Method for detecting fatigue resistance of rubber plate

Abstract

This invention discloses a fatigue resistance testing method for rubber sheets, relating to the field of rubber material performance testing technology. Its key technical points are: the testing area is selected as the intact rubber surface in front of the initial cut, eliminating the need to directly target the crack area for acquisition and crack contour identification. It is unaffected by the pre-cut processing precision or the irregular extension morphology of the crack; image acquisition can be completed with only a conventional industrial camera and stable supplementary lighting, reducing the equipment investment threshold and minimizing environmental interference with the test results. Secondly, this method does not require continuous loading until the sample crack extends to the failure length; only grayscale data of the surface testing area needs to be acquired in the early stages of cyclic loading to complete the entropy evolution stage determination and feature parameter extraction, thereby estimating the complete fatigue life. This effectively compresses the testing time for a single sample and is suitable for rapid screening of large batches of samples.

Description

Technical Field

[0001] This invention relates to the field of rubber material performance testing technology, and in particular to a method for testing the fatigue resistance of rubber sheets. Background Technology

[0002] Rubber sheets, as basic components for sealing, shock absorption, and load bearing, are widely used in engineering machinery, rail transportation, power infrastructure and other fields. Fatigue resistance is the core indicator that determines its service life and operational reliability, and it is also a mandatory inspection item for incoming materials inspection and batch quality sampling. Currently, the industry generally adopts the flexural fatigue test scheme for testing. In this process, a standard specimen with a pre-made notch is first prepared. The specimen is then clamped in a fatigue testing machine and a reciprocating tensile load is continuously applied. The crack propagation state is monitored at the notch position throughout the process until the crack extends to the preset failure length. Finally, the fatigue life of the specimen is characterized by the total number of cycles. Under this implementation mode, the entire testing process needs to fully cover all stages from internal micro-damage accumulation and macro-crack initiation to crack propagation failure. It cannot output life results in the early stage of damage accumulation. The testing cycle for a single specimen is long, and the testing efficiency is limited when dealing with large batches of samples.

[0003] Some improved solutions introduce image recognition technology to replace manual crack observation. During implementation, the imaging device is aimed at the notch area, and the crack outline is extracted and the length change is calculated through image segmentation algorithms. However, these solutions only optimize the execution method of crack observation. The detection accuracy always depends on the imaging quality of the crack area. The flatness of the pre-made notch, the stains on the sample surface, and slight fluctuations in ambient light can all interfere with the crack identification results. At the same time, in order to capture the propagation of minute cracks, high-magnification microscopic imaging equipment is usually required, which results in relatively high detection costs and environmental requirements. Therefore, we propose a fatigue resistance detection method for rubber sheets. Summary of the Invention

[0004] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a fatigue resistance testing method for rubber sheets. It characterizes the accumulation of internal micro-damage by analyzing the evolution of grayscale distribution entropy values ​​in the intact region in front of the crack. It can complete the full life prediction without loading until the crack fails. At the same time, it improves the detection accuracy through multi-layer data processing optimization, shortening the detection cycle, reducing detection costs, and ensuring detection reliability.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A fatigue resistance testing method for rubber sheets, comprising the following steps: Prepare a test sample with an initial incision, and delineate a surface detection area of ​​a fixed size on the intact rubber surface in front of the incision; Place the test sample in a sealed, light-proof, constant-temperature environment and let it stand until the overall temperature of the test sample is uniform and stable. A periodic reciprocating tensile load with constant parameters is applied to the test sample, and image data of the surface detection area is acquired at the zero-stress reset node of each cyclic loading cycle. For each image, perform gray-level segmentation and pixel proportion statistics on the surface detection area, calculate the gray-level distribution entropy value of the corresponding loading cycle, and generate an entropy value time sequence by arranging them in the loading order; Based on the periodic incremental change pattern of entropy time series, three consecutive entropy evolution stages are divided, and the linear slope of the gray-scale distribution entropy change within a specified evolution stage is extracted as the feature slope. By establishing a pre-defined relationship between characteristic slope and fatigue life, and combining it with the weight correction of the stationary segment, the fatigue life of the test sample after calibration is estimated. Output the fatigue resistance test results of the sample to be tested.

[0006] Preferably, the specific steps for preparing the test sample with an initial incision and delineating a surface detection area of ​​a fixed size on the intact rubber surface in front of the incision are as follows: The rubber sheet to be tested is cut into sheet-like samples with length, width, and thickness all conforming to the preset dimensions; An initial cut with a depth and width that meets the set values ​​is pre-made in the middle of one edge of the sheet-like specimen, and the extension direction of the initial cut is perpendicular to the force direction of the subsequent tensile loading. Starting from the tip of the initial incision, a rectangular area of ​​fixed length and width is delineated on the intact rubber surface in front of the incision along the direction of tensile force as a surface inspection area. The surface inspection area does not cover the initial incision body or the subsequent crack path.

[0007] Preferably, the steps for placing the test sample in a sealed, light-proof, constant-temperature environment until the overall temperature of the test sample is uniform and stable are as follows: Place the sample to be tested horizontally in a sealed, light-proof, constant-temperature chamber, set the internal temperature of the chamber to the preset detection temperature, and keep the air inside the chamber still. The continuous static time of the test sample under this temperature environment shall not be less than the preset static threshold. Once the overall temperature of the test sample is consistent with the internal temperature of the cavity, environmental pretreatment is completed.

[0008] Preferably, the steps for applying a constant cyclic reciprocating tensile load to the test sample and acquiring image data of the surface detection area at the zero-stress reset node of each cyclic loading cycle are as follows: The two ends of the pre-treated test sample are fixed to the two clamping ends of the tensile fixture. The reciprocating stroke of the tensile fixture is set to a fixed value so that the maximum elongation of the test sample after each stretching meets the preset deformation ratio. The reciprocating motion frequency of the tension fixture is set to a fixed value, which drives the test sample to continuously perform reciprocating tension motion. In each cycle of loading, the test sample is stretched from a stress-free state to the maximum elongation state, and then reset to a stress-free state. Image acquisition is triggered at the reset endpoint of each cyclic loading cycle, the instant the test sample is in a zero-stress static state. During the acquisition process, the supplementary light intensity remains constant, and the acquisition angle is perpendicular to the surface detection area to obtain distortion-free and clear image data of the surface detection area.

[0009] Preferably, the steps for performing grayscale level division and pixel proportion statistics on the surface detection area of ​​each image, calculating the grayscale distribution entropy value of the corresponding loading cycle, and arranging the entropy value time sequence according to the loading order are as follows: Using the pixel coordinates of the initial incision tip as a reference, rigid registration and coordinate alignment calibration are performed on all acquired images to make the surface detection area coordinates of each image coincide. The pixel grayscale values ​​in the surface detection area are divided into a preset number of equally spaced grayscale levels within the full grayscale range of 0 to 255, and the proportion of the number of pixels contained in each grayscale level to the total number of pixels in the surface detection area is calculated. Take the common logarithm to the base 10 for the pixel proportion of each gray level, multiply the resulting logarithmic value by the pixel proportion of the corresponding level, sum the product of all gray levels and take the negative value to obtain the gray level distribution entropy value of the corresponding loading cycle. Arrange all grayscale distribution entropy values ​​in the order of the cyclic loading cycles to generate a complete entropy value time series.

[0010] Preferably, based on the periodic incremental change pattern of the entropy time series, the steps for dividing the entropy evolution into three consecutive stages are as follows: The difference between the grayscale distribution entropy values ​​of two adjacent cycles is calculated sequentially according to the cyclic loading cycle order to obtain the single-cycle entropy increment; Specifically, the interval where the single-cycle entropy increment is consistently lower than the first preset threshold and for no less than a preset number of cycles is defined as the entropy stable stage, corresponding to the initial state where micro-damage inside the rubber has not yet begun to emerge in large quantities; the continuous interval where the single-cycle entropy increment is between the first preset threshold and the second preset threshold and the fluctuation amplitude between adjacent cycles does not exceed the limit value is defined as the entropy slowly rising stage, corresponding to the transitional state where micro-damage inside the rubber is accumulating stably; and the interval where the single-cycle entropy increment is consistently higher than the second preset threshold and for no less than a preset number of cycles is defined as the entropy sharply rising stage, corresponding to the state of near failure where macroscopic cracks are rapidly propagating. The three stages are connected sequentially according to the order of the cyclic loading cycle, with no overlapping or gaps.

[0011] Preferably, the steps for extracting the linear slope of the gray-level distribution entropy change within a specified evolutionary stage as the feature slope are as follows: Extract the middle section of data during the entropy increase phase, and extract the starting and ending points from the start and end points of the phase by a preset proportional length, removing the fluctuating data at the phase transition nodes. Perform least-squares linear fitting on the extracted grayscale distribution entropy data, and control the determination coefficient of the fitting result to be no less than the preset fitting threshold. The entropy increment value corresponding to the unit cyclic loading cycle obtained from the fitting is used as the feature slope.

[0012] Preferably, the steps for constructing the pre-established correspondence between characteristic slope and fatigue life are as follows: Using standard specimens of the same material, specifications, and preparation process as the test specimen, multiple sets of complete fatigue tests were conducted under consistent test temperature and deformation ratio conditions. The characteristic slope and corresponding standard fatigue life of each set of standard specimens were recorded. The standard fatigue life is the total number of loading cycles when the standard specimen is loaded until the crack length reaches the preset critical failure length. By fitting multiple sets of measured data, a positive correlation between the reciprocal of the characteristic slope and the standard fatigue life was obtained as a power function relationship, which is used as the correspondence between the characteristic slope and the fatigue life. By substituting the characteristic slope of the test sample into the corresponding relationship, a preliminary prediction of fatigue life can be obtained.

[0013] Preferably, the steps for calculating the calibrated fatigue life of the test sample after adjustment, combined with the weight correction of the stationary segment, are as follows: The proportion of the total number of cycles in the steady-state phase of entropy to the preliminary predicted fatigue life is calculated. The correction coefficient is matched according to the range in which the ratio is located. If the ratio is in the preset standard range, the correction coefficient is taken as the base value 1. If the ratio is higher than the upper limit of the range, the correction coefficient is adjusted upward by a fixed step size. If the ratio is lower than the lower limit of the range, the correction coefficient is adjusted downward by a fixed step size. The fluctuation range of the correction coefficient is limited to the preset range. The calibrated fatigue life is obtained by multiplying the initial predicted fatigue life by the correction factor.

[0014] Preferably, the steps for outputting the fatigue resistance test results of the test sample are as follows: The termination condition for cyclic loading is that the entropy evolution of the test sample enters the entropy value slowly rising stage, and the number of completed cyclic loading cycles reaches the preset proportion of the total length of this stage calculated based on the trend of the entropy value change of the collected gray scale distribution. It is not necessary to continue loading to the entropy value steep rising stage or crack failure state. The final output of the test includes three types of data: predicted fatigue life, characteristic slope, and complete entropy value time series after calibration. All output data are stored in a one-to-one correspondence with the test temperature and deformation ratio parameters of this test.

[0015] (III) Beneficial Effects The detection area is selected as the intact rubber surface in front of the initial incision. It is not necessary to directly target the crack area to collect and identify the crack outline. It is not affected by the processing accuracy of the pre-made incision or the irregular extension shape of the crack. Only a conventional industrial camera with stable supplementary lighting is needed to complete the image acquisition. This reduces the equipment investment threshold for detection and reduces the interference of environmental factors on the detection results. Secondly, this solution does not require continuous loading until the crack of the sample extends to the failure length. It only needs to collect the grayscale data of the surface detection area in the early stage of cyclic loading, complete the entropy evolution stage determination and feature parameter extraction, and then calculate the complete fatigue life. It can effectively compress the detection time of a single sample and is suitable for rapid screening of large batches of samples.

[0016] Furthermore, by quantifying the accumulation of micro-damage inside the rubber through grayscale distribution entropy, extracting characteristic parameters by dividing the evolution stage into stable segments, and finally superimposing the weighted calibration of the proportion of stable segments, the result deviation caused by the difference in the initial state of the sample can be offset, ensuring the stability and reliability of the test results; all test data can be stored in association with the corresponding test conditions, which is convenient for subsequent result traceability and batch performance comparison. Attached Figure Description

[0017] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart illustrating the overall process of the fatigue resistance testing method for rubber sheets in this invention. Figure 2 This is a schematic diagram of the arrangement of the test sample and the surface detection area in this invention. Detailed Implementation

[0019] This application provides a fatigue resistance testing method for rubber sheets. The method characterizes the accumulation of internal micro-damage by the evolution of the gray-scale distribution entropy value of the intact area in front of the crack. It can complete the full life prediction without loading until the crack fails. At the same time, it improves the detection accuracy through multi-layer data processing optimization, shortens the detection cycle, reduces the detection cost, and ensures the reliability of the detection.

[0020] Current conventional rubber fatigue testing generally employs the flexural test method, which requires continuous loading of the specimen until the macroscopic crack propagates to the failure length before results can be obtained, resulting in a single specimen testing cycle of several days. Some improved solutions use image recognition to directly observe cracks, but still require waiting for crack initiation and are greatly affected by incision accuracy and lighting interference. To address this issue, we focus on eliminating multi-dimensional interference, quantifying micro-damage accumulation, stably extracting characteristic parameters, and calibrating and predicting the entire lifespan. Furthermore, parameters can be adjusted to adapt to different rubber materials and operating conditions according to actual testing needs. Specimen cutting and initial incision pre-fabrication are common existing technologies in rubber fatigue testing, used to create standardized stress concentration points to ensure stable crack propagation from preset locations and ensure the comparability of test results. Figure 1 As shown, the specific content is as follows: A 6mm thick nitrile rubber sheet was selected for testing. The cutting direction was consistent with the calendering direction of the rubber raw material. After calendering, the molecular chains of the rubber will be oriented along the calendering direction. The fatigue performance in the parallel orientation direction is 10%-15% higher than that in the perpendicular direction. Unifying the cutting direction can eliminate the performance dispersion caused by the orientation difference. The sheet was cut into standard sheet samples with a length of 150mm and a width of 25mm using a rubber-specific sheeting machine. The cutting edges were kept flat and burr-free to avoid stress concentration at the edges interfering with the crack propagation path. A total of 5 parallel samples were prepared in the same batch. After all samples were cut, they were placed in a standard laboratory environment of 23℃ and 50% relative humidity for 24 hours to eliminate the internal residual stress generated during the cutting process and ensure that the initial stress state of all samples was uniform.

[0021] Subsequently, an initial incision is pre-made in the middle of the long side of each sample. A single-edged blade and a special positioning fixture are used to control the incision parameters. The incision depth is 2 mm, the width is 0.5 mm, and the radius of the incision tip is controlled within 0.1 mm to ensure the consistency of stress concentration points. The extension direction of the initial incision is strictly perpendicular to the direction of the subsequent tensile loading to ensure that the stress concentration point is located at the tip of the incision during tension. The crack extends stably along the direction perpendicular to the force, avoiding deviation in the force direction caused by incision skew, and ensuring the comparability of test results for samples in the same batch.

[0022] After completing the incision preparation, the surface inspection area is delineated, such as... Figure 2 As shown, the general design rules for this area are: The length is 8-12 times the initial incision depth, and the width is 50%-70% of the sample width. This ensures complete coverage of the stress concentration zone in front of the incision (the core area where micro-damage accumulates earliest) while avoiding the introduction of noise from irrelevant areas. Starting from the tip of the initial incision, a detection area is delineated on the intact rubber surface in front of the incision along the direction of tensile force, within the range of the undamaged rubber surface. To avoid the markings themselves interfering with grayscale acquisition, extremely fine matte positioning marks are made only at the outer corners of the four diagonals of the detection area, without touching the detection area itself, providing a unified positioning reference for subsequent image acquisition. This surface detection area is used to calculate the grayscale distribution entropy through subsequent grayscale data acquisition. By characterizing the internal invisible micro-damage through the microscopic changes in the intact area, it is the core basic design for achieving early prediction before crack initiation. Among these features, constant temperature control is used to eliminate the influence of temperature on the mechanical properties of rubber; light-proof and humidity-controlled, and secondary static placement in the chamber are used to eliminate the interference of environmental factors on grayscale acquisition.

[0023] The prepared test samples are transferred to a sealed, light-proof high and low temperature test chamber for environmental pretreatment. This step meets the initial state requirements of sample preparation and aims to eliminate the interference of external factors such as temperature and light on the grayscale state of the rubber surface, providing stable baseline conditions for subsequent image acquisition. The test chamber has a built-in tensile fixture and a sealed acquisition window. The inner wall is lined with a matte black film. Closing the observation window and the light shielding plate creates a light-proof internal environment, avoiding grayscale baseline drift caused by long-term light exposure leading to photoaging or gloss changes on the rubber surface. The standard test temperature inside the chamber is set at 23°C, and the relative humidity inside the chamber is simultaneously controlled to be stable at 50%. The dimensional change rate of nitrile rubber is less than 0.1% within the humidity range of 30%-70%. 50% is the standard environmental parameter, which can reduce surface state changes caused by rubber moisture absorption or loss. The air inside the chamber is kept still without forced convection to prevent airflow from causing slight displacement of the sample or uneven surface temperature.

[0024] Based on the thermal conductivity of 6mm thick nitrile rubber, a preset settling threshold of 2 hours was set to ensure that the continuous settling time of the test sample under this temperature environment was not less than the preset threshold. After settling, the temperature deviation was confirmed to be no more than ±0.5℃ by a miniature platinum resistance temperature sensor pre-attached to the sample surface. After the overall temperature of the sample was consistent with the temperature inside the chamber, the sample was clamped in the chamber. During clamping, the center line of the sample was ensured to coincide with the force axis of the clamp, the clamping length at both ends was 25mm, and the clamping pressure was controlled at 0.3MPa. Thin rubber pads were placed at the contact points between the clamp and the sample to ensure no relative slippage during clamping and to avoid local stress concentration at the clamping position, which could cause premature sample breakage. After clamping, the sample was allowed to stand for an additional 30 minutes to eliminate the slight temperature fluctuations and contact stress caused by the clamping operation and to further stabilize the initial state of the sample. The chamber door was not opened throughout the entire pretreatment process to completely eliminate the interference of temperature fluctuations and light differences on the subsequent grayscale acquisition results from an environmental perspective.

[0025] Cyclic loading and fixed-point image acquisition: After clamping and secondary static placement, the cyclic loading and image acquisition process is started. The reciprocating stroke of the tension fixture is set to a fixed value, and the loading waveform adopts a standard sine wave to ensure that the maximum elongation of the test sample after each stretching meets the preset deformation ratio of 20%. The reciprocating frequency is set to 5Hz, which is an industry-standard parameter for fatigue testing of rubber seals. The constant parameter of the periodic reciprocating tensile load ensures that the stress state is consistent in each loading cycle, meets the constant amplitude loading requirements of fatigue testing, and ensures that the micro-damage accumulation rate is stable and controllable. The fixture drives the test sample to continuously perform reciprocating tensile motion. In each cyclic loading cycle, the test sample is uniformly stretched from a stress-free state to the maximum elongation state, and then uniformly reset to the stress-free state. The loading parameters remain constant throughout the process without dynamic adjustment.

[0026] A 5-megapixel global shutter industrial camera, paired with a ring-shaped diffused white light source, is used. The camera is fixedly mounted on a dedicated bracket directly above the sample through the transparent observation window of the test chamber. The global shutter avoids the rolling shutter effect, ensuring clear imaging at the moment of motion reset. The resolution of 5 megapixels in a 20mm×15mm detection area is approximately 17 pixels / mm, which is sufficient to capture grayscale changes caused by microscopic surface irregularities. The lens optical axis is strictly perpendicular to the surface detection area, and the working distance is kept constant at 150mm. Before acquisition, the camera's exposure time, gain, and white balance parameters are locked, and the supplementary light intensity is kept constant throughout the process. The acquisition angle is perpendicular to the surface detection area to avoid grayscale fluctuations caused by changes in illumination angle or intensity.

[0027] The acquisition trigger signal is provided by the position output module of the testing machine, with synchronization accuracy controlled at the millisecond level. The camera is triggered at the reset end of each cyclic loading cycle, at the instant when the test sample is in a zero-stress static state, to acquire clear image data of the surface detection area. Under tension, the rubber undergoes elastic deformation, and the surface grayscale will change due to the deformation and stretching. It is impossible to distinguish whether it is caused by deformation or micro-damage. Only when the deformation is restored to the zero-stress static state, the deformation is restored. At this time, the acquired grayscale change is only caused by the surface micro-unevenness caused by the accumulation of internal micro-damage. This ensures the accuracy of subsequent grayscale distribution entropy analysis from the acquisition level.

[0028] Gray-scale distribution entropy calculation and time series: rigid registration and 3σ outlier correction are existing general image processing and data processing methods, which are used to eliminate positional offset and sudden interference; gray-scale distribution entropy characterizes the accumulation of micro-damage.

[0029] After image acquisition, the background data processing program simultaneously performs analysis. This step takes the acquired raw image as input, and after multi-step preprocessing to eliminate interference, outputs stable entropy time series data, providing a reliable foundation for subsequent stage division. First, using the pixel coordinates of the initial incision tip as the reference feature point, the SIFT feature matching algorithm combined with RANSAC mismatch removal is used to perform rigid registration and coordinate alignment calibration on all acquired images. Only through rigid transformations of translation and rotation, the feature points of all images are aligned to the same coordinate position, offsetting the positional deviation of the detection area caused by slight clamping offset and slight slippage of the sample, ensuring that the surface detection area coordinates of each image coincide, and ensuring that grayscale data of different periods are comparable laterally.

[0030] After registration, a weighted grayscale algorithm is used to convert the original color image into a single-channel grayscale image. Then, the pixel grayscale values ​​in the surface detection area are divided into 16 equally spaced grayscale levels within the full grayscale range of 0 to 255, with each level corresponding to 16 consecutive grayscale values. The purpose of dividing the grayscale into levels is to reduce the interference of random noise in single pixels and amplify the overall grayscale distribution trend. In practical applications, the number of levels can be adjusted to 8 or 32 depending on the detection accuracy, with 16 levels being the preferred value that balances accuracy and computational efficiency. Subsequently, the proportion of pixels contained in each grayscale level to the total number of pixels in the surface detection area is calculated to obtain the pixel proportion data for all levels.

[0031] Next, the gray-level distribution entropy value for the corresponding cyclic loading cycle is calculated. Gray-level distribution entropy is originally a quantitative indicator that characterizes the degree of gray-level disorder in an image. It is used for the first time for rubber fatigue micro-damage detection. As cyclic loading proceeds, microcracks and micropores inside the rubber continuously germinate and accumulate, which leads to an increase in the surface micro-unevenness, an increase in the diffuse reflection ratio of light, and an increase in the disorder of gray-level distribution. The entropy value continues to grow. Therefore, the rate of change of the entropy value can directly reflect the accumulation rate of internal micro-damage.

[0032] The specific calculation method is as follows: take the common logarithm with base 10 for the pixel proportion of each gray level, multiply the obtained logarithmic value by the pixel proportion of the corresponding level, and then sum the product results of all gray levels and take the negative value, which is the gray level distribution entropy value of the current period. The reason for using the common logarithm instead of the natural logarithm is that the numerical range is more regular, and the relative change trend of the entropy value is not affected by the logarithmic base, so it does not affect the characterization of the damage rate.

[0033] It is important to note that the characteristic slope represents the rate of entropy increment and is independent of the absolute value of the initial entropy. The baseline differences caused by different initial surface roughness of different samples only affect the intercept of the fitted line, not the slope result, and will not cause detection bias. All gray-scale distribution entropy values ​​are arranged in the order of the cyclic loading cycles to form a complete entropy time series, which serves as the basis for subsequent stage division. In practical applications, if there is a lot of dust or vibration interference in the detection environment, an outlier correction step can be added after generating the entropy time series. The 3σ criterion is used to verify the time series data point by point. If the gray-scale distribution entropy value of a certain cycle deviates from the mean of the five adjacent cycles by more than three times the standard deviation, it is determined to be an abnormal jump point caused by external interference such as instantaneous vibration, dust falling, and light source flicker. The outlier value is replaced by the average entropy value of the three normal cycles before and after the outlier point. After the correction is completed, subsequent analysis is carried out.

[0034] Entropy evolution stage division: Based on the corrected entropy value time series, the feature parameter extraction stage is entered. This stage takes over the previous entropy value time series data, selects stable growth segments through stage division, and extracts core parameters that can represent the damage rate. First, the difference between the gray scale distribution entropy values ​​of two adjacent cycles is calculated in the order of the cyclic loading cycle to obtain the single-cycle entropy increment. The stage is divided by rolling judgment method. The stage status is re-verified every 5 cycles of data to avoid misjudgment caused by single data fluctuations.

[0035] The first preset threshold is set to 0.002 per cycle, and the second preset threshold is set to 0.01 per cycle. The number of consecutive judgment cycles is set to 5, meaning that 5 cycles at a 5Hz loading frequency correspond to 1 second. This filters out instantaneous random fluctuations without causing excessive lag in stage judgment. A corresponding stage can only be determined if at least a preset number of consecutive cyclic loading cycles meet the requirements of the corresponding interval. After stage judgment, the first cycle that meets the interval requirements is traced back as the starting point of the corresponding stage to ensure the accuracy of stage division. The specific judgment rules and physical meanings of the three stages are as follows: (1) Stable entropy stage: When the single-cycle entropy increment of 5 consecutive cycles is lower than the first preset threshold of 0.002 per cycle, it is determined to enter this stage, which corresponds to the initial state in which micro-damage inside the rubber has not yet been generated in large quantities. In this stage, the rubber only undergoes repeated elastic deformation, the initial internal defects expand slowly, the entropy increases very slowly and fluctuates little, usually accounting for 15%-25% of the total fatigue life.

[0036] (2) Gradual increase in entropy: When the single-cycle entropy increment of 5 consecutive cycles is between 0.002 and 0.01 per cycle, and the fluctuation range of entropy increment between adjacent cycles does not exceed 20%, it is determined to enter this stage, which corresponds to the transitional state of stable accumulation of micro-damage inside the rubber. In this stage, microcracks and micropores are uniformly germinated and expanded, and the entropy value shows a stable linear growth trend. It is the best range for extracting the damage accumulation rate, which usually accounts for 40%-50% of the total fatigue life.

[0037] (3) Entropy surge stage: When the single-cycle entropy increment of 5 consecutive cycles is higher than the second preset threshold of 0.01 per cycle, it is determined that the stage has been entered, which corresponds to the failure nearing state of rapid macro crack propagation. The starting point of this stage is the critical point for macro crack initiation. The surface gray disorder increases sharply, and the entropy growth rate increases sharply, usually accounting for 25%-35% of the total fatigue life.

[0038] The three stages are sequentially connected according to the order of the cyclic loading cycle, with no overlapping or gaps, completely corresponding to the full-cycle evolution of rubber fatigue damage from initiation and stabilization to rapid failure, which is consistent with the three-stage theory of rubber fatigue in materials science. Here, the core principle is reiterated: the increase of gray-scale distribution entropy value directly corresponds to the accumulation of internal micro-damage, and the stage division is essentially a precise identification of different stages of damage development.

[0039] Feature slope extraction: After determining that the entropy value has entered the gradual increase phase, data is continuously collected and slope stability is fitted simultaneously. When the number of collected gradual increase phase cycles reaches 60% of the estimated total length of the gradual increase phase, the feature slope extraction process is initiated. The estimated total length of the gradual increase phase is calculated as follows: based on the existing linear fitting trend of entropy increment in the gradual increase phase, it is extrapolated to the corresponding cycle point where the single-cycle entropy increment reaches the second preset threshold. This is used as the estimated endpoint of the gradual increase phase. The relative error of this extrapolation method is less than 3%, which does not affect the accuracy of the termination determination. Since the entropy value gradually increases in the gradual increase phase exhibits a stable linear growth pattern, a sufficiently long segment of data can fit a slope result consistent with the entire segment, without waiting for the phase to end.

[0040] During extraction, data from the existing gradual increase phase is extracted, and transition fluctuation data within 10% of the phase start point are removed. The remaining stable growth segment is retained for calculation. The reason for removing the initial segment is that the entropy growth rate at the phase transition node is unstable and fluctuates greatly. Directly including it in the fitting would reduce the slope accuracy. Extracting the stable segment ensures the reliability of the feature parameters. Least square linear fitting is performed on the extracted gray-scale distribution entropy data. Least square fitting is a common mathematical method, and its purpose is to obtain the optimal linear fitting result. This scheme controls the coefficient of determination of the fitting result to be no less than a preset fitting threshold of 0.95. A coefficient of determination ≥ 0.95 indicates that the linearity of the gradual increase phase exceeds 95%, indicating that the linearity of the gradual increase phase is good and conforms to the characteristics of stable accumulation of micro-damage, ensuring that the fitting accuracy meets the requirements.

[0041] An additional fallback termination rule is added: if the coefficient of determination for fitting still does not reach the 0.95 threshold when the loading reaches 80% of the estimated total length in the gradual increase phase, the data acquisition is terminated, the existing data is used to calculate the results and marked as low reliability, and a comprehensive judgment is required in conjunction with the results of parallel experiments; finally, the entropy increment value corresponding to the unit cyclic loading cycle obtained by fitting is determined as the characteristic slope. The characteristic slope, that is, the entropy increment per unit cycle, directly reflects the stable accumulation rate of micro-damage and is the core intermediate parameter connecting entropy change and fatigue life, as well as the only core input for subsequent life estimation.

[0042] After extracting the characteristic slope, the fatigue life estimation and calibration stage begins. This stage takes over the characteristic slope parameters from the previous stage, outputs the preliminary life through a pre-built model, and then corrects for sample dispersion through a stable section, ultimately obtaining a reliable test result. The characteristic slope of the test sample is substituted into the pre-established correspondence between characteristic slope and fatigue life to obtain a preliminary predicted fatigue life. This correspondence needs to be constructed in advance through standard sample tests. The construction process is as follows: using standard samples of the same material, specifications, and preparation process as the test sample, at least 15 sets of parallel standard tests are completed under consistent test temperature and deformation ratio conditions. Statistically, the power function confidence of fitting with more than 12 sets of samples can reach 95%, and 15 sets can fully guarantee reliability. Each set of standard tests is fully loaded to failure. Simultaneously, the characteristic slope is extracted using a rule of removing 10% from each end of the full gradual rise section. The total number of loading cycles when the initial notch expansion crack length reaches the critical failure length of 5mm is recorded as the standard fatigue life. The 5mm failure length is the industry-standard judgment for rubber fatigue testing. After removing outlier data with excessive dispersion, fitting multiple sets of measured data reveals a power-law positive correlation between the reciprocal of the characteristic slope and the standard fatigue life. In this embodiment, the goodness of fit reaches 0.98. An example fitting formula is T=12860×(1 / k)^{1.17}, where T is the standard fatigue life in cycles and k is the characteristic slope. The reason for using the reciprocal of the characteristic slope is that a larger slope indicates a faster rate of micro-damage accumulation, corresponding to a shorter total life, and the two are inversely proportional. The power-law fitting conforms to the empirical law of the classic SN curve of rubber fatigue life. The corresponding relationship is condition-specific. Different materials, different testing temperatures, and different deformation ratios require the construction of corresponding relationship curves. During testing, the matched curve can be directly called, and cross-condition application is not allowed to ensure the accuracy of the estimation.

[0043] After obtaining the preliminary predicted fatigue life, further calibration is performed through the weight correction of the steady segment. First, the proportion of the total number of cycles in the steady segment of entropy value to the preliminary predicted fatigue life is calculated. The length of the steady segment directly reflects the initial defect level of the sample. The fewer the initial defects of the sample, the longer the time required for micro-damage initiation, and the higher the proportion of the steady segment to the total life. Conversely, the more initial defects the sample, the lower the proportion of the steady segment. Therefore, by correcting the proportion of the steady segment, the dispersion difference of the initial state of different samples can be offset.

[0044] Both the correction coefficient and the standard range were determined using statistical results from a large sample of over 200 standard samples of the same material. The statistics showed that 90% of the qualified samples had a stable range of 15%-25%. Based on this, the correction rules were set: In this embodiment, the preset standard range is 15% to 25%, the correction step size is 0.05, and the fluctuation range of the correction coefficient is limited to 0.8 to 1.2. When the proportion is within the preset standard range, the correction coefficient takes the baseline value of 1. When the proportion is higher than the upper limit of the range, the correction coefficient is adjusted upward by a fixed step size. When the proportion is lower than the lower limit of the range, the correction coefficient is adjusted downward by a fixed step size. Setting the fluctuation range of the correction coefficient can prevent over-correction and avoid the results deviating from the actual reasonable range due to individual abnormal data.

[0045] For example, if a sample's initial predicted fatigue life is 100,000 cycles, and the stable period is 22,000 cycles (22%), which is within the standard range, the correction factor is 1, and the calibrated life remains 100,000 cycles. If the stable period is 28,000 cycles (28%), which is higher than the upper limit of the standard range, the correction factor is increased to 1.05, and the calibrated life is 105,000 cycles. If the stable period is 12,000 cycles (12%), which is lower than the lower limit of the standard range, the correction factor is decreased to 0.95, and the calibrated life is 95,000 cycles. Multiplying the initial predicted fatigue life by the correction factor yields the calibrated predicted fatigue life, which is the core test result for the sample.

[0046] Detection Termination and Result Output: Cyclic loading does not need to continue until the sample fails. After fitting the characteristic slope that meets the accuracy requirements and completing the lifetime calibration, loading and image acquisition can be stopped. The corresponding termination condition is that the entropy evolution of the test sample enters the entropy value gradual increase stage, and the number of completed cyclic loading cycles reaches 60% of the preset proportion of the total length of this stage calculated based on the entropy value change trend of the acquired grayscale distribution. There is no need to continue loading to the entropy value steep increase stage or crack failure state. Here, it echoes the design logic of the detection area again: since the entropy value steep increase stage corresponds to the initiation of macroscopic cracks, this scheme terminates the detection in the gradual increase stage. At this time, the steep increase stage has not yet been entered, and no macroscopic cracks have been generated. Therefore, the surface detection area will not be covered by crack paths throughout the entire process, which meets the setting requirements of the detection area. There is no need to worry about cracks intruding into the detection area and affecting the accuracy of the data.

[0047] Choosing a 60% ratio is an optimized value that balances detection efficiency and accuracy. When this ratio is reached, there is enough stable data in the gradual increase phase to complete high-precision feature slope fitting, and there is no need to continue loading to subsequent phases. Based on the lifetime ratio of the three phases, it is estimated that only about 50% of the total lifetime loading cycle needs to be completed to obtain the result.

[0048] The final output of the test includes three types of data: calibrated predicted fatigue life, characteristic slope, and complete entropy time series. The calibrated predicted fatigue life is the core test result, used to determine whether the sample meets the performance requirements. The characteristic slope, i.e., the entropy increment per unit period, represents the micro-damage accumulation rate and is used to compare the consistency of damage rates among samples in the same batch. The complete entropy time series is used for subsequent data tracing and secondary analysis. All output data are stored one-to-one with the test temperature, deformation ratio, and sample number parameters of this test, facilitating subsequent result tracing and iterative optimization of the test model. For multiple parallel samples in the same batch, the final batch test result is the arithmetic mean of the calibrated predicted fatigue life of all parallel samples. The single sample result and the batch dispersion coefficient are output simultaneously. If the dispersion coefficient exceeds 10%, the number of parallel samples is automatically increased to ensure the reliability of the results. This batch processing rule can be flexibly adjusted according to the actual sampling requirements.

[0049] When the testing scenario is adjusted to high or low temperature environment or fatigue testing of rubber sheets of other materials, only the constant temperature chamber temperature, preset static threshold, loading deformation ratio and corresponding characteristic slope-life relationship curve need to be adjusted accordingly. The rest of the testing process remains the same, which can adapt to the testing needs of different materials and different working conditions. The solution has strong versatility.

[0050] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method of detecting fatigue resistance of a rubber sheet, characterized by, The detection method includes the following steps: Prepare a test sample with an initial incision, and delineate a surface detection area of ​​a fixed size on the intact rubber surface in front of the incision; Place the test sample in a sealed, light-proof, constant-temperature environment and let it stand until the overall temperature of the test sample is uniform and stable. A periodic reciprocating tensile load with constant parameters is applied to the test sample, and image data of the surface detection area is acquired at the zero-stress reset node of each cyclic loading cycle. For each image, perform gray-level segmentation and pixel proportion statistics on the surface detection area, calculate the gray-level distribution entropy value of the corresponding loading cycle, and generate an entropy value time sequence by arranging them in the loading order; Based on the periodic incremental change pattern of entropy time series, three consecutive entropy evolution stages are divided, and the linear slope of the gray-scale distribution entropy change within a specified evolution stage is extracted as the feature slope. By establishing a pre-defined relationship between characteristic slope and fatigue life, and combining it with the weight correction of the stationary segment, the fatigue life of the test sample after calibration is estimated. Output the fatigue resistance test results of the sample to be tested.

2. The method of claim 1, wherein The specific steps for preparing the test sample with an initial incision and delineating a surface inspection area of ​​fixed size on the intact rubber surface in front of the incision are as follows: The rubber sheet to be tested is cut into sheet-like samples with length, width, and thickness all conforming to the preset dimensions; An initial cut with a depth and width that meets the set values ​​is pre-made in the middle of one edge of the sheet-like specimen, and the extension direction of the initial cut is perpendicular to the force direction of the subsequent tensile loading. Starting from the tip of the initial incision, a rectangular area of ​​fixed length and width is delineated on the intact rubber surface in front of the incision along the direction of tensile force as a surface inspection area. The surface inspection area does not cover the initial incision body or the subsequent crack path.

3. The method of claim 1, wherein The steps for placing the test sample in a sealed, light-proof, constant-temperature environment until the overall temperature of the test sample is uniform and stable are as follows: Place the sample to be tested horizontally in a sealed, light-proof, constant-temperature chamber, set the internal temperature of the chamber to the preset detection temperature, and keep the air inside the chamber still. The continuous static time of the test sample under this temperature environment shall not be less than the preset static threshold. Once the overall temperature of the test sample is consistent with the internal temperature of the cavity, environmental pretreatment is completed.

4. The fatigue resistance testing method for rubber sheets according to claim 1, characterized in that, The steps for applying a constant cyclic tensile load to the test sample and acquiring image data of the surface detection area at the zero-stress reset node of each cyclic loading cycle are as follows: The two ends of the pre-treated test sample are fixed to the two clamping ends of the tensile fixture. The reciprocating stroke of the tensile fixture is set to a fixed value so that the maximum elongation of the test sample after each stretching meets the preset deformation ratio. The reciprocating motion frequency of the tension fixture is set to a fixed value, which drives the test sample to continuously perform reciprocating tension motion. In each cycle of loading, the test sample is stretched from a stress-free state to the maximum elongation state, and then reset to a stress-free state. Image acquisition is triggered at the reset endpoint of each cyclic loading cycle, the instant the test sample is in a zero-stress static state. During the acquisition process, the supplementary light intensity remains constant, and the acquisition angle is perpendicular to the surface detection area to obtain distortion-free and clear image data of the surface detection area.

5. The fatigue resistance testing method for rubber sheets according to claim 1, characterized in that, The steps for performing grayscale level division and pixel proportion statistics on the surface detection area of ​​each image, calculating the grayscale distribution entropy value for the corresponding loading cycle, and arranging the entropy value time series according to the loading order are as follows: Using the pixel coordinates of the initial incision tip as a reference, rigid registration and coordinate alignment calibration are performed on all acquired images to make the surface detection area coordinates of each image coincide. The pixel grayscale values ​​in the surface detection area are divided into a preset number of equally spaced grayscale levels within the full grayscale range of 0 to 255, and the proportion of the number of pixels contained in each grayscale level to the total number of pixels in the surface detection area is calculated. Take the common logarithm to the base 10 for the pixel proportion of each gray level, multiply the resulting logarithmic value by the pixel proportion of the corresponding level, sum the product of all gray levels and take the negative value to obtain the gray level distribution entropy value of the corresponding loading cycle. Arrange all grayscale distribution entropy values ​​in the order of the cyclic loading cycles to generate a complete entropy value time series.

6. The fatigue resistance testing method for rubber sheets according to claim 5, characterized in that, Based on the periodic incremental change pattern of entropy time series, the steps to divide the entropy evolution into three consecutive stages are as follows: The difference between the grayscale distribution entropy values ​​of two adjacent cycles is calculated sequentially according to the cyclic loading cycle order to obtain the single-cycle entropy increment; Specifically, the interval where the single-cycle entropy increment is consistently lower than the first preset threshold and for no less than a preset number of cycles is defined as the entropy stable stage, corresponding to the initial state where micro-damage inside the rubber has not yet begun to emerge in large quantities; the continuous interval where the single-cycle entropy increment is between the first preset threshold and the second preset threshold and the fluctuation amplitude between adjacent cycles does not exceed the limit value is defined as the entropy slowly rising stage, corresponding to the transitional state where micro-damage inside the rubber is accumulating stably; and the interval where the single-cycle entropy increment is consistently higher than the second preset threshold and for no less than a preset number of cycles is defined as the entropy sharply rising stage, corresponding to the state of near failure where macroscopic cracks are rapidly propagating. The three stages are connected sequentially according to the order of the cyclic loading cycle, with no overlapping or gaps.

7. The fatigue resistance testing method for rubber sheets according to claim 6, characterized in that, The steps to extract the linear slope of the gray-level distribution entropy change within a specified evolutionary stage as the feature slope are as follows: Extract the middle section of data during the entropy increase phase, and extract the starting and ending points from the start and end points of the phase by a preset proportional length, removing the fluctuating data at the phase transition nodes. Perform least-squares linear fitting on the extracted grayscale distribution entropy data, and control the determination coefficient of the fitting result to be no less than the preset fitting threshold. The entropy increment value corresponding to the unit cyclic loading cycle obtained from the fitting is used as the feature slope.

8. The fatigue resistance testing method for rubber sheets according to claim 7, characterized in that, The steps for constructing the pre-established relationship between characteristic slope and fatigue life are as follows: Using standard specimens of the same material, specifications, and preparation process as the test specimen, multiple sets of complete fatigue tests were conducted under consistent test temperature and deformation ratio conditions. The characteristic slope and corresponding standard fatigue life of each set of standard specimens were recorded. The standard fatigue life is the total number of loading cycles when the standard specimen is loaded until the crack length reaches the preset critical failure length. By fitting multiple sets of measured data, a positive correlation between the reciprocal of the characteristic slope and the standard fatigue life was obtained as a power function relationship, which is used as the correspondence between the characteristic slope and the fatigue life. By substituting the characteristic slope of the test sample into the corresponding relationship, a preliminary prediction of fatigue life can be obtained.

9. The fatigue resistance testing method for rubber sheets according to claim 8, characterized in that, The steps for calculating the predicted fatigue life of the test sample after calibration, based on the weight correction of the stationary phase, are as follows: The proportion of the total number of cycles in the steady-state phase of entropy to the preliminary predicted fatigue life is calculated. The correction coefficient is matched according to the range in which the ratio is located. If the ratio is in the preset standard range, the correction coefficient is taken as the base value 1. If the ratio is higher than the upper limit of the range, the correction coefficient is adjusted upward by a fixed step size. If the ratio is lower than the lower limit of the range, the correction coefficient is adjusted downward by a fixed step size. The fluctuation range of the correction coefficient is limited to the preset range. The calibrated fatigue life is obtained by multiplying the initial predicted fatigue life by the correction factor.

10. The fatigue resistance testing method for rubber sheets according to claim 9, characterized in that, The steps for outputting the fatigue resistance test results of the test sample are as follows: The termination condition for cyclic loading is that the entropy evolution of the test sample enters the entropy value slowly rising stage, and the number of completed cyclic loading cycles reaches the preset proportion of the total length of this stage calculated based on the trend of the entropy value change of the collected gray scale distribution. It is not necessary to continue loading to the entropy value steep rising stage or crack failure state. The final output of the test includes three types of data: predicted fatigue life, characteristic slope, and complete entropy value time series after calibration. All output data are stored in a one-to-one correspondence with the test temperature and deformation ratio parameters of this test.

Images