Caterpillar track section quality detection method and system based on laser measurement

Abstract

The invention relates to the technical field of caterpillar track section quality detection, in particular to a caterpillar track section quality detection method and system based on laser measurement. The method comprises the following steps: carrying out data collection on the surface morphology of a track link section through laser measurement equipment, and identifying and processing crack feature distribution, so as to generate aperture crack feature distribution mapping data. On the basis of the data, driving wheel meshing power loss simulation is carried out, power loss data are obtained, and the eccentric wear fracture risk probability is calculated according to the data. And then, analyzing the eccentric wear fracture risk probability data by using a strategy gradient algorithm, constructing a caterpillar track section quality detection model, and finally, feeding back the model to a terminal to realize caterpillar track section quality detection. According to the method, the caterpillar track link quality detection technology is optimized, so that the caterpillar track link quality detection technology is more perfect.

Description

Technical Field

[0001] The present invention relates to the technical field of track segment quality detection, and in particular to a track segment quality detection method and system based on laser measurement. Background Art

[0002] Laser measurement technology offers the advantages of high precision, high resolution, and non-contact performance. It can accurately capture surface morphology data for track links, providing a reliable data foundation for subsequent crack identification and defect analysis. Laser measurement equipment efficiently captures surface morphology data for track links. Advanced data processing algorithms can identify minute surface cracks and machining defects, accurately assessing machining quality. Track link quality inspection is not limited to detecting surface defects; it also requires comprehensive consideration of the track link's performance under actual operating conditions. For example, surface cracks in a track link can directly affect the loss of drive wheel meshing power, thereby impacting the overall operational efficiency and safety of rail transit. Therefore, track link quality inspection methods based on laser measurement data need to be further integrated with dynamic simulation, crack growth analysis, and risk assessment to establish a comprehensive quality inspection system. By accurately identifying the characteristics of machining cracks in the track link and simulating drive wheel meshing loss, the risks of the track link during actual use can be assessed. However, traditional track link quality inspection methods based on laser measurement suffer from inaccurate crack growth analysis and fracture risk assessment, resulting in large quality inspection errors. Summary of the Invention

[0003] Based on this, it is necessary to provide a track link quality detection method and system based on laser measurement to solve at least one of the above technical problems.

[0004] To achieve the above object, a method for detecting the quality of a track link based on laser measurement is provided, the method comprising the following steps: Step S1: Collecting factory surface morphology laser measurement data of the track segment using a laser measurement device to obtain track segment surface morphology laser measurement data; performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; performing aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; Step S2: simulating the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain drive wheel meshing power loss data; performing eccentric wear fracture risk probability calculation based on the drive wheel meshing power loss data to obtain eccentric wear fracture risk probability data; Step S3: constructing a track segment quality detection model based on the eccentric wear fracture risk probability data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

[0005] Preferably, step S1 includes the following steps: Step S11: Collecting factory surface morphology laser measurement data of the track link using a laser measurement device to obtain the surface morphology laser measurement data of the track link; Step S12: extracting the track segment aperture state from the track segment surface morphology laser measurement data to obtain the track segment aperture state data; Step S13: performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; Step S14: performing aperture crack distribution mapping on the track segment aperture state data based on the track segment machining crack characteristic distribution data to obtain aperture crack characteristic distribution mapping data.

[0006] Preferably, step S13 includes the following steps: Step S131: performing machining crack roughness distribution identification on the track link surface morphology laser measurement data to obtain machining crack roughness distribution data; Step S132: performing self-similarity fractal feature analysis on the processing crack roughness distribution data to obtain crack roughness distribution fractal feature data; Step S133: Calculating the variance of the texture surface curvature radius based on the crack roughness distribution fractal feature data to obtain the variance of the crack texture surface fractal curvature radius; Step S134: performing crack kink disorder state identification on the crack roughness distribution fractal feature data according to the variance of the crack surface fractal curvature radius to obtain crack kink disorder state data; Step S135: performing machining crack characteristic distribution identification based on the crack surface fractal curvature radius variance and crack kink disorder state data to obtain the track link machining crack characteristic distribution data.

[0007] Preferably, step S134 includes the following steps: Performing equal-partition curvature profile projection processing on the variance data of the fractal curvature radius of the crack surface to obtain the equal-partition curvature profile data of the crack surface; Based on the crack surface equal-area curvature profile data, the crack roughness distribution fractal feature data is used to calculate the curvature change trend differential gradient between adjacent crack sections, and the curvature change trend differential gradient between adjacent sections is obtained; The random process of crack kink angle expansion is analyzed by using the differential gradient of curvature variation trend to obtain the random process data of kink angle expansion. The crack kink disorder state is identified based on the curvature change trend differential gradient and kink angle expansion random process data to obtain the crack kink disorder state data.

[0008] Preferably, step S2 includes the following steps: Step S21: performing crack trend topology analysis on the aperture crack characteristic distribution mapping data to obtain aperture crack characteristic trend topology data; Step S22: simulating the drive meshing power loss of the track link based on the aperture crack characteristic trend topology data to obtain the drive wheel meshing power loss data; Step S23: predicting the crack strike depth / width expansion ratio based on the aperture crack characteristic distribution mapping data according to the driving wheel meshing power loss data to obtain the crack strike depth / width expansion ratio; Step S24: Calculate the eccentric wear fracture risk probability based on the driving wheel meshing power loss data and the crack depth / width expansion ratio to obtain eccentric wear fracture risk probability data.

[0009] Preferably, step S22 includes the following steps: Step S221: performing spatial intersection density analysis on the aperture crack feature trend topology data to obtain the crack spatial intersection density; Step S222: quantifying the aperture load pressure angle offset of the aperture crack characteristic trend topology data based on the crack space intersection density to obtain aperture load pressure angle offset data; Step S223: performing tooth profile force deviation calculation on the aperture load pressure angle offset data to obtain load tooth profile force deviation data; Step S224: performing power transmission intermittent fluctuation simulation based on the load tooth profile force deviation data and the aperture load pressure angle offset data to obtain power transmission intermittent fluctuation data; Step S225: Simulating the drive meshing power loss of the track link according to the load tooth profile force deviation data, the aperture load pressure angle offset data and the power transmission intermittent fluctuation data to obtain the drive wheel meshing power loss data.

[0010] Preferably, step S24 includes the following steps: Step S241: identifying a power transmission deviation concentration area based on the driving wheel meshing power loss data to obtain the power transmission deviation concentration area; Step S242: Deducing the stress increment index of the concentrated area of power transmission deviation based on the driving wheel meshing power loss data to obtain the stress increment index of the concentrated area; Step S243: performing a crack propagation stress critical inference based on the concentrated area stress increment index and the crack strike depth / width expansion ratio to obtain crack propagation stress critical data; Step S244: performing eccentric wear fracture risk probability calculation on the critical crack propagation stress data to obtain eccentric wear fracture risk probability data.

[0011] Preferably, step S3 includes the following steps: Step S31: normalizing the eccentric wear fracture risk probability data to obtain normalized eccentric wear fracture risk probability data; Step S32: performing risk probability association learning based on the normalized data of the eccentric wear fracture risk probability to obtain risk probability association learning data; Step S33: performing a multiple regression analysis on the risk probability association learning data to obtain risk probability association regression data; Step S34: constructing a track segment quality detection model for the risk probability associated regression data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

[0012] Preferably, the present invention further provides a track segment quality detection system based on laser measurement, which is used to perform the track segment quality detection method based on laser measurement as described above. The track segment quality detection system based on laser measurement includes: The aperture crack distribution mapping module is used to collect factory surface morphology laser measurement data of the track segment using laser measurement equipment to obtain track segment surface morphology laser measurement data; perform machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; perform aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; The fracture risk probability calculation module is used to simulate the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain the drive wheel meshing power loss data; and to calculate the eccentric wear fracture risk probability based on the drive wheel meshing power loss data to obtain the eccentric wear fracture risk probability data; The quality detection model construction module is used to construct a track segment quality detection model based on the eccentric wear and fracture risk probability data according to the policy gradient algorithm, obtain the track segment quality detection model, and feed back the track segment quality detection model to the terminal to perform track segment quality detection.

[0013] The beneficial effect of the present invention is that by accurately measuring the surface morphology of the track link using laser measurement equipment, the surface data of the track link can be accurately obtained. This data not only reflects the geometric morphology of the track link, but also provides a basis for subsequent crack identification. Processing crack feature distribution identification can effectively identify small cracks and defects on the surface of the track link, and further aperture crack distribution mapping based on these feature data can accurately determine the distribution and severity of the cracks. The core advantage of this step is the high-precision, non-contact measurement method, which avoids the errors or damage caused by traditional contact detection, while improving the accuracy and detection efficiency of crack identification. Based on the aperture crack feature distribution mapping data, the track link drive wheel engagement power loss simulation is carried out, which can deeply analyze the dynamic performance of the track link under actual working conditions. By simulating the drive wheel engagement loss, the impact of cracks on power transmission can be evaluated, thereby more comprehensively understanding the potential threat of cracks to the operation of the rail transit system. Further eccentric wear fracture risk probability calculation can quantify the risk of track links in long-term use, thereby providing a scientific basis for risk warning and maintenance decision-making. The advantage of this step is that it combines crack analysis with actual working condition simulation, improving the accuracy and predictive ability of the track segment quality assessment. The policy gradient algorithm is used to analyze the probability data of eccentric wear and fracture risk and construct a track segment quality detection model. This process can automatically establish an accurate quality detection model based on historical data and simulation results, greatly improving the intelligence and automation level of detection. By feeding back to the terminal for track segment quality detection, quality monitoring without human intervention can be achieved throughout the process, which not only improves the detection efficiency but also ensures the high accuracy of the detection results. In addition, the application of the policy gradient algorithm can dynamically adjust the model according to real-time data to adapt to changes in different environments, ensuring that the model has good adaptability and stability, and further improving the reliability and practicality of the track segment quality monitoring system. Therefore, the present invention is an optimization of a traditional track segment quality detection method based on laser measurement, which solves the problem that a traditional track segment quality detection method based on laser measurement has inaccurate crack propagation analysis and fracture risk assessment, thereby causing large quality detection errors, improves the accuracy of crack propagation analysis and fracture risk assessment, and reduces the error of quality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The figure is a flow chart of the steps of a method for detecting the quality of a track link based on laser measurement; Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG. Figure 3 for Figure 2 Detailed implementation steps of step S22 are shown in the flowchart. DETAILED DESCRIPTION

[0015] See also Figures 1 to 3 , a method for detecting the quality of a track link based on laser measurement, the method comprising the following steps: Step S1: Collecting factory surface morphology laser measurement data of the track segment using a laser measurement device to obtain track segment surface morphology laser measurement data; performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; performing aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; Step S2: simulating the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain drive wheel meshing power loss data; performing eccentric wear fracture risk probability calculation based on the drive wheel meshing power loss data to obtain eccentric wear fracture risk probability data; Step S3: constructing a track segment quality detection model based on the eccentric wear fracture risk probability data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

[0016] In the embodiment of the present invention, reference Figure 1 The above is a schematic flow chart of the steps of a track segment quality detection method based on laser measurement of the present invention. In this example, the track segment quality detection method based on laser measurement includes the following steps: Step S1: Collecting factory surface morphology laser measurement data of the track segment using a laser measurement device to obtain track segment surface morphology laser measurement data; performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; performing aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; In an embodiment of the present invention, high-precision laser scanning equipment is used to collect laser measurement data of the track segment's surface morphology at the factory. During the measurement process, the laser scanning equipment is mounted on a three-dimensional motion platform, with a scanning speed of 2500 points / second, a scanning resolution of 0.05mm, and a laser wavelength of 635nm. The scanning angle covers the entire surface area of the track segment, including the tooth surface, the hole perimeter, and the link joint. After acquisition, the point cloud data is processed into a three-dimensional grid using a spatial reconstruction algorithm to remove obstructing or reflective interference points. Median filtering and B-spline fitting are then used to repair defective areas in the point cloud, resulting in a complete and continuous track segment surface morphology data model. This data model is uniformly calibrated within the coordinate system, ensuring that each track segment sample has a consistent geometric reference frame. During the identification of machining crack feature distribution, the processed three-dimensional track segment surface morphology data is used using a roughness gradient recognition algorithm to initially identify suspected crack areas. The algorithm calculates the average gradient and standard deviation of surface elevation changes within each local grid region. Any change in a region exceeding a set threshold (e.g., average gradient exceeding 0.15 mm / mm, standard deviation greater than 0.1 mm) is marked as a suspected crack point. Subsequently, a fractal dimension recognition method is applied, using a box-counting binning algorithm to analyze the spatial distribution of suspected crack points. The fractal dimension of each region is extracted, and the fractal dimension range is limited to 2.2 to 2.6 to identify machining crack regions. Furthermore, a Gabor texture direction extraction method is used to obtain the crack extension direction. The Gabor filter direction angle is set from 0° to 180°, with every 15° grouping, for a total of 12 groups of directions. Convolution calculations are performed on the crack region in sequence, and the main crack direction is determined by the maximum directional response, forming a crack feature distribution dataset. In the aperture crack distribution mapping step, spatial coordinate registration technology is used to spatially fuse the machining crack feature distribution data with the track link aperture position. First, all structural holes in the track link are identified. Circular areas with a diameter greater than 6 mm and an edge integrity greater than 95% in the measured model are used as the identification criteria for structural holes. The 3D coordinates of the crack region are matched to the aperture region using the ICP (Iterative Closest Point) algorithm. Aperture cracks are identified when the angle between the main crack direction and the tangent direction of the aperture boundary is less than 30°, and the crack extension region crosses more than 30% of the aperture boundary circumference. Finally, the matched crack features are normalized and spatially mapped, outputting the aperture crack feature distribution map data.

[0017] Step S2: simulating the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain drive wheel meshing power loss data; performing eccentric wear fracture risk probability calculation based on the drive wheel meshing power loss data to obtain eccentric wear fracture risk probability data; In an embodiment of the present invention, during the simulation of drive wheel meshing power loss, a finite element static analysis method was used to construct a meshing contact model between the track link and the drive wheel based on the aperture crack characteristic distribution mapping data. The material parameters were set to 45# steel, Young's modulus to 210GPa, and Poisson's ratio to 0.3. The model meshing contact load was set to 1500N, and the loading direction was along the tangential direction of the drive wheel rotation. The crack area was assigned a stress concentration factor K = 2.5 to 5 based on the mapping data. The crack location was enhanced by mesh refinement to ensure that the unit size of the crack area was controlled within 0.05mm and the remaining area was 0.3mm. During the analysis process, the contact surface load transfer path was recorded, and the force transfer efficiency of the meshing node was compared under the conditions of crack presence and absence. The overall power loss rate was calculated by extracting the load offset, contact area change, and edge meshing number. In the calculation of the probability of eccentric wear fracture risk, the above-mentioned meshing loss data was collected and the crack distribution parameters, including crack depth, width, direction, and length, were recorded. A stress-displacement response curve was constructed for each crack. Crack propagation was simulated using the Monte Carlo method, with initial microcrack propagation parameters randomly generated each time. The crack growth trend during loading was evaluated based on the Paris law framework. The number of iterations was set to 5000, and the number of cycles from propagation to fracture was recorded each time. The probability of eccentric wear fracture risk was calculated by counting the number of times the track link reached failure within the simulation cycle. This probability was output as a specific value and assigned to the corresponding data label for the track link.

[0018] Step S3: constructing a track segment quality detection model based on the eccentric wear fracture risk probability data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

[0019] In an embodiment of the present invention, during the stage of constructing a quality detection model for track links, the data on the probability of eccentric wear and fracture risk are first normalized to the minimum and maximum values and mapped to the range of 0-1. The principal component analysis (PCA) dimensionality reduction method is then used to process multiple influencing variables such as crack direction, length, number, and meshing loss rate, and the first three principal components are extracted as input variables. Combined with the policy gradient algorithm, a quality classification discriminator is constructed in a reinforcement learning framework, and the objective function is set to maximize the accuracy of identifying high-risk track links. The discrimination output is divided into three categories: high risk (P ≥ 0.7), medium risk (0.4 ≤ P < 0.7), and low risk (P < 0.4). After the model training is completed, it is exported as an executable structured recognition rule, and the model is deployed to the edge detection terminal to automatically identify and detect the quality level of the track links during the track link warehousing process.

[0020] Step S1 includes the following steps: Step S11: Collecting factory surface morphology laser measurement data of the track link using a laser measurement device to obtain the surface morphology laser measurement data of the track link; Step S12: extracting the track segment aperture state from the track segment surface morphology laser measurement data to obtain the track segment aperture state data; Step S13: performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; Step S14: performing aperture crack distribution mapping on the track segment aperture state data based on the track segment machining crack characteristic distribution data to obtain aperture crack characteristic distribution mapping data.

[0021] In this embodiment of the present invention, the track segment is placed on a 3D laser scanning platform before shipment, and laser measurement equipment performs non-contact laser scanning on it. The laser emission frequency of the equipment is set to 2500 Hz, the laser wavelength is 635 nm, the vertical scanning range is 270°, and the horizontal resolution is 0.05 mm. The measurement platform uses a high-precision displacement slide to move the track segment through the scanning area at a constant speed of 0.2 m / s, ensuring that the scan covers all structural surfaces of the track segment, including the tooth root, pitch surface, joint hole edge, and outer surface. The raw point cloud data is subjected to noise reduction processing using an edge filtering module embedded in the laser measurement system. A third-order wavelet filter is used to smooth the point cloud data edges, remove flying points and background interference points, and reconstruct it into a 3D dot matrix image format with uniform density. The point cloud data is calibrated along the coordinate axes, setting the bottom surface of the track segment as the reference plane. The X-axis is defined as the length direction, the Y-axis as the width direction, and the Z-axis as the height direction. This completes the collection and standardization of the laser measurement data of the track segment surface morphology. Aperture state extraction is performed on this standardized 3D point cloud data. First, a complete triangular mesh model was constructed using a mesh reconstruction algorithm based on normal vector consistency. The normal vector direction was extracted within each mesh face, and a set of continuous edge points along the hole boundary was identified. Subsequently, the Hough transform method was used for geometric identification. The detection circle parameters were set in three-dimensional space with a diameter range of 6 mm to 20 mm. The scanning angle was rotated in steps of 1°, and multiple circles were fitted to the closed edge points. The RANSAC algorithm was used to optimize the fitting accuracy and remove outliers. When the fitting residual of the fitted circular area was less than 0.1 mm and the edge point closure rate was greater than 95%, it was determined to be a track link structural hole. For each identified aperture area, its spatial position, aperture size, center coordinates, and edge normal vector change rate were recorded to generate aperture status data, including parameters such as position identification, orientation angle, roundness deviation, and eccentricity. Laser measurement data of the track link surface morphology was used to identify the distribution of machining crack characteristics. The process first calculates the local normal gradient of the point cloud data, dividing the entire point cloud into 2mm×2mm grid cells. The variance of the point cloud normal vector is calculated within each cell. If the normal variance exceeds a set threshold (e.g., 0.08 rad²), the area is marked as a rough anomaly. The SIFT three-dimensional feature extraction algorithm is further applied to these areas to analyze the local structural response of each anomaly and extract stable extreme points in scale space. Crack regions are represented as linear clusters with high normal variation rates and severe local texture heterogeneity. Canny edge detection is then used to identify connected high-density extreme value chains on the spatial grid. A Gabor filter is then used to calculate the texture response in 12 different directions to identify the crack direction and width. Crack length, strike angle, average width, and edge grayscale difference are used as feature vectors to construct a processing crack feature distribution dataset and uniformly encode and manage them.The machining crack feature distribution data generated in step S13 is used with the aperture state data of step S12 to map the aperture crack distribution. The mapping process adopts a spatial registration method. First, the center point coordinates of the crack main axis and the aperture center coordinates are matched with the spatial nearest neighbor through the KD-Tree index method, and the crack segments with a distance less than 2mm and a direction angle less than 30° are pre-matched with the aperture boundary. Then, the tangent intersection point of the crack main axis passing through the aperture boundary is calculated. If the crack extension length exceeds 25% of the aperture boundary perimeter and the projection value of the crack width in the normal direction of the hole edge is greater than 0.2mm, the crack is marked as an aperture crack, and its spatial features are bound to the aperture identifier. Finally, an aperture crack feature distribution mapping data file containing parameters such as aperture number, crack number, crack penetration angle, crack length as a percentage of aperture perimeter, crack direction and hole edge angle is generated and output to the subsequent meshing loss simulation processing module in a spatial indexing manner.

[0022] Step S13 includes the following steps: Step S131: performing machining crack roughness distribution identification on the track link surface morphology laser measurement data to obtain machining crack roughness distribution data; Step S132: performing self-similarity fractal feature analysis on the processing crack roughness distribution data to obtain crack roughness distribution fractal feature data; Step S133: Calculating the variance of the texture surface curvature radius based on the crack roughness distribution fractal feature data to obtain the variance of the crack texture surface fractal curvature radius; Step S134: performing crack kink disorder state identification on the crack roughness distribution fractal feature data according to the variance of the crack surface fractal curvature radius to obtain crack kink disorder state data; Step S135: performing machining crack characteristic distribution identification based on the crack surface fractal curvature radius variance and crack kink disorder state data to obtain the track link machining crack characteristic distribution data.

[0023] In an embodiment of the present invention, the roughness distribution of machining cracks is identified using the laser measurement point cloud data of the track link surface morphology. First, the three-dimensional point cloud data is projected into a local Gaussian surface coordinate system, and the surface is evenly divided, with each crack detection window set to a fixed size of 5mm×5mm. Within each window area, a local extreme value search algorithm is used to extract the peak and valley points of the surface, and then a spatial gradient operator (such as a three-dimensional Sobel operator) is used to calculate the surface height gradient. The roughness parameters use the arithmetic mean height Ra and the peak-to-valley height difference Rz as evaluation criteria, where Ra is calculated by averaging the absolute values of the local height deviations from the mean along the Z-axis, and Rz is calculated by calculating the difference between the maximum peak height and the deepest valley value in each window. When the Ra value in a certain area exceeds 12μm and the Rz value is greater than 50μm, the area is identified as a rough abnormality area with cracks. A moving window overlap method was used to completely scan all point cloud areas, ultimately generating a crack roughness distribution image of the track link surface. The spatial coordinates, Ra values, Rz values, maximum profile depth, and average height variance of all crack regions were archived and output to construct a machined crack roughness distribution dataset. The self-similarity fractal characteristics of the crack roughness distribution data were analyzed using fractal geometry analysis. Specifically, the box-counting dimension (BCD) of the crack region surface was first calculated. Using the center point of the crack region as a reference, a grid of equally spaced cubes was constructed in three-dimensional space, with the cube side length increasing from 0.1 mm to 1.5 mm. The number of crack surface points contained within each cube size was statistically analyzed, and the slope of the logarithmic coordinates was calculated, representing the BCD of the crack surface. The D value reflects the self-similarity complexity of the crack surface; a closer D value to 2.0 indicates a rougher surface. This process combines multi-scale and multi-directional wavelet transforms to perform a supplementary analysis of the roughness of the crack surface in all directions. High-frequency fluctuation amplitudes at different scales within the crack region are extracted and combined with the D value to form a fractal feature vector for the crack roughness distribution. This ultimately generates a fractal feature dataset consisting of the crack number, crack region location, box dimension D, and a scale response amplitude sequence. This fractal feature data is then used to calculate the variance of the texture surface curvature radius. First, each crack region is smoothed using a quadratic surface fitting model based on surface fitting techniques to obtain the principal and minor curvatures of the crack surface. At each crack point, the principal curvature radius and its variation along the crack axis are calculated. The curvature radius values are sampled at 0.1 mm intervals along the crack principal axis and recorded to form a one-dimensional curvature sequence. The variance of this curvature sequence is then calculated as the variance of the texture surface curvature radius within the crack region, reflecting the degree of unevenness of the crack surface. If the curvature variance of a crack area exceeds 0.03mm², it means that the surface of the area is volatile and the shape is unstable.The output of this step is a dataset of the variance of the crack surface fractal curvature radius, including the crack number, principal axis length, average curvature radius, variance, and sampling sequence. The kink disorder state of the crack is identified based on the variance of the crack surface fractal curvature radius. Using the crack principal axis fitting method, principal component analysis (PCA) is performed on all points in the crack region. After determining the principal axis direction, the local tangent direction variation is calculated for each 1mm segment. A kink event is considered if the angle between two adjacent tangent directions exceeds 15°. The spatial discontinuity of the kink region is determined by combining the curvature variance data. A threshold is set: when the number of kink events exceeds 30% of the total number of principal axis segments and the average angle variation exceeds 20°, a highly disordered kink crack is defined. The spectral density distribution of the crack contour is further analyzed using Fourier transform to determine whether its frequency distribution exhibits a non-periodic high-frequency clustering. The crack kink disorder state vector is constructed by combining indicators such as kink event density, mean angle variation, and spectral dominant frequency. The output data includes crack number, number of kink segments, angle standard deviation, spectrum dominant frequency and disorder judgment label. The variance of the crack surface fractal curvature radius and the crack kink disorder state data are combined to comprehensively identify the distribution of processing crack characteristics. For each crack area, the feature fusion analysis method is used to construct a crack feature vector space composed of multi-dimensional vectors such as box dimension D, Ra, Rz, curvature variance, kink angle standard deviation, spectrum dominant frequency, etc. The crack feature vectors are classified and clustered using an improved clustering algorithm based on K-means++, and the cracks are divided into three categories: process residual cracks, fatigue cracks, and thermal cracks. At the same time, the spatial position, total crack length, average width, number of crack-intensive areas and adjacent aperture distance of each type of crack are marked. Finally, the chain track segment processing crack characteristic distribution data is generated and stored in the data management platform as a standard structured data set, and used for simulation analysis of the impact of cracks on dynamics in subsequent steps.

[0024] Step S134 includes the following steps: Performing equal-partition curvature profile projection processing on the variance data of the fractal curvature radius of the crack surface to obtain the equal-partition curvature profile data of the crack surface; Based on the crack surface equal-area curvature profile data, the crack roughness distribution fractal feature data is used to calculate the curvature change trend differential gradient between adjacent crack sections, and the curvature change trend differential gradient between adjacent sections is obtained; The random process of crack kink angle expansion is analyzed by using the differential gradient of curvature variation trend to obtain the random process data of kink angle expansion. The crack kink disorder state is identified based on the curvature change trend differential gradient and kink angle expansion random process data to obtain the crack kink disorder state data.

[0025] In an embodiment of the present invention, during the equally partitioned curvature profile projection process of the crack surface fractal curvature radius variance data, the crack principal axis is first used as the reference line. Sectioning planes are divided along the principal axis at a fixed interval of 0.2 mm. The sectioning operation is performed using planes orthogonal to the principal axis. Each sectioning plane intersects with the crack's three-dimensional point cloud data. All intersection points on the section are extracted and curvature fitting is performed. The crack contour on the section is reconstructed using cubic spline interpolation. On each section contour line, the contour curvature is calculated using the three-point circle fitting method to obtain the curvature value of each sampling point on the section. The maximum and minimum curvature locations are recorded through extreme value extraction. The curvature values on each section are uniformly sampled to 100 points, forming a crack surface equally partitioned curvature profile data matrix. Each row represents a section, and each column corresponds to the curvature at a specific location. The output result is a three-dimensional matrix of size n × 100, where n is the total number of sections along the crack principal axis. When calculating the differential gradient of the curvature change trend between adjacent crack sections based on the crack surface equal-area curvature profile data, the forward first-order difference method is used as the calculation basis, and position-corresponding differential processing is performed on two adjacent rows of data in the profile data matrix. Each differential value reflects the curvature change rate between the two previous and next sections at the same position. The average gradient change at that position is obtained by averaging each column. Then, a local root mean square gradient (RMSG) calculation is performed within every five columns using the sliding window method to analyze the degree of drastic change in the local crack area. By statistically analyzing the maximum RMSG value, average gradient absolute value, gradient standard deviation, and maximum number of mutation points between sections in the entire crack area, a differential gradient data set of the curvature change trend between adjacent crack sections is generated. When analyzing the random process of crack kink angle expansion using the curvature change trend differential gradient, the Markov chain random process theory is used to model the crack profile change state. First, the curvature difference sequence between sections is discretely quantized, and the state threshold is set to The curvature gradient change of the crack profile is divided into five levels (for example, from -2 to +2), and a state transition matrix is constructed. By statistically analyzing the state sequence of the entire crack profile and calculating the transition probability between any two states, the presence of high-frequency asymmetric state jumps is determined. When a high-curvature state frequently transitions to a higher-gradient state, it is considered to be a trend of kink angle expansion. Furthermore, an extreme value distribution fitting method is introduced, using the generalized extreme value distribution (GEV) to fit the RMSG peak sequence. The growth rate of the peak interval is estimated by fitting parameters, and the confidence interval of its shape parameter is calculated. If the shape parameter is greater than 0.3, the crack is considered to have an angle expansion trend. The final output of the kink angle expansion random process data includes the state transition matrix, extreme value fitting parameters, curvature expansion rate, and high-risk segment identification. In the process of identifying the disordered state of the crack kink based on the curvature change trend differential gradient and the kink angle expansion random process data, a weighted discriminant criterion is used to comprehensively evaluate the degree of crack irregularity. The residual mean square error (RMS) between the angle deflection curves of each crack profile along the main axis and the fitted straight line is set as the basic irregularity evaluation index. The discrimination criteria are also combined with the number of regions where the RMSG extreme value exceeds the set threshold, the total proportion of off-diagonal elements in the crack state transition matrix, and the tail thickness parameter of the extreme value fitting function. A logistic regression discriminant method is used to combine and classify multiple indicators, with stable and unstable crack states as the binary classification labels. The training data consists of 120 crack profiles from actual chain link crack samples in industrial applications. The optimal discrimination threshold is determined through 5-fold cross-validation. The final identified crack kink disorder state data is output as a structured dataset, including the crack number, profile irregularity value, state off-diagonal ratio, fitted shape parameters, final identification label, and a list of high-kink risk profile numbers.

[0026] Step S2 includes the following steps: Step S21: performing crack trend topology analysis on the aperture crack characteristic distribution mapping data to obtain aperture crack characteristic trend topology data; Step S22: simulating the drive meshing power loss of the track link based on the aperture crack characteristic trend topology data to obtain the drive wheel meshing power loss data; Step S23: predicting the crack strike depth / width expansion ratio based on the aperture crack characteristic distribution mapping data according to the driving wheel meshing power loss data to obtain the crack strike depth / width expansion ratio; Step S24: Calculate the eccentric wear fracture risk probability based on the driving wheel meshing power loss data and the crack depth / width expansion ratio to obtain eccentric wear fracture risk probability data.

[0027] As an example of the present invention, refer to Figure 2 As shown, in this example, step S2 includes: Step S21: performing crack trend topology analysis on the aperture crack characteristic distribution mapping data to obtain aperture crack characteristic trend topology data; In an embodiment of the present invention, during the crack direction topological analysis of aperture crack feature distribution mapping data, a two-dimensional polar coordinate system is first established with the center of the track link aperture as the coordinate origin, and the crack distribution map is angularly projected. The principal axis direction of each crack is extracted by analyzing the polar angle variation trend of the crack endpoints in the polar coordinate system, and a vector set is formed according to the crack principal axis direction. Using the coordinates of the crack vector start and end points as input parameters, the shortest path algorithm (Dijkstra algorithm) is applied to analyze the crack path connectivity, and the spatial connection relationship and average direction deviation between cracks are statistically calculated. A topological graph modeling approach is further introduced, abstracting each crack vector as an edge in a graph structure, and constructing a crack topology graph using the crack intersections and endpoints as nodes. The degree distribution is used in the topological graph to analyze the intersection density of cracks, and the average crack extension length is extracted by statistically analyzing the path lengths in the graph. The crack strike concentration is analyzed by combining the spatial distribution density function, and the main crack direction is statistically modeled using the Von Mises distribution function. The characteristic strike topology data of the aperture crack is output, including the strike main axis angle set, the node degree distribution of the crack topology graph, the density of the crack intersection point, and the angle value of the main concentrated area of the crack extension direction.

[0028] Step S22: simulating the drive meshing power loss of the track link based on the aperture crack characteristic trend topology data to obtain the drive wheel meshing power loss data; In an embodiment of the present invention, in the process of simulating the power loss of the track segment drive meshing based on the topological data of the aperture crack characteristic trend, the meshing geometry model between the track segment and the drive wheel is first constructed according to the actual engineering assembly parameters, the number of drive wheel teeth is set to 12, the diameter of the track segment aperture and the gear pin is matched to 18.2 mm, and the meshing angle is 20 degrees. During the modeling process, the crack topology data is introduced into the aperture edge contour, the aperture contour disturbance caused by the crack distribution is corrected, and the curvature perturbation superposition method is used to simulate the effect of the crack on the meshing surface. The loading process is simulated by the finite element method, and the material properties are set to 45 steel, the elastic modulus is 210 GPa, and the Poisson's ratio is 0.3. During the load application process, a radial meshing force of 5000 N is applied, and the contact area change and contact pressure distribution when the gear pin is inserted into the aperture are recorded. The power loss is calculated by the energy dissipation method, and the nonlinear contact analysis module is used to obtain the sliding friction energy loss per unit time and integrate it to obtain the total power loss. Output drive wheel meshing power loss data, including sliding contact rate in crack disturbance area, total contact area change rate, meshing energy consumption under unit load, and power loss rate percentage.

[0029] Step S23: predicting the crack strike depth / width expansion ratio based on the aperture crack characteristic distribution mapping data according to the driving wheel meshing power loss data to obtain the crack strike depth / width expansion ratio; In an embodiment of the present invention, when predicting the crack strike depth / width expansion ratio based on the aperture crack feature distribution mapping data based on the drive wheel meshing power loss data, the crack vectors in the aperture crack topology map are first sorted according to the principal axis direction, and the depth growth rate and tangential width expansion rate in the normal direction are calculated based on the crack edge distribution. The crack growth pattern is inferred from the energy consumption per unit area trend in the power loss data. The gray correlation analysis method is used to analyze the correlation between the contact energy distribution and crack geometry changes under different crack directions. For cracks with an angle between the crack strike direction and the meshing force direction between 0° and 30°, depth expansion is dominant, while cracks with an angle between 60° and 90° are dominant in width expansion. A crack growth direction vector is established based on the initial crack depth, topological principal axis direction, and load direction. The depth / width expansion ratio of each crack is then calculated using a multivariate linear fitting method. The resulting crack strike depth / width expansion ratio dataset includes the crack number, principal axis direction angle, initial crack geometry, maximum energy dissipation path angle, and the final predicted depth / width expansion ratio.

[0030] Step S24: Calculate the eccentric wear fracture risk probability based on the driving wheel meshing power loss data and the crack depth / width expansion ratio to obtain eccentric wear fracture risk probability data.

[0031] In an embodiment of the present invention, in the process of calculating the probability of eccentric wear fracture risk based on the drive wheel meshing power loss data and the crack depth / width expansion ratio, a mapping relationship between crack extension and fracture risk is established based on fatigue fracture mechanics theory. First, the stress concentration degree of each crack after applying a periodic load is calculated using the crack stress intensity factor range estimation method, and the geometric stability of the crack extension path is evaluated in combination with the crack depth / width expansion ratio data. A Poisson distribution model is further introduced to describe the frequency of crack occurrence in the track segment during the meshing process, and the overlap rate between the crack position and the power loss hotspot area is used as a risk weighting factor. The probability of aperture failure caused by crack penetration is calculated using the Bayesian posterior estimation method. Finally, a eccentric wear fracture risk probability data set is constructed, which includes the track segment number, the total number of cracks, the crack direction distribution concentration, the depth-width expansion ratio average, the crack-meshing force angle distribution statistics, the power loss distribution moment, the comprehensive fracture risk probability value and the threshold comparison label. This risk probability value can be used for subsequent track segment quality grading detection and screening processing.

[0032] Step S22 includes the following steps: Step S221: performing spatial intersection density analysis on the aperture crack feature trend topology data to obtain the crack spatial intersection density; Step S222: quantifying the aperture load pressure angle offset of the aperture crack characteristic trend topology data based on the crack space intersection density to obtain aperture load pressure angle offset data; Step S223: performing tooth profile force deviation calculation on the aperture load pressure angle offset data to obtain load tooth profile force deviation data; Step S224: performing power transmission intermittent fluctuation simulation based on the load tooth profile force deviation data and the aperture load pressure angle offset data to obtain power transmission intermittent fluctuation data; Step S225: Simulating the drive meshing power loss of the track link according to the load tooth profile force deviation data, the aperture load pressure angle offset data and the power transmission intermittent fluctuation data to obtain the drive wheel meshing power loss data.

[0033] As an example of the present invention, refer to Figure 3 As shown, in this example, step S22 includes: Step S221: performing spatial intersection density analysis on the aperture crack feature trend topology data to obtain the crack spatial intersection density; In an embodiment of the present invention, in the process of performing spatial intersection density analysis on the topological data of the characteristic trend of the aperture crack, the crack topological vector diagram of the link link aperture is first imported into the three-dimensional geometric analysis environment to construct a three-dimensional crack vector field based on the aperture edge. In this vector field, based on the start and end coordinates and the main axis direction of each crack in the topological diagram, the crack vector is mapped to the aperture geometric boundary through the spatial vector projection algorithm. The intersection of the main axis of the crack is used as the spatial intersection point, and the analysis area is set as a circular annulus with a diameter of 5 mm. Each area is a grid unit of 0.5 mm × 0.5 mm, and the number of internal crack intersections is counted unit by unit. The kernel density estimation method is used to calculate the distribution density of the crack intersection in the circumferential direction of the aperture, and the Gaussian kernel function is used to calculate the distribution density of the crack intersection in the circumferential direction of the aperture. = 0.2 mm is a smoothing parameter, and the corresponding crack spatial intersection density value for each angular unit (with a resolution of 1°) is output. This process marks areas with more than three crack intersections per 0.5 mm² as high-density areas. This generates a crack spatial intersection density distribution map and a corresponding data matrix containing angular displacement, intersection density values, and density gradient direction.

[0034] Step S222: quantifying the aperture load pressure angle offset of the aperture crack characteristic trend topology data based on the crack space intersection density to obtain aperture load pressure angle offset data; In an embodiment of the present invention, in the process of quantifying the aperture load pressure angle offset of the aperture crack characteristic trend topological data based on the crack space intersection density, a chain track segment meshing loading mechanical model is first established, and the initial contact pressure angle is set to 20 degrees, with the normal direction of the driving gear tooth surface as the standard direction. According to the distribution of the crack space intersection density, the polar coordinate position of each high-density area is counted, and the local boundary disturbance vector field of the crack area is constructed in combination with the crack main axis direction. The vector synthesis method is used to map the crack main direction disturbance vector to the unit normal of the aperture edge to obtain the local normal deflection angle. The crack intersection density value is then weighted by the crack disturbance vector angle to calculate the mechanical disturbance intensity in different directions on the local unit circle. Based on the disturbance intensity field, the center of gravity offset in the pressure angle direction on the unit circle is calculated. The specific angle value of the mechanical load direction offset from the standard direction (20 degrees) is obtained through sector integration, and the aperture load pressure angle offset data is output. The data content includes the load offset angle, the direction of the maximum offset angle, the average offset angle range, and the corresponding crack density weighted vector set in the offset direction.

[0035] Step S223: performing tooth profile force deviation calculation on the aperture load pressure angle offset data to obtain load tooth profile force deviation data; In an embodiment of the present invention, during the tooth profile force deviation calculation process for the aperture load pressure angle offset data, the tooth profile of the track segment in meshing contact with the drive gear in the actual assembly state is used as a reference, and the track segment tooth profile is set to an involute profile with a module of 6, a pressure angle of 20 degrees, and a tooth width of 20 mm. The pressure angle offset data is introduced into the contact tooth surface model to construct the contact force direction vector field under the actual offset load. Finite element static contact analysis is used to simulate the unit stress distribution of the tooth surface under different pressure angle offset angles. The contact surface material is set to 45 steel, the contact stiffness is 1.5e8 N / m, the friction coefficient is 0.15, and the loading force is a standard load of 5000 N. The contact stress distribution per unit area of the tooth surface under the standard pressure angle and the offset angle is compared to calculate the position change of the maximum stress concentration area, the total contact area change rate, and the increment of the force unevenness factor. The deformation vector decomposition method is introduced to decompose the pressure vector into the normal compressive stress component and the tangential friction stress component, and the difference analysis is performed on all the force points per unit area of the tooth surface. Output load tooth profile force deviation data, including the maximum stress increment value of the tooth surface, the contact area change rate, the force uniformity coefficient change rate, the maximum stress area displacement under the offset angle, and the tooth surface indentation deformation path change vector set, for subsequent chain track segment fatigue life assessment and fracture risk deduction.

[0036] Step S224: performing power transmission intermittent fluctuation simulation based on the load tooth profile force deviation data and the aperture load pressure angle offset data to obtain power transmission intermittent fluctuation data; In this embodiment of the present invention, during the simulation of intermittent power transmission fluctuations based on load tooth profile force deviation data and aperture load pressure angle offset data, the meshing state of the track segment within a standard motion cycle is first used as a reference. The number of contact pairs between the track segment and the gear teeth completed per drive wheel revolution is defined as the base cycle. The base speed is set to 180 rpm, corresponding to a single track segment engagement duration of 20 ms. Within this cycle, the tooth surface unit contact stress variation curve from the load tooth profile force deviation data is synchronized with the angular offset sequence from the aperture load pressure angle offset data to form a tooth profile contact perturbation time series. Subsequently, a dynamic system stiffness matrix is established. Based on a 0.1 ms time step within the meshing cycle, a variable stiffness dynamic integration method is used to iteratively calculate the actual force response of the track segment meshing contact point frame by frame. The effect of the pressure angle perturbation on the normal force distribution at the contact point is introduced within each time step, and the tangential slip offset caused by uneven tooth profile force is superimposed. The single-point contact stiffness of the track segment is updated by combining the Hertzian contact theory in elastic mechanics. During the iterative process, the torque transmission path is tracked in real time, and the actual torque change per unit time is recorded. This value is then used to construct a time series tensor. Finally, a fast Fourier transform analysis is performed on the torque data sequence for 20 consecutive meshing cycles to extract parameters such as the characteristic frequency of the intermittent spectral fluctuations, the main fluctuation amplitude, the instantaneous energy pulse increment, and the fluctuation phase offset. This generates power transmission intermittent fluctuation data, which includes the fluctuation amplitude spectrum corresponding to the time step, the average torque deviation rate, the peak-to-valley difference of the torque change, the fluctuation duration interval, and a sequence of energy loss estimation vectors.

[0037] Step S225: Simulating the drive meshing power loss of the track link according to the load tooth profile force deviation data, the aperture load pressure angle offset data and the power transmission intermittent fluctuation data to obtain the drive wheel meshing power loss data.

[0038] In this embodiment of the present invention, when simulating the power loss of a track segment drive meshing engagement based on load tooth profile force deviation data, aperture load pressure angle offset data, and power transmission intermittent fluctuation data, a mathematical model for meshing contact power transmission is first established. The drive wheel input power is defined as the product of the linear velocity of the meshing point under a standard load of 5000 N, and the tooth surface contact radius is set to 95 mm. Under standard conditions, the meshing power input and output values are calculated for the absence of cracks and offsets and recorded as the baseline efficiency parameter. Force concentration values at different locations in the tooth profile force deviation data are then incorporated, and the surface forces of the track segment tooth profile are binned using a numerical integration method, with each bin being divided into 0.1 mm² units. The effective power transfer path length per unit area of each bin is recalculated based on the torque transmission direction and the stress distribution per unit contact area, superimposed on the contact normal offset angle caused by the aperture pressure angle offset. The main frequency points of the fluctuations in the power transmission intermittent fluctuation data are matched, and the difference between the input and output work at the peak time of the fluctuation within each meshing cycle is calculated and periodically integrated. Finally, the difference between the power loss integral curve and the standard input power curve is normalized to obtain the total drive wheel meshing power loss data. This data includes a set of parameters such as the energy loss curve within the meshing cycle, the meshing efficiency drop rate under superposition of fluctuations, the power loss per unit area in the crack-affected area, the power conversion path offset distribution caused by pressure angle deviation, and the instantaneous power jump value in the stress concentration area of the tooth profile. This data is used to predict the failure life of the subsequent track link and evaluate the reliability of the meshing transmission.

[0039] Step S24 includes the following steps: Step S241: identifying a power transmission deviation concentration area based on the driving wheel meshing power loss data to obtain the power transmission deviation concentration area; Step S242: Deducing the stress increment index of the concentrated area of power transmission deviation based on the driving wheel meshing power loss data to obtain the stress increment index of the concentrated area; Step S243: performing a crack propagation stress critical inference based on the concentrated area stress increment index and the crack strike depth / width expansion ratio to obtain crack propagation stress critical data; Step S244: performing eccentric wear fracture risk probability calculation on the critical crack propagation stress data to obtain eccentric wear fracture risk probability data.

[0040] In this embodiment of the present invention, when identifying areas of concentrated power transfer deviation from drive wheel meshing power loss data, the spatial distribution sequence of power loss per unit time is first extracted from the drive wheel meshing power loss data. The two-dimensional topological grid of the track link meshing contact surface is divided into 0.1 mm × 0.1 mm cells. The cumulative energy loss value for each grid cell over a complete meshing cycle is calculated. This energy loss value is converted into a grid heat map. Using a thresholding method and local extreme value detection, grid cells with energy densities exceeding the global average by more than 30% and reaching extreme values in a local 3×3 neighborhood are marked as initial high-energy cells. Connectivity analysis is then used to merge adjacent high-energy cells to form concentrated power transfer deviation areas. This ultimately creates a data structure for concentrated power transfer deviation areas, including parameters such as region number, grid location index, area, and maximum and average energy density. To derive the stress increment index for concentrated power transmission offset regions based on drive wheel meshing power loss data, each identified concentrated power transmission offset region is first selected. The energy loss curve for all mesh units within the region is then read. Based on the energy conversion law per unit area, the energy loss value is converted into an instantaneous increment of unit contact pressure stress. The total number of steps within the meshing cycle is set to N = 200. The energy change rate within each time step is differentiated to obtain a stress growth rate vector. The maximum value of this vector over the entire cycle is then calculated to form a stress increment ratio index by comparing it to its average value. Combining the region's area and the degree of force concentration, an exponentially weighted stress increment index is then derived for each concentrated power transmission offset region. This index reflects the risk of stress jumps due to contact distortion or load offset in that region. Output includes the region number, stress increment index value, maximum stress jump amplitude, and stress slope per unit area. In the process of inferring the critical stress of crack growth based on the concentrated regional stress increment index and the crack strike depth / width expansion ratio, the values of each region in the concentrated regional stress increment index were mapped to the crack strike depth / width expansion ratio data on a region-by-region basis. Crack segments with an angle of less than 15° between the crack strike direction and the principal axis of force were selected as candidate crack units for growth. The depth / width expansion ratio of these crack units was normalized and used as a weighting parameter for the change rate of the crack tip plastic zone. An incremental evaluation method based on the equivalent stress intensity factor from elastic-plastic fracture mechanics was introduced. The stress increment index was multiplied by the crack expansion ratio to obtain the equivalent crack tip stress rise. This was then compared with the material's fracture toughness parameter (e.g., the K_IC value for Q235B steel is 50 MPa√m) to identify crack units exceeding the limit as critical crack growth units. This generated critical stress data for crack growth, including information such as crack number, crack location, equivalent stress rise value, propagation direction, and stress accumulation rate per unit time.To calculate the probability of eccentric wear fracture risk based on critical crack propagation stress data, a statistical fatigue life method and a fracture probability density method were combined. First, a polynomial fit was performed on the stress accumulation rate of all critical crack propagation stress elements, and the crack failure growth rate was calculated using its time derivative. The crack failure criterion was set as a propagation depth of 10 mm or an expansion ratio of 3.5. Monte Carlo simulation was used to simulate the stress variation path of each critical element over 10,000 operating cycles. The eccentric wear fracture probability value was calculated by dividing the number of samples that reached fracture conditions during the simulation by the total number of samples. Furthermore, considering the area of the concentrated force transmission offset region, the probability of overlap between the crack distribution density and the failure path, the initially obtained single-point fracture probability was regionally weighted to generate eccentric wear fracture risk probability data. This data includes structured outputs such as a fracture risk distribution map, regional fracture probability values, fracture time distribution curves corresponding to crack numbers, and failure priority ranking.

[0041] Step S3 includes the following steps: Step S31: normalizing the eccentric wear fracture risk probability data to obtain normalized eccentric wear fracture risk probability data; Step S32: performing risk probability association learning based on the normalized data of the eccentric wear fracture risk probability to obtain risk probability association learning data; Step S33: performing a multiple regression analysis on the risk probability association learning data to obtain risk probability association regression data; Step S34: constructing a track segment quality detection model for the risk probability associated regression data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

[0042] In an embodiment of the present invention, a maximum-minimum normalization method is first used to preprocess the data for eccentric wear fracture risk probability data. This method extracts the maximum and minimum values from the dataset and linearly maps all risk probability data to a range of 0 to 1, ensuring that risk data collected from different batches of track links has a uniform numerical scale. During the normalization process, the maximum and minimum values of the fracture risk probability for all collected samples are read. Then, for each individual data point, the minimum value is subtracted and divided by the difference between the maximum and minimum values. This process ensures that the normalized data avoids deviations caused by dimensional inconsistencies in subsequent algorithm processing and improves computational stability and convergence speed. The normalized results serve as input for the next step of risk probability association learning. Based on the normalized data for eccentric wear fracture risk probability, principal component analysis (PCA) and correlation coefficient matrix calculation techniques are used for risk probability association learning. PCA technology is used to reduce the dimensionality of the normalized data, extracting the main variables and their linear combinations, reducing redundant information, and enhancing the ability to express key features. Next, the Pearson correlation coefficients between the variables in the normalized data were calculated to construct a risk probability association matrix, which quantitatively describes the intrinsic correlations between multiple factors, such as crack depth, width expansion ratio, and load loss. This process involves performing covariance calculations on the normalized multidimensional risk parameter matrix, extracting principal components through eigenvalue decomposition, and selecting principal components with a cumulative contribution rate of at least 95% for subsequent analysis. The risk probability association learning data contains the weight coefficients of each principal component and the corresponding variable contribution values. Multiple linear regression analysis was performed on the risk probability association learning data to establish a mathematical mapping between each risk factor and fracture probability. The least squares method was used to fit the normalized risk index and its principal component weights as independent variables, and the fracture risk probability as the dependent variable. Regression coefficients were calculated to analyze the contribution of each risk factor to fracture probability and its significance level. In the experiment, a sample size of at least 500 groups was selected, and cross-validation was used to verify the accuracy and stability of the regression model, ensuring that the regression residuals met the normal distribution assumption and lacked multicollinearity. The output was the set of multiple regression equation coefficients and their statistical significance indicators. Based on risk probability-associated regression data, a policy gradient algorithm is used to construct a track segment quality detection model. First, the state space of the detection model is defined as a multidimensional risk factor vector, and the action space is defined as the track segment quality judgment category. The policy gradient algorithm iteratively adjusts the model parameters to maximize the overall track segment detection accuracy index. The specific operation includes calculating the gradient vector and updating the policy parameters using the stochastic gradient descent optimization method until the detection model loss function converges to the preset threshold. During the training process, risk data processed by multivariate regression analysis is input, and the track segment quality detection results are output. Finally, the generated track segment quality detection model is stored in the form of a weight matrix and threshold, and transmitted to the terminal detection device for on-site real-time quality detection.The model has the ability to comprehensively judge complex risk parameters, realizing high-precision track link defect identification and quality classification.

[0043] The present invention further provides a track segment quality detection system based on laser measurement, which is used to perform the track segment quality detection method based on laser measurement as described above. The track segment quality detection system based on laser measurement includes: The aperture crack distribution mapping module is used to collect factory surface morphology laser measurement data of the track segment using laser measurement equipment to obtain track segment surface morphology laser measurement data; perform machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; perform aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; The fracture risk probability calculation module is used to simulate the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain the drive wheel meshing power loss data; and to calculate the eccentric wear fracture risk probability based on the drive wheel meshing power loss data to obtain the eccentric wear fracture risk probability data; The quality detection model construction module is used to construct a track segment quality detection model based on the eccentric wear and fracture risk probability data according to the policy gradient algorithm, obtain the track segment quality detection model, and feed back the track segment quality detection model to the terminal to perform track segment quality detection.

[0044] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting the quality of a track link based on laser measurement, characterized in that: The following steps are involved: Step S1: Collecting factory surface morphology laser measurement data of the track segment using a laser measurement device to obtain track segment surface morphology laser measurement data; performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; performing aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; Step S2: simulating the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain drive wheel meshing power loss data; performing eccentric wear fracture risk probability calculation based on the drive wheel meshing power loss data to obtain eccentric wear fracture risk probability data; Step S3: constructing a track segment quality detection model based on the eccentric wear fracture risk probability data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

2. The method for detecting the quality of a track segment based on laser measurement according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Collecting factory surface morphology laser measurement data of the track link using a laser measurement device to obtain the surface morphology laser measurement data of the track link; Step S12: extracting the track segment aperture state from the track segment surface morphology laser measurement data to obtain the track segment aperture state data; Step S13: performing machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; Step S14: performing aperture crack distribution mapping on the track segment aperture state data based on the track segment machining crack characteristic distribution data to obtain aperture crack characteristic distribution mapping data.

3. The method for detecting the quality of a track segment based on laser measurement according to claim 2, characterized in that: Step S13 includes the following steps: Step S131: performing machining crack roughness distribution identification on the track link surface morphology laser measurement data to obtain machining crack roughness distribution data; Step S132: performing self-similarity fractal feature analysis on the processing crack roughness distribution data to obtain crack roughness distribution fractal feature data; Step S133: Calculating the variance of the texture surface curvature radius based on the crack roughness distribution fractal feature data to obtain the variance of the crack texture surface fractal curvature radius; Step S134: performing crack kink disorder state identification on the crack roughness distribution fractal feature data according to the variance of the crack surface fractal curvature radius to obtain crack kink disorder state data; Step S135: performing machining crack characteristic distribution identification based on the crack surface fractal curvature radius variance and crack kink disorder state data to obtain the track link machining crack characteristic distribution data.

4. The method for detecting the quality of a track segment based on laser measurement according to claim 3, characterized in that: Step S134 includes the following steps: Performing equal-partition curvature profile projection processing on the variance data of the fractal curvature radius of the crack surface to obtain the equal-partition curvature profile data of the crack surface; Based on the crack surface equal-area curvature profile data, the crack roughness distribution fractal feature data is used to calculate the curvature change trend differential gradient between adjacent crack sections, and the curvature change trend differential gradient between adjacent sections is obtained; The random process of crack kink angle expansion is analyzed by using the differential gradient of curvature variation trend to obtain the random process data of kink angle expansion. The crack kink disorder state is identified based on the curvature change trend differential gradient and kink angle expansion random process data to obtain the crack kink disorder state data.

5. The method for detecting the quality of a track segment based on laser measurement according to claim 2, characterized in that: Step S2 includes the following steps: Step S21: performing crack trend topology analysis on the aperture crack characteristic distribution mapping data to obtain aperture crack characteristic trend topology data; Step S22: simulating the drive meshing power loss of the track link based on the aperture crack characteristic trend topology data to obtain the drive wheel meshing power loss data; Step S23: predicting the crack strike depth / width expansion ratio based on the aperture crack characteristic distribution mapping data according to the driving wheel meshing power loss data to obtain the crack strike depth / width expansion ratio; Step S24: Calculate the eccentric wear fracture risk probability based on the driving wheel meshing power loss data and the crack depth / width expansion ratio to obtain eccentric wear fracture risk probability data.

6. The method for detecting the quality of a track segment based on laser measurement according to claim 5, characterized in that: Step S22 includes the following steps: Step S221: performing spatial intersection density analysis on the aperture crack feature trend topology data to obtain the crack spatial intersection density; Step S222: quantifying the aperture load pressure angle offset of the aperture crack characteristic trend topology data based on the crack space intersection density to obtain aperture load pressure angle offset data; Step S223: performing tooth profile force deviation calculation on the aperture load pressure angle offset data to obtain load tooth profile force deviation data; Step S224: performing power transmission intermittent fluctuation simulation based on the load tooth profile force deviation data and the aperture load pressure angle offset data to obtain power transmission intermittent fluctuation data; Step S225: Simulating the drive meshing power loss of the track link according to the load tooth profile force deviation data, the aperture load pressure angle offset data and the power transmission intermittent fluctuation data to obtain the drive wheel meshing power loss data.

7. The method for detecting the quality of a track segment based on laser measurement according to claim 5, characterized in that: Step S24 includes the following steps: Step S241: identifying a power transmission deviation concentration area based on the driving wheel meshing power loss data to obtain the power transmission deviation concentration area; Step S242: Deducing the stress increment index of the concentrated area of power transmission deviation based on the driving wheel meshing power loss data to obtain the stress increment index of the concentrated area; Step S243: performing a crack propagation stress critical inference based on the concentrated area stress increment index and the crack strike depth / width expansion ratio to obtain crack propagation stress critical data; Step S244: performing eccentric wear fracture risk probability calculation on the critical crack propagation stress data to obtain eccentric wear fracture risk probability data.

8. The method for detecting the quality of a track segment based on laser measurement according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: normalizing the eccentric wear fracture risk probability data to obtain normalized eccentric wear fracture risk probability data; Step S32: performing risk probability association learning based on the normalized data of the eccentric wear fracture risk probability to obtain risk probability association learning data; Step S33: performing a multiple regression analysis on the risk probability association learning data to obtain risk probability association regression data; Step S34: constructing a track segment quality detection model for the risk probability associated regression data according to the policy gradient algorithm to obtain the track segment quality detection model, and feeding the track segment quality detection model back to the terminal to perform track segment quality detection.

9. A track link quality detection system based on laser measurement, characterized in that: For executing the track segment quality detection method based on laser measurement according to claim 1, the track segment quality detection system based on laser measurement comprises: The aperture crack distribution mapping module is used to collect factory surface morphology laser measurement data of the track segment using laser measurement equipment to obtain track segment surface morphology laser measurement data; perform machining crack feature distribution identification on the track segment surface morphology laser measurement data to obtain track segment machining crack feature distribution data; perform aperture crack distribution mapping based on the track segment machining crack feature distribution data to obtain aperture crack feature distribution mapping data; The fracture risk probability calculation module is used to simulate the drive wheel meshing power loss of the track link based on the aperture crack characteristic distribution mapping data to obtain the drive wheel meshing power loss data; and to calculate the eccentric wear fracture risk probability based on the drive wheel meshing power loss data to obtain the eccentric wear fracture risk probability data; The quality detection model construction module is used to construct a track segment quality detection model based on the eccentric wear and fracture risk probability data according to the policy gradient algorithm, obtain the track segment quality detection model, and feed back the track segment quality detection model to the terminal to perform track segment quality detection.

Images