Performance test method for dynamic bending of flexible connection sheet

Abstract

The invention relates to a performance testing method for dynamic bending of a flexible connecting piece in the field of intelligent manufacturing, and the method comprises the steps: collecting multi-modal data of the flexible connecting piece under a dynamic bending working condition through a sensor array, and obtaining an initial collection data set; performing de-noising processing on the initial acquisition data set to generate a de-noised resistance sequence and a de-noised clear image sequence; analyzing correlation characteristics of fluctuation amplitude and bending frequency according to the denoised resistance sequence, and determining an abnormal fluctuation classification label; extracting structural damage features through the clear image sequence, and generating a damage distribution diagram; according to the damage distribution diagram and the abnormal fluctuation classification label, fusing a contact state index and a structure damage feature, calculating a performance degradation degree, and obtaining a comprehensive degradation evaluation value; aiming at the comprehensive degradation evaluation value, constructing an acceleration test scheme in combination with standard spectrum statistical characteristics, and generating an acceleration test result; and according to the acceleration test result, mapping an expected degradation path, and determining a long-term performance degradation trend.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing, and more particularly to flexible connectors for new energy vehicle batteries, specifically to a performance testing method for dynamic bending of flexible connectors. Background Technology

[0002] In the fields of new energy vehicles and power equipment, flexible conductive components, as core elements for connecting and conducting electrical energy, directly affect the safety and stability of equipment operation. With technological advancements, traditional rigid connection structures are gradually being replaced by flexible connecting sheets composed of multiple layers of stacked metal sheets, a structure favored for its flexibility. However, ensuring the durability of these components during long-term use has become a critical issue that the industry urgently needs to address; its importance is self-evident, directly impacting the reliability and lifespan of the equipment.

[0003] Currently, although there is some accumulated testing methods for flexible connectors in the market, most are limited to testing the basic properties of the material, neglecting the complex performance of components in real-world usage environments. Especially when faced with common conditions such as repeated bending, existing methods often fail to fully capture the performance degradation process of components over time, and are also unable to reveal potential problems hidden in the internal structure. This deficiency in testing methods leaves product design and quality verification without sufficient basis, thus affecting the reliability of the final product. Focusing on the technical level, the performance evaluation of flexible connectors under dynamic bending faces two core challenges.

[0004] First, the real-time changes in electrical performance during bending are difficult to accurately monitor. Subtle fluctuations in electrical properties, such as resistance, often reflect a deterioration in the internal contact state, which may stem from material fatigue or wear at interlayer contact surfaces. Second, these changes in electrical performance are closely related to structural integrity. Minor damage to the surface or between layers, such as wrinkles or cracks, can further exacerbate the decline in electrical performance, creating a mutually reinforcing cycle of deterioration. These two factors are closely interrelated, jointly contributing to the complexity of performance degradation. For example, in the connection of new energy vehicle battery packs, flexible connectors must frequently withstand vibration and bending. If increased resistance or surface damage cannot be detected in time, it may lead to localized overheating or even connection failure.

[0005] Therefore, how to simultaneously capture the evolution of electrical performance changes and structural damage during dynamic bending under simulated real-world working conditions, and reveal the relationship between the two, has become a key issue in improving the durability assessment and design optimization of flexible connectors. Summary of the Invention

[0006] This invention provides a performance testing method for dynamic bending of flexible connectors, mainly including:

[0007] Multimodal data of the flexible connector under dynamic bending conditions is acquired using a sensor array to obtain an initial dataset. The initial dataset is then denoised to generate a denoised resistance sequence and a clear image sequence. The correlation between fluctuation amplitude and bending frequency is analyzed based on the denoised resistance sequence to determine abnormal fluctuation classification labels. Structural damage features are extracted from the clear image sequence to generate a damage distribution map. Based on the damage distribution map and the abnormal fluctuation classification labels, contact state indicators and structural damage features are fused to calculate the performance degradation degree and obtain a comprehensive degradation assessment value. An accelerated testing scheme is constructed based on the comprehensive degradation assessment value, combined with standard spectral statistical features, to generate accelerated testing results. Based on the accelerated testing results, the expected degradation path is mapped to determine the long-term performance degradation trend. Furthermore, the step of acquiring multimodal data of the flexible connector under dynamic bending conditions using a sensor array to obtain an initial acquisition dataset includes: recording the bending angle range and dynamic load direction in real time using the sensor array, while simultaneously acquiring resistance fluctuation amplitude data and high-resolution surface image data to form the initial acquisition dataset; extracting resistance fluctuation amplitude data from the initial acquisition dataset, filtering outliers using a timestamp matching method, and determining the fluctuation trend within the bending angle range; fusing the high-resolution surface image data with the fluctuation trend, and obtaining the correspondence between multimodal data through pixel-to-pixel mapping; if the correspondence exceeds a preset threshold, adjusting the recording of the dynamic load direction; constructing an extended dataset structure based on the correspondence, adding filter labels, and generating the final acquisition dataset. Furthermore, the step of denoising the initial acquired dataset to generate a denoised resistance sequence and a clear image sequence includes: obtaining the noise interference type from the initial acquired dataset, using median filtering to purify the resistance fluctuation amplitude data to obtain a denoised resistance sequence; for the denoised resistance sequence, obtaining the surface image resolution, using a super-resolution reconstruction method to enhance and optimize it through low-resolution image interpolation and detail restoration to generate a clear image sequence; extracting pixel distribution features from the clear image sequence, and combining them with the denoised resistance sequence to determine the sequence trend change; and for the sequence trend change, obtaining material surface defect indicators to generate an optimized resistance image fusion sequence. Furthermore, the step of analyzing the correlation characteristics between fluctuation amplitude and bending frequency based on the denoised resistance sequence to determine the abnormal fluctuation classification label includes: calculating the resistance value difference through the denoised resistance sequence to obtain a fluctuation amplitude value sequence; calculating the frequency of bending points for the fluctuation amplitude value sequence and statistically analyzing the bending frequency distribution curve; extracting the correlation feature vector between the fluctuation amplitude value and the bending frequency from the bending frequency distribution curve; if the correlation feature vector exceeds a preset threshold range, it is marked as an abnormal fluctuation state; and classifying the abnormal fluctuation state to determine the abnormal fluctuation classification label.Furthermore, the step of extracting structural damage features from the clear image sequence and generating a damage distribution map includes: extracting structural damage features from the clear image sequence using an edge detection algorithm to determine the location of surface damage; if the location of surface damage contains wrinkles, determining the crack extension direction by pixel grayscale difference and obtaining crack location coordinates; determining the damage boundary from the crack location coordinates using a preset threshold to generate a damage boundary map; integrating the surface damage locations of wrinkles and cracks according to the damage boundary map, calculating the damage area ratio, and obtaining a damage severity assessment; and fusing the damage severity assessment into the distribution map through coordinate mapping to generate a damage distribution map. Furthermore, the step of calculating the performance degradation degree and obtaining a comprehensive degradation assessment value by fusing contact state indicators and structural damage characteristics based on the damage distribution map and the abnormal fluctuation classification labels includes: obtaining the coordinates of surface damage points through the damage distribution map and determining the abnormal fluctuation region by combining the abnormal fluctuation classification labels; extracting contact state indicators based on the abnormal fluctuation region and generating a damage fusion vector by fusing structural damage characteristics; calculating the load response curve under the dynamic load direction for the damage fusion vector to determine the performance degradation degree; if the performance degradation degree exceeds a preset threshold, adjusting the load response curve using degradation trend prediction to obtain a degradation correction value; and integrating the damage fusion vector with the degradation correction value to generate a comprehensive degradation assessment value. Furthermore, the step of constructing an accelerated testing scheme and generating accelerated testing results based on the comprehensive degradation assessment value and standard spectral statistical characteristics includes: calculating the initial degradation level by combining the comprehensive degradation assessment value with standard spectral statistical characteristics; obtaining the accelerated bending frequency based on the initial degradation level, constructing an accelerated testing scheme, and determining the bending parameter set; increasing the accelerated bending amplitude according to the bending parameter set, adjusting the stress input within the linear range of the material response, and obtaining a linear response curve; conducting compression tests for a durability verification cycle using the linear response curve, monitoring the degradation threshold, and generating intermediate durability data; and generating accelerated testing results by combining the intermediate durability data with bending stress distribution analysis. Furthermore, the step of mapping the expected degradation path and determining the long-term performance degradation trend based on the accelerated test results includes: obtaining initial degradation data through the accelerated test results, calculating the equivalent usage time with reference to the equivalent lifetime model, and determining the initial degradation rate; extracting spectral distribution parameters from the initial degradation rate based on standard spectral statistical characteristics, and mapping the curve shape of the expected degradation path; if the curve shape deviates from a preset threshold, correcting the spectral distribution parameters; using the corrected spectral distribution parameters, using simulation methods to predict reliability assessment indicators, and generating a multi-scenario degradation sequence; determining the coefficient of variation of the long-term performance degradation trend based on the multi-scenario degradation sequence, constructing a trend projection model, obtaining the trend stability interval, and mapping the expected degradation path.

[0008] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0009] This invention discloses a performance testing method for flexible connectors, focusing on performance degradation assessment and prediction under dynamic bending conditions. It aims to address the business challenge of accurately evaluating the performance degradation degree of flexible connectors under dynamic loads and predicting their long-term performance trends through multimodal data acquisition and analysis. The invention collects resistance fluctuation and surface image data using a sensor array, employs signal smoothing and image enhancement techniques to purify the data, analyzes the correlation between resistance fluctuation and bending frequency, marks abnormal states, extracts surface damage features, generates a damage distribution map, and fuses contact states and damage features to calculate a comprehensive degradation assessment value. Subsequently, based on an accelerated testing scheme and an equivalent life model, the expected degradation path is mapped to determine the long-term performance trend. The core innovation of this invention lies in the combination of multimodal data fusion and accelerated testing, ensuring assessment accuracy and prediction reliability, and providing a scientific basis for the durability design and maintenance of flexible connectors. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a performance testing method for dynamic bending of a flexible connector sheet according to the present invention. Detailed embodiments.

[0011] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0012] like Figure 1 The performance testing method for dynamic bending of a flexible connector in this embodiment may specifically include:

[0013] Step S101: Multimodal data of the flexible connector is collected under simulated dynamic bending conditions using a sensor array. The bending angle range and dynamic load direction are recorded in real time. At the same time, raw data with high resistance fluctuation amplitude and surface image resolution are acquired to form an initial acquisition dataset.

[0014] Multimodal data of the flexible connector is collected under simulated dynamic bending conditions using a sensor array. Real-time recording is performed on the bending angle range and dynamic load direction. Simultaneously, high-resolution raw data of resistance fluctuation amplitude and surface image are acquired to obtain an initial dataset. Resistance fluctuation amplitude data is obtained from this initial dataset. A data synchronization verification method is used to filter outomas by timestamp matching to determine the fluctuation trend within the bending angle range. Based on this fluctuation trend, the high-resolution surface image data is fused, and the correspondence between multimodal data is obtained through pixel-to-pixel mapping. If the correspondence exceeds a preset threshold, the recording of the dynamic load direction is adjusted. An expanded dataset structure is constructed based on this correspondence, and a final dataset containing anomaly filtering is obtained by adding filter labels.

[0015] Specifically, in one implementation, multimodal data of the flexible connector is acquired using a sensor array under simulated dynamic bending conditions. The first step is to construct a simulation environment. The flexible connector is a connecting component made of flexible materials, commonly used in electronic devices to withstand bending stress. Simulating dynamic bending conditions can be achieved using a robotic arm or a bending test platform capable of controlling the bending angle from 0 to 180 degrees and applying dynamic loads in different directions, such as vertical or horizontal forces. The sensor array, including resistive sensors and image sensors, is positioned at key locations on the flexible connector to capture multimodal data.

[0016] Specifically, the multimodal data encompasses resistance signals and surface images. Resistance sensors monitor internal electrical changes in the material, while image sensors record external deformation. Furthermore, the process of real-time recording for the bending angle range and dynamic load direction is as follows.

[0017] For example, in the simulation environment, the bending angle range is set to 30 degrees to 150 degrees, and the dynamic load direction includes clockwise and counterclockwise bending. A real-time monitoring system collects data once per second to ensure synchronized recording of angle changes and load direction.

[0018] For example, the bending angle is measured by an angle sensor, and the dynamic load direction is detected as vector information by a force sensor. These records form time-series data, which is used for subsequent analysis of the durability of the flexible connector.

[0019] Preferably, acquiring both high-resolution raw data on resistance fluctuation and surface image is a crucial step. Resistance fluctuation refers to the range of resistance change during bending, for example, within ±10% of a reference value, measured using a high-precision ohmmeter. High-resolution raw data on the surface image is captured using a high-definition camera with a resolution of 1920×1080 pixels, ensuring the capture of minute cracks or deformation details.

[0020] In one possible implementation, this data is acquired synchronously via a data acquisition card to avoid latency issues and form an initial dataset. This dataset includes timestamps, angle values, load direction, resistance fluctuation data, and image files, and is stored in a local database.

[0021] It should be noted that the purpose of forming the initial dataset is to provide a basis for the performance evaluation of the soft connector.

[0022] In one embodiment, when collecting data under simulated operating conditions for a flexible connector in an electronic device, a noise filtering module can be introduced to preprocess the resistance signal to remove environmental interference.

[0023] Specifically, noise filtering employs a low-pass filter to preserve fluctuation amplitude information, while image data undergoes edge enhancement algorithms to improve sharpness. This approach ensures the accuracy of the dataset, reflecting the resistive stability and surface integrity of the flexible connector under repeated bending in actual testing.

[0024] Specifically, in another implementation, the sensor array can be optimized into a grid-like distribution, for example, by uniformly placing 16 resistance sensors and 4 image sensors on the surface of the flexible connector. When recording the bending angle range in real time, the dynamic changes are displayed using a software interface, with the load direction represented by a vector diagram. Simultaneously acquired resistance fluctuation data is quantified by calculating the peak-to-valley difference, for example, the fluctuation amplitude is the absolute difference in resistance change. The raw data with high surface image resolution is then used to form a video sequence through continuous shooting, with the resolution controlled to be no less than 2000 pixels wide per frame. This grid arrangement enhances the spatial coverage of the data and improves the comprehensiveness of the dataset.

[0025] For example, during data acquisition under simulated dynamic bending conditions, if the bending angle range extends to extreme values, such as close to 180 degrees, the dynamic load direction can simulate multi-axial force applications. Using a sensor array, the resistance fluctuation amplitude is calculated in real time as the standard deviation of the resistance value, and the surface image is illuminated by multiple light sources to improve resolution. The resulting initial dataset includes multiple sets of records, each corresponding to one bending cycle. This method is applicable to flexible connecting sheets of different thicknesses, demonstrating the versatility of the technical solution. Furthermore...

[0026] Understandably, the technical advantage of this acquisition process lies in providing high-fidelity data to support fatigue analysis of flexible connecting sheets.

[0027] In one embodiment, for thin flexible connectors, resistance fluctuation amplitude monitoring focuses on minute changes, and image resolution emphasizes surface texture details; the resulting dataset can be used to predict breakage points. In another embodiment, for thick flexible connectors, simulated operating conditions emphasize high load directional changes; after the dataset is formed, multimodal information is integrated through a data fusion algorithm to ensure the correspondence between resistance and image data.

[0028] Step S102: For the noise interference type in the initial acquired dataset, signal smoothing technology is used to clean up the resistance fluctuation amplitude data, and the surface image resolution is enhanced and optimized to generate a denoised resistance sequence and a clear image sequence.

[0029] Noise interference types are identified from the initial dataset. Median filtering is used to clean the data by replacing outliers with the median value of the resistance fluctuation amplitude data, resulting in a denoised resistance sequence. For this denoised resistance sequence, the surface image resolution is acquired. Super-resolution reconstruction is used to enhance and optimize the surface image resolution through low-resolution image interpolation and detail restoration, resulting in a clear image sequence. Pixel distribution features are obtained from the clear image sequence. These features are combined with the denoised resistance sequence to determine the sequence trend. When the sequence trend changes, material surface defect indices are acquired, and an optimized resistance image fusion sequence is generated using these indices.

[0030] Specifically, in one implementation, the type of noise interference needs to be identified before preprocessing the initial acquired dataset.

[0031] Specifically, by performing spectral analysis on the leakage current resistance sequence, noise can be classified into high-frequency electromagnetic interference, low-frequency temperature drift noise, and random white noise.

[0032] For example, resistance data collected in coastal areas often contains strong power frequency harmonic interference, while in plateau areas it is dominated by impulse noise. By observing the energy distribution through Fast Fourier Transform, the dominant noise type is identified, providing a basis for selecting a targeted smoothing method. Based on the identified noise type, signal smoothing techniques are used to clean the resistance fluctuation amplitude data.

[0033] In one possible implementation, wavelet denoising is preferred when high-frequency noise is dominant.

[0034] Specifically, the db8 wavelet basis is used to perform a 5-level decomposition of the resistance sequence. A soft threshold is applied to the high-frequency coefficients. This threshold is adaptively adjusted based on a general threshold formula and the signal-to-noise ratio (SNR) of the resistance signal, effectively removing high-frequency interference while preserving the spike characteristics of flashover precursors. After processing, the SNR of the resistance sequence can be improved by more than 15 dB, while the error at the abrupt change edge position is controlled within 3 sampling points, ensuring the reliability of subsequent feature extraction. Furthermore, when low-frequency drift noise exists, a Savitzky-Golay filter can be used for smoothing. This filter removes slow drift while maintaining the overall trend of the resistance waveform through local polynomial least squares fitting.

[0035] For example, selecting a window length of 51 points and a third-order polynomial can restore the resistance sequence to the actual level of contamination leakage in scenarios with baseline drift caused by slow temperature changes, with an error of less than 0.5%. In another implementation, a combined smoothing strategy can be used for resistance sequences containing multiple types of noise. First, empirical mode decomposition is used to decompose the signal into multiple intrinsic mode functions. Then, wavelet thresholding is applied to the high-frequency mode functions for noise reduction, and trend term correction is applied to the low-frequency mode functions. Finally, the purified resistance sequence is reconstructed. This method is stable in coastal humid environments with strong electromagnetic interference. In parallel with the resistance sequence processing, resolution enhancement optimization is performed on the insulator surface image.

[0036] In one embodiment, a single-image super-resolution algorithm based on residual networks is used to enlarge the original 480×480 pixel dirty image to 1920×1920 pixels.

[0037] Specifically, the network contains 16 residual blocks and uses subpixel convolutional layers for upsampling. During training, an equivalent salt-dense labeled image is used as a high-resolution label, which makes the enhanced image more clearly present the gray-dense grains and salt-dense crystal texture.

[0038] Preferably, on edge devices with limited computing resources, a lightweight enhancement method combining bicubic interpolation and guided filtering can be used. First, a 4x bicubic interpolation upscaling is performed, and then the original low-resolution image is used as a guide image for edge-preserving filtering, which sharpens the boundaries of the dirt layer and significantly improves the details of the gray density distribution. The processing time for a single image is less than 80 ms.

[0039] It should be noted that both image enhancement methods described above were completed simultaneously with the resistance sequence processing within the same batch, ultimately generating denoised resistance sequence files and high-resolution image sequence files. The file naming convention includes the acquisition timestamp and insulator number, facilitating subsequent multimodal fusion analysis of pollution conditions. Through these preprocessing methods, the noise amplitude of the resistance sequence was reduced to below 8% of the original, and the peak signal-to-noise ratio of the image was increased to above 35 dB, providing a high-quality data foundation for accurately identifying the pollution levels of various insulator models, such as XP-70 and FC-100P.

[0040] Step S103: Based on the denoised resistance sequence, analyze the correlation characteristics between the resistance fluctuation amplitude and the bending frequency distribution. If the resistance fluctuation amplitude exceeds the preset threshold range, it is marked as an abnormal fluctuation state, and the abnormal fluctuation classification label is determined.

[0041] By analyzing the denoised resistance sequence, the difference between each resistance value and its adjacent values ​​is calculated to obtain a fluctuation amplitude value sequence. For this fluctuation amplitude value sequence, the frequency of bending points in the sequence is calculated, and a bending frequency distribution curve is statistically analyzed. From the bending frequency distribution curve, a correlation feature vector between the fluctuation amplitude value and the bending frequency is extracted. If the correlation feature vector exceeds a preset threshold range, it is marked as an abnormal fluctuation state. Based on the abnormal fluctuation state, the abnormal fluctuations are classified into types, and an abnormal fluctuation classification label is determined.

[0042] Specifically, in one implementation, the resistance fluctuation amplitude is first extracted based on the denoised resistance sequence.

[0043] Specifically, a resistance sequence refers to a continuous sequence of resistance values ​​collected by a sensor, after noise interference has been removed. The fluctuation amplitude can be quantified by calculating the difference or standard deviation between adjacent points in the sequence.

[0044] For example, a sliding window process can be applied to the sequence, calculating the maximum and minimum differences within each window to obtain the amplitude sequence. This extraction helps capture changes in resistance caused by external bending. Further analysis of the bending frequency distribution is then performed. The bending frequency distribution refers to the statistical distribution of the frequency of bending events in the resistance sequence; it can be understood as performing a Fourier transform or peak detection on the sequence to identify periodic fluctuations caused by bending.

[0045] For example, in flexible cable monitoring scenarios, the bending frequency can be distributed by counting the number of resistance peaks per unit time, generating a histogram to represent the probability of occurrence in different frequency ranges. This distribution reflects the regularity of bending, and its correlation with resistance amplitude lies in the fact that high-frequency bending often leads to larger fluctuations.

[0046] Preferably, the correlation characteristics between the resistance fluctuation amplitude and the bending frequency distribution are analyzed. These correlation characteristics can be quantified by calculating the correlation coefficient or mutual information.

[0047] For example, an amplitude vector and a frequency distribution vector are constructed, and Pearson correlation analysis is performed to obtain the correlation strength value. If the correlation strength is higher than a preset threshold, it indicates that the bending frequency directly affects the amplitude change.

[0048] In one possible implementation, this analysis is applied to wearable device sensors, and correlated features help identify differences in patterns between normal use and abnormal bending.

[0049] It should be noted that if the resistance fluctuation exceeds the preset threshold range, it will be marked as an abnormal fluctuation state. The threshold range can be set based on historical data.

[0050] For example, the upper limit is 1.5 times the normal amplitude, and the lower limit is 0.5 times. The marking process involves comparing the current amplitude with the threshold; if it exceeds the threshold, a status flag is set to abnormal. This marking ensures timely detection of potential damage.

[0051] For example, in a cable bending monitoring embodiment, an abnormal fluctuation classification label is determined. The classification label can be divided into minor anomaly, moderate anomaly, and severe anomaly, based on the degree of amplitude deviation and the strength of the associated characteristics.

[0052] Specifically, a decision tree algorithm is used as input for amplitude and frequency features, and outputs labels. For example, amplitudes exceeding 20% ​​and a high proportion of high-frequency bending are marked as severe anomalies. This classification supports subsequent maintenance decisions. In another implementation, for flexible electronic materials, correlation feature analysis can be extended to multidimensional statistics.

[0053] For example, introducing a covariance matrix to calculate the joint distribution of amplitude and frequency enhances feature robustness. This approach demonstrates versatility within the same domain, such as sensor applications under different bending intensities. Furthermore, the entire process logic starts from the resistance sequence input, proceeds through amplitude extraction, frequency distribution calculation, correlation analysis, to threshold judgment and tag determination, ensuring continuous monitoring.

[0054] For example, in practical deployments, this method can enable accurate identification of resistance changes induced by bending, thereby improving equipment reliability.

[0055] In one embodiment, the threshold can be dynamically adjusted, and the correlation feature analysis is optimized based on machine learning feedback to adapt to varying bend environments. This flexibility enhances the applicability of the solution.

[0056] Understandably, through the above steps, the technical solution provides effective anomaly detection in the field of resistance signal processing, supports preventive maintenance, and does not introduce subjective optimization.

[0057] Step S104: If the abnormal fluctuation classification label indicates an abnormality, then the structural damage features of the clear image sequence are extracted using image processing technology, with a focus on detecting the location of surface damage such as wrinkles or cracks, and a damage distribution map is generated.

[0058] Clear image sequences are obtained through abnormal fluctuation labels. For these clear image sequences, an edge detection algorithm is used to extract structural damage features and determine the surface damage locations. If the surface damage locations contain wrinkles, the crack locations are determined by pixel comparison to obtain the damage type distribution. From the damage type distribution, a preset threshold is used to determine the damage severity, generating a preliminary damage layer. Based on the preliminary damage layer, the location coordinates are integrated to generate a damage distribution map. For cases where abnormal fluctuation labels indicate anomalies, the clear image sequences are processed. After obtaining the clear image sequences, the Canny edge detection algorithm is used. This algorithm first calculates the image gradient, then applies non-maximum suppression to suppress non-edge pixels, and finally uses double thresholds to connect edges to extract structural damage features. After obtaining the structural damage features, surface damage locations are detected. If wrinkles are detected, the crack extension direction is determined by pixel grayscale differences to obtain the crack location coordinates. From the crack location coordinates, a preset threshold is used to determine the damage boundary, generating a damage boundary map. Based on the damage boundary map, the surface damage locations of wrinkles and cracks are integrated, and the damage area ratio is calculated to obtain a damage severity assessment. The severity assessment of the injury is integrated into the distribution map generation, and the location of the injury is visualized through coordinate mapping to generate the injury distribution map.

[0059] Specifically, in one implementation, when the abnormal fluctuation classification label indicates an anomaly, the system activates the image processing module to extract structural damage features from the acquired clear image sequence. This process first involves image preprocessing to improve image quality, such as by denoising and enhancing contrast to ensure the damage features are clearly visible.

[0060] Specifically, clear image sequences typically originate from power transmission line monitoring equipment, such as high-definition cameras mounted on poles. These images capture the surface condition of conductors under conditions such as wind or freezing. Furthermore, the focus of extracting structural damage features is on detecting the location of surface damage such as wrinkles or cracks.

[0061] For example, edge detection algorithms can be used to identify discontinuous lines in an image, such as using the Canny operator to detect edge intensity, thereby locating potential cracks.

[0062] It should be noted that wrinkle detection is based on texture analysis principles, identifying surface deformation areas by calculating the grayscale change rate of local image regions. This method helps distinguish between normal textures and abnormal damage, ensuring that the extracted features accurately reflect structural problems. In actual operation, for monitoring power lines, the system analyzes the image sequence frame by frame, marking the damage initiation point and expansion direction, forming a preliminary damage coordinate set.

[0063] Preferably, after feature extraction, the damage distribution map is generated by mapping the detected damage locations to a two-dimensional or three-dimensional coordinate system.

[0064] For example, projecting damage points from an image sequence onto a conductor model generates a heatmap-like distribution map, where color depth indicates the severity of the damage. This distribution map can visually display areas of concentrated damage, facilitating quick problem location for maintenance personnel.

[0065] In one possible implementation, if multiple images in a sequence show the evolution of a crack at the same location, the system accumulates a damage index to generate a dynamic distribution map to track damage progression.

[0066] Understandably, the application of this image processing technology in the field of power line monitoring can support a variety of scenarios, such as real-time detection in strong wind environments.

[0067] Specifically, for different types of conductors, such as aluminum stranded wire or steel-cored aluminum stranded wire, the extraction process can adjust parameters to adapt to material differences. For example, for crack detection of steel-cored conductors, the edge threshold can be increased to avoid noise interference, thereby ensuring the robustness of feature extraction.

[0068] In one embodiment, considering a nighttime monitoring scenario, the system integrates infrared image sequences to assist visible light images, extracting the correlation between thermal anomalies and surface damage.

[0069] For example, when a tag indicates an anomaly, thermal distribution analysis is first performed on the infrared image, and then combined with crack detection in the visible light image to generate a comprehensive damage distribution map. This approach expands the applicability of the technology, effectively identifying wrinkle locations even under low-light conditions. Furthermore, the generation of the damage distribution map includes post-processing steps, such as smoothing boundaries to remove isolated noise points.

[0070] Specifically, morphological operations such as expansion and corrosion are used to refine damaged areas, ensuring the accuracy of the distribution map. In power transmission line inspection operations, this map can be output to the monitoring center to support remote decision-making.

[0071] For example, in scenarios implemented under freezing weather conditions, the system focuses on detecting ice cracks on the surface of the conductor. By comparing the differences between consecutive frames in an image sequence, dynamic damage features are extracted, and a distribution map covering the entire line segment is generated. This method helps predict potential line breakage risks and provides a basis for preventative maintenance.

[0072] It should be noted that the entire extraction and generation process is based on a modular design, with the image processing module and the anomaly labeling module closely integrated to ensure logical consistency from anomaly indication to distribution map output. In practical applications, this technical solution enables timely visualization of structural damage, supporting the stable operation of power systems.

[0073] Step S105: Based on the damage distribution map and abnormal fluctuation classification labels, the contact state index and structural damage characteristics are integrated to calculate the performance degradation degree of the soft connector under dynamic load direction and obtain a comprehensive degradation assessment value.

[0074] The coordinates of damage points on the surface of the flexible connector are obtained by using a damage distribution map, and abnormal fluctuation regions are identified by fusing abnormal fluctuation classification labels. Contact state indicators are extracted from these abnormal fluctuation regions, and structural damage features are fused to obtain a damage fusion vector. The load response curve under dynamic load direction is calculated based on the damage fusion vector to determine the degree of performance degradation. If the degree of performance degradation exceeds a preset threshold, degradation trend prediction is used to adjust the load response curve to obtain a degradation correction value. The degradation correction value is then integrated with the damage fusion vector to obtain a comprehensive degradation assessment value.

[0075] Specifically, in one implementation, a damage distribution map is first obtained. This map is a two-dimensional or three-dimensional representation obtained by scanning and imaging the surface of the flexible connector, reflecting the location, area, and severity of damage in the structure.

[0076] For example, in the monitoring of blade connection components in wind power equipment, damage distribution maps can be generated in the form of heat maps based on data collected by optical scanning equipment. The color depth represents damage density, thus providing basic data support for subsequent fusion. The generation process of this distribution map includes data acquisition, image processing, and damage identification steps, ensuring accurate capture of minute cracks or deformations caused by dynamic loads. Furthermore, abnormal fluctuation classification labels are the analysis results of vibration signals from flexible connecting pieces under dynamic loads.

[0077] Specifically, abnormal fluctuations refer to abnormal amplitude or frequency deviations caused by load changes. Vibration data is monitored in real time by sensors, and then classification algorithms such as support vector machines are applied to label the fluctuations, for example, classifying them into three categories: slight abnormality, moderate abnormality, and severe abnormality. This classification helps identify potential risks. In the operational process, time-series data is first collected, then features such as peak values ​​and spectra are extracted, and then labels are generated according to preset thresholds. Ensuring the accuracy of the labels directly affects the reliability of the degradation assessment.

[0078] Preferably, the contact state index is used to describe the tightness of the connection between the flexible connector and adjacent components, including quantitative values ​​such as the coefficient of friction, gap size, and pressure distribution.

[0079] In one possible implementation, these metrics are measured using embedded sensors; for example, in wind power equipment, pressure sensors monitor real-time pressure at the contact surfaces and calculate a numerical vector of the contact state. The acquisition of these metrics involves calibrating the sensors, filtering noise, and calculating averages to provide a stable data foundation for fusion.

[0080] It should be noted that structural damage features are quantitative descriptions extracted from damage distribution maps, such as feature vectors for crack length, depth, and propagation direction. The specific process includes image segmentation, edge detection, and feature quantization. For example, edge detection algorithms are used to identify crack outlines, and then geometric parameters are calculated to form a feature set. This feature extraction emphasizes changes under dynamic loading; for example, under repeated wind loads, features may show propagation trends, laying the foundation for the fusion step.

[0081] In one embodiment, the process of fusing contact state indicators and structural damage characteristics employs a weighted summation or neural network method to map the two into a unified space.

[0082] For example, the contact status index is first normalized, then concatenated with the damage feature vector, and finally a comprehensive vector is calculated using a fusion model. This fusion effectively integrates multi-source information, improves the comprehensiveness of the assessment, reduces misjudgments in business operations, and ensures more reliable maintenance decisions for flexible connectors.

[0083] For example, when calculating the performance degradation of the soft connector under dynamic load direction, a degradation model is introduced based on the fusion result.

[0084] Specifically, the dynamic load direction refers to the direction of the main axis of external forces such as wind. The degree of degradation is calculated using a formula, such as multiplying the fused vector by the load coefficient to obtain a degradation score. This process includes: first, determining the load direction vector; then, projecting the fused features onto this direction to quantify the degradation impact. For example, in wind power equipment, the fatigue accumulation of connecting pieces due to wind speed changes is considered, thus deriving a numerical representation of the degree of degradation. This calculation emphasizes directionality, ensuring the assessment is tailored to a specific load scenario. Further, obtaining the comprehensive degradation assessment value is the output step that combines the degree of degradation with abnormal fluctuation classification labels. In one implementation, all inputs are integrated through a multi-classifier, such as using a decision tree model to input fused data and labels, outputting an assessment value from 0 to 1, where a higher value indicates more severe degradation. This assessment value is used in operations to predict maintenance time; for example, triggering an alarm when the value exceeds a threshold to achieve preventative intervention.

[0085] Understandably, in another embodiment, the method can be applied to the monitoring of tower connection plates in similar wind power equipment, adjusting the fusion weights to adapt to different load intensities.

[0086] For example, in light-load scenarios, contact indicators are emphasized, while in heavy-load scenarios, damage characteristics are prioritized, thus demonstrating the flexibility of the solution.

[0087] Specifically, the logical flow of the entire process from data acquisition to evaluation output ensures continuity. Through the above steps, objective monitoring of the performance of the flexible connector can be achieved, improving equipment lifespan and safety in practical applications.

[0088] Step S106: Based on the comprehensive degradation assessment value, combined with the statistical characteristics of the standard spectrum and the accelerated bending frequency, an accelerated testing scheme is constructed. By increasing the accelerated bending amplitude within the linear range of the material response, a compression test is conducted for the durability verification cycle, and accelerated test results are generated.

[0089] By collecting initial material state data and combining it with standard spectral statistical characteristics (where the mean and variance are extracted from preset spectral data), a comprehensive degradation assessment value is calculated to obtain the initial degradation level. For this initial degradation level, the accelerated bending frequency is obtained, and an accelerated testing scheme is constructed to determine the bending parameter set. Based on the bending parameter set, the accelerated bending amplitude is increased, and the stress input is adjusted within the linear range of the material response. The stress input adjustment is achieved by obtaining a linear response curve through incremental load simulation. Using the linear response curve, compression tests are performed for a durability verification cycle, monitoring the degradation threshold. The degradation threshold is obtained from a preset degradation model to generate intermediate durability data. From this intermediate durability data, combined with bending stress distribution analysis, the accelerated test results are generated by mapping stress points to a distribution map.

[0090] Specifically, in one implementation, for the field of material durability testing, a comprehensive degradation assessment value is first obtained, which is calculated by integrating multiple degradation indicators.

[0091] Specifically, the comprehensive degradation assessment value is a quantitative index obtained by weighting and summing factors such as fatigue damage, crack propagation rate, and surface corrosion degree accumulated by the material under normal use conditions.

[0092] For example, when testing metallic alloys, initial strength data and post-use strength decay data can be collected. The degradation percentage can be calculated using a formula, and then corrected for environmental factors such as temperature and humidity, resulting in a comprehensive value reflecting the overall degradation state of the material. This assessment helps in the precise design of subsequent acceleration programs, ensuring that the testing approach is based on actual degradation. Furthermore, standard spectral statistical features are incorporated. These features are statistical parameters extracted from standard load spectra, including mean, standard deviation, and peak frequency. A standard spectrum refers to the typical load distribution that a material experiences in real-world applications; for example, in structural engineering testing, a standard spectrum can represent the distribution pattern of repeated bending loads. The process of extracting statistical features involves performing a Fourier transform on the load spectrum data to identify dominant frequencies, and then calculating the statistical distribution of these frequencies to obtain a feature vector. This feature helps quantify the randomness and intensity distribution of the load, providing a data foundation for accelerated testing.

[0093] Preferably, an accelerating bending frequency is introduced, which accelerates the degradation process by increasing the standard bending frequency. Specifically, the accelerating bending frequency is obtained by multiplying the bending frequency in the standard spectrum by an acceleration factor.

[0094] For example, if the standard frequency is 10Hz, it can be accelerated to 20Hz or higher to shorten the test cycle.

[0095] It should be noted that when selecting the acceleration factor, the fatigue limit of the material must be considered to ensure that increasing the frequency does not lead to a nonlinear response, thereby maintaining the reliability of the test results. This frequency adjustment is a key step in constructing an accelerated testing scheme, enabling the simulation of long-term usage effects within a finite time.

[0096] In one possible implementation, the accelerated testing scheme combines a comprehensive degradation assessment value with standard spectral statistical characteristics and accelerated bending frequencies. The specific process involves first setting a target degradation threshold based on the degradation assessment value, then optimizing the load distribution using statistical characteristics, and finally adjusting the bending frequency to match the acceleration requirements.

[0097] For example, when testing composite materials, these elements can be fused using iterative algorithms to generate a test protocol that includes time, amplitude, and frequency parameters. This approach ensures the targeted nature and efficiency of the test, enabling precise control of degradation simulation through a data-driven approach.

[0098] For example, compression testing for durability verification cycles can be performed by increasing the accelerated bending amplitude within the linear range of the material's response. This linear range refers to the region where the material's stress-strain curve remains linear, typically defined by experimentally determining the material's elastic modulus. Increasing the amplitude involves gradually increasing the bending load, for example, starting with 1.2 times the standard amplitude, while monitoring the material's response curve to ensure it does not exceed the linear limit. This method can compress verification cycles that would otherwise take months into weeks. For instance, in verifying the durability of aerospace materials, increasing the amplitude can simulate thousands of bending cycles, thereby accelerating the verification of the material's long-term performance. Furthermore, during the compression test, the material's response is monitored in real time to verify the linear range.

[0099] Specifically, sensors are used to collect strain data during the bending process, which is then compared with a linear model. If the deviation exceeds a threshold, the amplitude is adjusted. This monitoring ensures the validity of the test data and avoids deviations in results caused by nonlinear effects.

[0100] In one embodiment, for testing high-strength steel, the bending radius can be increased to 1.5 times the standard value while maintaining an accelerated frequency, reducing the cycle from 1000 hours to 200 hours.

[0101] Understandably, generating accelerated test results involves summarizing test data and calculating final degradation metrics. Specifically, this process compares measurements from compression tests, such as crack length and strength loss, with initial assessment values ​​to generate a report.

[0102] For example, the results can show the time it takes for a material to reach a preset degradation level under accelerated conditions, thus assessing its durability. These results generate decisions that support subsequent material optimization. In another implementation, the scheme parameters can be adjusted for different material types; for example, for polymer materials, the acceleration factor can be reduced to accommodate their lower linear range threshold. This flexibility demonstrates the versatility of the scheme, ensuring application in multiple scenarios within the same testing domain.

[0103] Preferably, the entire process can be integrated into an automated testing system, with parameters adjusted via software control for efficient execution. This integration improves the repeatability and accuracy of the test. Finally, the accelerated test results obtained through the above steps can objectively reflect the durability performance of the material, providing a reliable basis for engineering applications.

[0104] Step S107: Based on the accelerated test results, and referring to the equivalent lifetime model and the predictive reliability assessment method, the expected degradation path of the soft connector under the standard spectrum statistical characteristics is mapped to determine the long-term performance degradation trend.

[0105] By accelerating the test results, the initial degradation data of the flexible connector under vibration load simulation is obtained. Referring to the equivalent life model, the equivalent service life is calculated by multiplying the acceleration factor by the test duration, thus determining the initial degradation rate. From this initial degradation rate, spectral distribution parameters are extracted based on standard spectral statistical characteristics to map the curve shape of the expected degradation path. If the curve shape deviates from a preset threshold, the spectral distribution parameters are corrected. The corrected spectral distribution parameters are obtained, and a Monte Carlo simulation method is used with these parameters as input to predict reliability assessment indicators, generating a multi-scenario degradation sequence of the flexible connector under material fatigue threshold conditions. Through these multi-scenario degradation sequences, the coefficient of variation of the long-term performance degradation trend is determined. A trend projection model is constructed based on the coefficient of variation, using it as input to obtain the stable interval of the trend. From the stable interval of the trend, the expected degradation path of the flexible connector under standard spectral statistical characteristics is mapped, determining the long-term performance degradation trend.

[0106] Specifically, in one implementation, the flexible connector is used as a component for flexible connections in electronic devices, and its accelerated testing results are experimental data obtained by simulating extreme environmental conditions.

[0107] Specifically, accelerated testing involves placing the flexible connector in high-temperature, high-humidity, or vibrating environments and recording changes in its performance parameters, such as resistance or mechanical strength degradation. These results provide foundational data for subsequent model applications.

[0108] It should be noted that the purpose of accelerated testing is to shorten the testing cycle, simulate long-term use effects, and thus efficiently obtain degradation information. Furthermore, an equivalent life model is used to convert accelerated test results into life estimates under standard operating conditions. This model is based on the Arrhenius equation or other thermodynamic principles, incorporating acceleration factors such as temperature or stress levels. For example...

[0109] In one possible implementation, an acceleration factor is first calculated, determined by the ratio of the test temperature to the standard temperature. The test time is then multiplied by this factor to obtain the equivalent lifetime value. This conversion process ensures the applicability of the test data and avoids biases caused by directly using accelerated data.

[0110] Understandably, the core of the equivalent lifetime model lies in the precise selection of parameters, such as the determination of the activation energy, which directly affects the accuracy of the mapping. In practical applications, such as for flexible connectors in battery packs, the model can help predict lifetimes under normal operating temperatures, reducing the need for field testing.

[0111] Preferably, the predictive reliability assessment method combines statistical tools to analyze equivalent lifetime data.

[0112] Specifically, the method involves constructing a reliability function, such as the Weibull distribution, to assess the probability of failure.

[0113] For example, key metrics, such as mean time to failure, are first extracted from accelerated testing results, and then Monte Carlo simulations are applied to generate reliability curves under various scenarios. This approach involves data fitting and parameter estimation to ensure the robustness of the predictions.

[0114] In one embodiment, standard spectral statistical characteristics of the flexible connector, such as the mean and variance of the load spectrum, are input into the model to generate a reliability assessment report. The advantage of this method is its ability to quantify uncertainty and provide confidence intervals, thereby supporting decision-making. In one implementation, standard spectral statistical characteristics refer to a statistical description of the flexible connector under standard operating conditions, including statistical values ​​of vibration frequency distribution or thermal cycling count.

[0115] Specifically, these features are obtained through historical data or standard specifications, such as extracting the average load spectrum from industry standards. The accelerated testing results are then mapped onto these features to form the expected degradation path.

[0116] It should be noted that the mapping process includes curve fitting, such as using multinomial regression, to connect test points into a continuous path. This path illustrates how performance parameters change over time, such as the fatigue crack propagation curve of a connecting piece. In business scenarios, this helps identify potential failure points and improves design iteration efficiency. Furthermore, the mapping of expected degradation paths is based on a combination of equivalent life models and standard spectral characteristics. For example...

[0117] In one possible implementation, the degradation data from accelerated testing is first converted to a standard timescale using an equivalent model, and then statistical perturbations of the standard spectrum are superimposed to generate a path curve. The specific process includes data normalization and interpolation to ensure a smooth path. This path can be used to simulate degradation under different usage intensities; for example, a path in a moderate vibration environment shows a slow increase in resistance, while high vibration accelerates crack formation.

[0118] Understandably, this mapping enhances the versatility of the predictions, making them applicable to a variety of flexible connector types, such as copper-based or aluminum-based materials.

[0119] In one embodiment, determining a long-term performance degradation trend involves trend analysis of the expected degradation path.

[0120] Specifically, a trend line is extracted from the path through linear regression or exponential fitting, for example, by calculating the slope value of the degradation rate.

[0121] For example, for flexible connectors, the trend might show slow degradation in the initial stage, accelerating later, reflecting material fatigue accumulation. The process of determining this trend includes setting thresholds, such as marking a high-risk point when performance degradation exceeds 20%. In practical applications, this trend analysis can guide maintenance planning and extend equipment life.

[0122] Preferably, in another embodiment, the mapping accuracy is improved by combining multiple sets of accelerated test results.

[0123] Specifically, test data from different batches of soft connectors are collected, and Bayesian update methods are applied to adjust the parameters of the equivalent lifetime model. This process involves setting the prior distribution and calculating the posterior distribution to ensure the model adapts to variability.

[0124] For example, in scenarios with changing environmental factors, path mapping can display a wider confidence band, providing a conservative estimate. This embodiment demonstrates the flexibility of the technical solution and supports quality control in mass production. Furthermore, the assessment of long-term performance degradation trends can be aided by machine learning, such as using neural networks to predict path extension.

[0125] It should be noted that this network takes standard spectral features as input, is trained on accelerated test data, and outputs a long-term trend curve. The process includes feature engineering and model validation to avoid overfitting.

[0126] In one embodiment, for applications of flexible connectors under continuous vibration, trends show a significant performance degradation after 5000 hours, which helps in setting warranty periods. The technical advantage of this method is that it improves the automation of prediction and reduces human error.

[0127] In one possible implementation, the entire process begins with accelerated testing and continues with a logical flow that ensures consistency as trends are established.

[0128] For example, test results are first processed using an equivalent model, then mapped to a standard spectrum to form a path, and finally the trend is analyzed. This connection is achieved through data flow, such as using software tools to integrate the various steps.

[0129] Understandably, this solution is applicable to flexible connectors in electronic modules, covering scenarios from design to maintenance.

[0130] For example, in actual business operations, degradation path mapping of soft connectors can generate reports for risk assessment.

[0131] Specifically, the report includes roadmaps and trend indicators to support decisions such as material optimization. This implementation emphasizes the method's practicality, extending to connecting pieces of different specifications within the same field to ensure broad applicability.

[0132] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for testing the performance of dynamic bending of a flexible connector, characterized in that, include: S101, multimodal data of the flexible connector under dynamic bending conditions are collected by a sensor array to obtain an initial collection dataset. The bending angle range and dynamic load direction are recorded in real time by the sensor array. At the same time, resistance fluctuation amplitude data and high-resolution surface image data are acquired to form the initial collection dataset. Resistance fluctuation amplitude data is extracted from the initial collection dataset. Outliers are filtered by a timestamp matching method to determine the fluctuation trend within the bending angle range. The high-resolution surface image data is fused according to the fluctuation trend. The correspondence between multimodal data is obtained by pixel correspondence mapping. If the correspondence exceeds a preset threshold, the dynamic load direction recording is adjusted. An extended dataset structure is constructed according to the correspondence. Filter labels are added to generate the final collection dataset. S102, Denoise the initial acquired dataset to generate a denoised resistance sequence and a clear image sequence; S103, Analyze the correlation characteristics between fluctuation amplitude and bending frequency based on the denoised resistance sequence to determine abnormal fluctuation classification labels; S104, Extract structural damage features from the clear image sequence to generate a damage distribution map; S105, Based on the damage distribution map and the abnormal fluctuation classification labels, fuse contact state indicators and structural damage features to calculate the degree of performance degradation and obtain a comprehensive degradation assessment value; S106, Based on the comprehensive degradation assessment value, Construct an accelerated testing scheme by combining standard spectrum statistical features to generate accelerated testing results; S107, Based on the accelerated testing results, Map the expected degradation path to determine the long-term performance degradation trend.

2. The performance testing method for dynamic bending of a flexible connector as described in claim 1, characterized in that, Step S102 further includes: The noise interference type is obtained from the initial dataset, and the resistance fluctuation amplitude data is purified by median filtering to obtain a denoised resistance sequence. For the denoised resistance sequence, the surface image resolution is obtained, and a super-resolution reconstruction method is used to enhance and optimize it through low-resolution image interpolation and detail restoration to generate a clear image sequence. Pixel distribution features are extracted from the clear image sequence, and the sequence trend change is determined in conjunction with the denoised resistance sequence. Based on the sequence trend change, material surface defect indices are obtained, and an optimized resistance image fusion sequence is generated.

3. The performance testing method for dynamic bending of a flexible connector as described in claim 1, characterized in that, Step S103 further includes: calculating the resistance value difference through the denoised resistance sequence to obtain a fluctuation amplitude value sequence; calculating the frequency of bending points for the fluctuation amplitude value sequence and statistically analyzing the bending frequency distribution curve; extracting the correlation feature vector between the fluctuation amplitude value and the bending frequency from the bending frequency distribution curve; if the correlation feature vector exceeds a preset threshold range, it is marked as an abnormal fluctuation state; classifying the abnormal fluctuation state according to its type and determining the abnormal fluctuation classification label.

4. The performance testing method for dynamic bending of a flexible connector as described in claim 1, characterized in that, Step S104 further includes: for the clear image sequence, using an edge detection algorithm to extract structural damage features and determine the surface damage location; if the surface damage location contains wrinkles, then determining the crack extension direction by pixel grayscale difference and obtaining the crack location coordinates; from the crack location coordinates, determining the damage boundary by a preset threshold and generating a damage boundary map; integrating the surface damage locations of wrinkles and cracks according to the damage boundary map, calculating the damage area ratio, and obtaining a damage severity assessment; and fusing the damage severity assessment into the distribution map through coordinate mapping to generate a damage distribution map.

5. The performance testing method for dynamic bending of a flexible connector as described in claim 1, characterized in that, Step S105 further includes: obtaining the coordinates of surface damage points through the damage distribution map, and determining the abnormal fluctuation region by combining the abnormal fluctuation classification label; extracting contact state indicators based on the abnormal fluctuation region, and generating a damage fusion vector by fusing structural damage features; calculating the load response curve under dynamic load direction for the damage fusion vector to determine the degree of performance degradation; if the degree of performance degradation exceeds a preset threshold, adjusting the load response curve using degradation trend prediction to obtain a degradation correction value; and integrating the damage fusion vector through the degradation correction value to generate a comprehensive degradation assessment value.

6. The performance testing method for dynamic bending of a flexible connector as described in claim 1, characterized in that, Step S106 further includes: calculating the initial degradation level by combining the comprehensive degradation assessment value with standard spectral statistical characteristics; obtaining the accelerated bending frequency for the initial degradation level, constructing an accelerated testing scheme, and determining the bending parameter set; increasing the accelerated bending amplitude according to the bending parameter set, adjusting the stress input within the linear range of the material response, and obtaining a linear response curve; conducting compression tests for durability verification cycles using the linear response curve, monitoring the degradation threshold, and generating intermediate durability data; and generating accelerated test results by combining the intermediate durability data with bending stress distribution analysis.

7. The performance testing method for dynamic bending of a flexible connector as described in claim 1, characterized in that, Step S107 further includes: obtaining initial degradation data through the accelerated test results, calculating the equivalent service life with reference to the equivalent lifetime model, and determining the initial degradation rate; extracting spectral distribution parameters from the initial degradation rate based on standard spectral statistical characteristics, and mapping the curve shape of the expected degradation path; if the curve shape deviates from a preset threshold, correcting the spectral distribution parameters; using the corrected spectral distribution parameters, using simulation methods to predict reliability assessment indicators, and generating a multi-scenario degradation sequence; determining the coefficient of variation of the long-term performance degradation trend based on the multi-scenario degradation sequence, constructing a trend projection model, obtaining the trend stability interval, and mapping the expected degradation path.

Images