle number in the cyclic load parameters, dynamic weight allocation is performed on the crack depth distribution and frequency domain energy amplitude within the grid cell to generate target risk level information; Based on the gradient direction consistent region in the second temperature distribution data and the target risk level information, combined with the amplitude attenuation characteristics of the electric field intensity distribution, the segmentation results of fatigue cracks and insulation aging defects of the metal component are generated respectively.

3. The image segmentation method for the cantilever system of the high-speed railway catenary support device according to claim 2, characterized in that, The aging region in the first temperature distribution data is subjected to boundary topology correction processing to generate the second temperature distribution data, including: Based on the thread profile at the bolt connection, the gradient direction angle of the electric field intensity distribution is transformed to generate a first direction angle distribution; Extract the fluctuation amplitude of the gradient direction angle corresponding to the aging region in the first temperature distribution data within the preset dynamic load frequency band, and combine it with the crack curvature of the thread profile to generate a second direction angle distribution; Based on the thread pitch and helix angle direction at the bolt connection, multi-cycle phase matching is performed on the difference between the gradient direction angles of the first and second direction angle distributions to obtain the difference distribution; For regions in the difference distribution where the gradient direction angle difference exceeds a preset adaptive threshold, a fracture boundary connection process is performed to generate second temperature distribution data.

4. The image segmentation method for the cantilever system of the high-speed railway catenary support device according to claim 3, characterized in that, For regions in the difference distribution where the gradient direction angle difference exceeds a preset adaptive threshold, fracture boundary connection processing is performed to generate second temperature distribution data, including: Based on the helix angle direction and crack curvature at the bolt connection, a helical path template corresponding to the number of helical turns of the thread profile is generated. At the same time, the difference distribution is meshed to generate the target mesh. Target break points are selected from the target grid; Based on the spiral path template, the target fracture point is extended by spiral path to generate a first spiral path and a second radius of curvature change curve. Based on the dynamic deviation between the preset first radius of curvature change curve and the second radius of curvature change curve, the first helical path is subjected to curvature correction processing to generate a second helical path. Combined with the helical cycle number of the thread profile, the second helical path is subjected to phase compensation processing to obtain a third helical path. Based on the gradient decay rate in the amplitude decay characteristics, the curvature smoothing process is applied to the third spiral path to generate second temperature distribution data.

5. The image segmentation method for the cantilever system of the high-speed railway catenary support device according to claim 4, characterized in that, Based on the dynamic deviation between the preset first radius of curvature change curve and the second radius of curvature change curve, the first helical path is subjected to curvature correction processing to generate a second helical path. Then, combined with the helical cycle number of the thread profile, the second helical path is subjected to phase compensation processing to obtain a third helical path, including: Based on the number of helical turns of the thread profile at the bolt connection, the preset first curvature radius variation curve is divided into multiple first curvature segments; Based on the number of helical turns of the thread profile at the bolted connection, multiple second curvature segments are extracted from the second curvature radius variation curve; Calculate the deviation between the second curvature segment and the corresponding first curvature segment of each thread turn to obtain the nominal deviation value. When the nominal deviation value exceeds the preset adaptive threshold, correct the second curvature radius change curve to obtain the third curvature radius change curve. Combine the helix angle direction to obtain the second helical path. Based on the number of helical cycles of the thread profile and the direction of the helical helix angle, the phase angle offset of the second helical path is calculated. Based on the phase angle offset, phase synchronization compensation is performed on the node spacing in the second helical path to obtain the third helical path.

6. The image segmentation method for the cantilever system of the high-speed railway catenary support device according to claim 1, characterized in that, Based on the cyclic load parameters of the metal component, the spectral characteristics of the stress wave signal are extracted. Combined with the geometric structural parameters of the bolted connection of the metal component, the energy distribution characteristics of the bolted connection in the frequency dimension are determined, including: Based on the correspondence between stress amplitude and number of cycles in the cyclic load parameters of the metal component, the time-domain waveform of the stress wave signal is divided into multiple vibration period segments; Based on the thread pitch at the bolted connection, a multi-channel filter bank is constructed with the helix angle direction consistent with that at the bolted connection. The vibration period segment is input into the multi-channel filter bank for frequency domain energy allocation to extract the target frequency band energy amplitude in each vibration period segment and generate an initial frequency domain energy distribution. Based on the crack curvature, the initial frequency domain energy distribution is corrected by curvature weighting to obtain the corrected frequency domain energy distribution. Based on the corrected frequency domain energy distribution, the energy distribution characteristics of the bolt connection in the frequency dimension are generated.

7. The image segmentation method for the cantilever system of the high-speed railway catenary support device according to claim 6, characterized in that, The vibration period segment is input into the multi-channel filter bank for frequency domain energy allocation to extract the target frequency band energy amplitude in each vibration period segment and generate an initial frequency domain energy distribution, including: The center frequency distribution of the multi-channel filter bank is determined based on the thread pitch and hole diameter at the bolt connection. Based on the center frequency distribution, the time-domain waveform of each vibration period segment is converted into a frequency-domain energy distribution; Based on the frequency domain energy distribution, the target frequency band energy amplitude corresponding to the center frequency distribution is extracted to generate an initial frequency domain energy distribution.

8. An image segmentation system for a cantilever system of a high-speed railway catenary support device, characterized in that, include: The acquisition module is used to scan the metal components of the carpal tunnel system, generate three-dimensional images, acquire stress wave signals of the metal components under dynamic load, and detect the electric field intensity distribution of abnormal temperature rise areas and aging areas on the surface of the metal components. The aging areas characterize the aging defect areas of non-metallic components connected to or wrapped with the metal components in the carpal tunnel system. The determination module is used to extract the spectral characteristics of the stress wave signal based on the cyclic load parameters of the metal component, and combine them with the geometric structural parameters of the bolt connection of the metal component to determine the energy distribution characteristics of the bolt connection in the frequency dimension. The fusion module is used to fuse the three-dimensional image, the energy distribution characteristics, the temperature rise anomaly region, and the electric field intensity distribution to generate a multi-source data set. Combined with the cyclic load parameters, it generates the segmentation results of fatigue cracks and insulation aging defects of the metal component.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the image segmentation method for the cantilever system of a high-speed railway catenary support device as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The device contains a computer program, which, when executed by a computer, implements an image segmentation method for the cantilever system of a high-speed rail contact network support device as described in any one of claims 1 to 7.

Images