Crack detection device and method for precise automobile parts

Abstract

The invention relates to the technical field of nondestructive testing, and discloses a crack detection device and method for precise automobile parts. The method comprises the following steps: carrying out ultrasonic multi-dimensional pre-scanning and acoustic impedance abnormal region identification on a to-be-detected automobile part, carrying out influence range calculation and sensitive threshold classification based on abnormal region distribution data, and obtaining stress-acoustic response time sequence data through partition grabbing force distribution and stress loading; three-dimensional stress field reconstruction and stress concentration area identification are realized; performing dual-frequency stress modulation excitation and multi-physics field synchronous detection on the stress concentration area to obtain a multi-physics field response signal, performing multi-parameter fusion calculation and multi-scale wavelet decomposition of acoustic components on the multi-physics field response signal, and performing feature dimension splicing to obtain crack feature classification data; and calculating and generating a crack detection result based on feature similarity matching and extension prediction. The precision and reliability of crack detection of the automobile parts are remarkably improved, and the damage risk in the detection process is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of nondestructive testing, and particularly to a crack detection device and method for precision automotive parts. Background Art

[0002] In the field of precision manufacturing technology, crack detection is a core link in the quality control and safety assessment of parts, and the accurate identification and location of microcracks are the most critical and challenging stages in crack detection. Especially in the automotive manufacturing field, the crack detection of key automotive parts such as engine parts, transmission system components, and brake system assemblies is directly related to driving safety and vehicle reliability. Automotive parts usually have complex geometric shapes, diverse material combinations, and strict working environment requirements, making crack detection face greater challenges. Therefore, an effective crack detection method to identify these microscopic defects is crucial for ensuring the structural integrity and service safety of automotive parts.

[0003] Currently, ultrasonic testing technology is used to analyze reflected signals and establish a defect identification model to evaluate the crack state, or machine vision technology is attempted to be applied to the surface defect detection process to better identify crack types and geometric features. At the same time, laser scanning technology realizes the three-dimensional reconstruction of surface defects through high-precision measurement, and eddy current testing technology detects near-surface cracks through the principle of electromagnetic induction. However, these methods still face challenges such as limited detection accuracy, inability to simultaneously handle surface and internal cracks, and possible damage to automotive parts during the detection process, and these detection methods often ignore some unique features of precision automotive parts, such as complex geometric shapes, material inhomogeneity, stress concentration effects, multi-scale crack distributions, etc., and these factors may have a significant impact on crack identification and location. That is, the existing crack detection methods for precision automotive parts ignore the multi-physical field coupling characteristics and the complex structural characteristics of automotive parts, resulting in lower accuracy and reliability of the final crack detection results. Summary of the Invention

[0004] The main purpose of the present invention is to solve the problem that the existing crack detection methods for precision automotive parts ignore the multi-physical field coupling characteristics and the complex structural characteristics of automotive parts, resulting in lower accuracy and reliability of the final crack detection results.

[0005] The first aspect of the present invention provides a method for crack detection of precision automotive components. The method for crack detection of precision automotive components includes: performing ultrasonic multi-dimensional pre-scanning on the automotive component to be tested to obtain acoustic reflection data, and identifying acoustic impedance abnormal regions from the acoustic reflection data to obtain abnormal region distribution data; based on the abnormal region distribution data, calculating the influence range of the abnormal regions on the automotive component to be tested and classifying sensitive thresholds to obtain a grasping sensitivity partition, and performing partition grasping force allocation and stress loading on the grasping sensitivity partition to obtain stress-acoustic response time series data; reconstructing the three-dimensional stress field of the automotive component to be tested from the stress-acoustic response time series data to obtain stress concentration regions, and performing dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration regions to obtain multi-physical field response signals; performing multi-parameter fusion calculation on the multi-physical field response signals to obtain crack geometric parameters, and performing multi-scale wavelet decomposition on the acoustic components in the multi-physical field response signals to obtain acoustic time-frequency characteristics, and splicing the crack geometric parameters and the acoustic time-frequency characteristics in terms of feature dimensions to obtain crack feature classification data; performing feature similarity matching and extended prediction calculation on the crack feature classification data to generate a crack detection result.

[0006] Optionally, in the first implementation manner of the first aspect of the present invention, the step of performing ultrasonic multi-dimensional pre-scanning on the automotive component to be tested to obtain acoustic reflection data, and identifying acoustic impedance abnormal regions from the acoustic reflection data to obtain abnormal region distribution data includes: performing full-circumference spiral ultrasonic scanning on the automotive component to be tested to obtain acoustic reflection data, and calculating the sound velocity from the reflection time information in the acoustic reflection data based on the material density parameters corresponding to the automotive component to be tested to obtain the acoustic impedance values of each scanning point; arranging the acoustic impedance values of each scanning point in a three-dimensional space to obtain a three-dimensional acoustic impedance distribution, and calculating the gradient difference between adjacent scanning points in the three-dimensional acoustic impedance distribution and detecting the scanning position points where the detected gradient difference exceeds a preset acoustic impedance change threshold to generate a set of acoustic impedance mutation points; performing aggregation connection of spatially adjacent mutation points in the set of acoustic impedance mutation points to obtain the contour of the acoustic impedance abnormal region, and calculating and extracting various geometric feature parameters of the contour of the acoustic impedance abnormal region to generate abnormal region distribution data.

[0007] Optionally, in the second implementation manner of the first aspect of the present invention, calculating the influence range of the abnormal area on the automotive part to be tested and classifying the sensitivity thresholds based on the abnormal area distribution data to obtain a grasping sensitivity partition, and performing partition grasping force allocation and stress loading on the grasping sensitivity partition to obtain stress-acoustic response time series data, includes: calculating the abnormal intensity value corresponding to each abnormal area in the abnormal area distribution data, and calculating the influence radius of the abnormal intensity value in the abnormal area distribution data to obtain the spatial influence range of each abnormal area; dividing the surface of the automotive part to be tested into regular grids to obtain surface grid partitions, and based on the spatial influence range, calculating the cumulative abnormal influence intensity of each grid unit in the surface grid partitions to obtain the abnormal influence intensity value of each grid unit, and extracting the acoustic impedance value corresponding to each grid unit in the abnormal area distribution data; based on the acoustic impedance value, calculating the sensitivity index weighted value of the abnormal influence intensity value to obtain the grasping sensitivity index of each grid, and based on a preset sensitivity threshold, classifying the grasping sensitivity index to obtain a grasping sensitivity partition, the grasping sensitivity partition including a grasping safety area, a grasping observation area, and a grasping prohibited area; based on a preset partition grasping force parameter, performing differential grasping force configuration on the grasping sensitivity partition to obtain a partition grasping force configuration scheme, and performing three-stage progressive stress loading execution and stress wave change numerical detection on the partition grasping force configuration scheme to obtain stress-acoustic response time series data.

[0008] Optionally, in the third implementation manner of the first aspect of the present invention, calculating the cumulative abnormal influence intensity of each grid unit in the surface grid partitions based on the spatial influence range to obtain the abnormal influence intensity value of each grid unit, includes: positioning and identifying the grid center coordinates of each grid unit in the surface grid partitions to obtain a grid coordinate index table, and calculating the distance from the grid to the abnormal area between the grid coordinate index table and the abnormal area distribution data to obtain a spatial distance matrix; comparing the distance values in the spatial distance matrix with the spatial influence range of each abnormal area to obtain an abnormal comparison result, and marking the grids in the spatial distance matrix whose distance values are less than the spatial influence range of the corresponding abnormal area as affected grids to obtain an influence relationship mapping table; calculating the influence intensity of the affected grids in the influence relationship mapping table to obtain the single-point influence intensity of the grid, and based on the single-point influence intensity of the grid, performing superposition summation numerical normalization of the influence intensity of multiple abnormal areas on each grid unit to obtain the abnormal influence intensity value of each grid unit.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present invention, reconstructing the three-dimensional stress field of the automotive component to be tested from the stress-acoustic response time series data to obtain a stress concentration region, and performing dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration region to obtain a multi-physical field response signal, including: extracting the spatial coordinates of the stress wave values in the stress-acoustic response time series data to obtain discrete stress measurement point data, and calculating the stress values of the voxel points inside the automotive component based on the preset voxel space to obtain the three-dimensional stress field distribution; calculating the stress gradient of each voxel point in the three-dimensional stress field distribution and identifying the voxel positions corresponding to the high stress gradients to obtain a high stress gradient region, and performing spatial clustering and merging and connecting of adjacent high stress gradient voxels on the high stress gradient region to obtain a stress concentration region sensitive to cracks; based on the preset frequency band parameters, performing dual-frequency modulation excitation and multi-physical field synchronous detection on the stress concentration region to obtain synchronous response signals of multiple physical fields, and performing phase synchronization calculation on the periods of the synchronous response signals to obtain a multi-physical field response signal.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present invention, calculating the stress values of the voxel points inside the automotive component based on the preset voxel space to obtain the three-dimensional stress field distribution, including: performing voxelized grid division on the three-dimensional space of the automotive component to obtain a voxel space coordinate system, and calculating the distances from each voxel point in the voxel space coordinate system to multiple stress measurement points to obtain a voxel-measurement point distance relationship table; performing radial basis function transformation on the distance values in the voxel-measurement point distance relationship table to obtain a voxel interpolation weight matrix, and performing weighted product calculation on the voxel interpolation weight matrix and the stress values of the corresponding discrete stress measurement points in the discrete stress measurement point data to obtain the voxel stress contribution values of each voxel point; performing summation calculation on the voxel stress contribution values to obtain voxel interpolation stress values, and performing spatial mapping combination on the voxel interpolation stress values based on the voxel space coordinate system to obtain the three-dimensional stress field distribution.

[0011] Optionally, in the sixth implementation manner of the first aspect of the present invention, the multi-parameter fusion calculation of the multi-physical field response signal to obtain crack geometric parameters, and the multi-scale wavelet decomposition of the acoustic component in the multi-physical field response signal to obtain acoustic time-frequency characteristics include: numerically extracting the acoustic modulation depth, electromagnetic response amplitude, and thermal infrared gradient anomaly value in the multi-physical field response signal to obtain response intensity parameters corresponding to three types of physical fields, and performing wavelet decomposition of multi-time-frequency characteristic parameters on the acoustic component in the multi-physical field response signal to obtain acoustic time-frequency characteristics, and performing weighted fusion calculation of each stress concentration region on each of the response intensity parameters to obtain a comprehensive response intensity distribution; based on the comprehensive response intensity distribution, identifying the positions of the response intensity peak points in each stress concentration region to obtain the coordinates of the crack center candidate points, and performing threshold segmentation on the coordinates of the crack center candidate points based on a preset intensity boundary threshold to obtain at least one crack space boundary contour; measuring and calculating various geometric parameters for each of the crack space boundary contours to obtain crack geometric parameters.

[0012] Optionally, in the seventh implementation manner of the first aspect of the present invention, the feature similarity matching and extended prediction calculation of the crack feature classification data to generate a crack detection result include: performing crack feature clustering on the crack feature classification data to obtain crack feature category labels, and calculating the crack risk level of each category label by performing risk scoring calculation on the crack feature category labels; performing historical case matching on the crack feature classification data based on a preset crack historical case library to obtain a reference case set, and calculating the expansion rate of the crack feature category labels based on the historical expansion records corresponding to the reference case set to generate a crack expansion prediction model; performing evolution prediction calculation of the future crack size on the crack risk level based on the time series parameters corresponding to the crack expansion prediction model to generate a crack detection result.

[0013] In a second aspect of the present invention, there is provided a crack detection device for precision automotive parts, and the crack detection device for precision automotive parts includes: a pre-scanning module, configured to perform ultrasonic multi-dimensional pre-scanning on a to-be-tested automotive part to obtain acoustic reflection data, and identify an abnormal acoustic impedance region from the acoustic reflection data to obtain abnormal region distribution data; an intelligent grasping module, configured to calculate the influence range of the abnormal region on the to-be-tested automotive part and classify sensitive thresholds based on the abnormal region distribution data to obtain a grasping sensitivity partition, and perform partition grasping force allocation and stress loading on the grasping sensitivity partition to obtain stress-acoustic response time series data; a stress excitation module, configured to reconstruct a three-dimensional stress field of the to-be-tested automotive part based on the stress-acoustic response time series data to obtain a stress concentration region, and perform dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration region to obtain multi-physical field response signals; a feature fusion module, configured to perform multi-parameter fusion calculation on the multi-physical field response signals to obtain crack geometric parameters, perform multi-scale wavelet decomposition on the acoustic components in the multi-physical field response signals to obtain acoustic time-frequency features, and splice the crack geometric parameters and the acoustic time-frequency features in terms of feature dimensions to obtain crack feature classification data; an intelligent diagnosis module, configured to perform feature similarity matching and extended prediction calculation on the crack feature classification data to generate a crack detection result.

[0014] The above crack detection device and method for precision automotive parts. In an embodiment of the present invention, by performing ultrasonic multi-dimensional pre-scanning and abnormal acoustic impedance region identification on a to-be-tested automotive part, abnormal region distribution data is obtained, and then based on the calculation of the influence range of the abnormal region and sensitive threshold classification, a grasping sensitivity partition is obtained and differential grasping force configuration is implemented; then, a three-dimensional stress field is reconstructed based on the stress-acoustic response time series data, a stress concentration region is identified and dual-frequency modulation excitation and multi-physical field synchronous detection are implemented; finally, multi-parameter fusion calculation and multi-scale wavelet decomposition are performed on the multi-physical field response signals, and similarity matching and extended prediction are performed in combination with crack geometric parameters and acoustic time-frequency features to output a crack detection result. Through hierarchical multi-physical field fusion detection and intelligent grasping control, the problem of accurate identification of cracks in precision automotive parts is solved. Especially in aspects such as stress concentration identification, non-destructive grasping, and synchronous detection of internal and external cracks, the complex geometric shape, material non-uniformity, and stress distribution characteristics of precision automotive parts are fully considered, effectively improving the detection accuracy; and by adopting a grasping sensitivity partition and an adaptive force control strategy, both non-destructive operation during the detection process is achieved and the reliability of crack identification is enhanced; in addition, through multi-physical field response fusion and predictive analysis, the crack type, geometric parameters, and development trend are accurately identified, thus overall achieving high-precision intelligent detection of cracks in precision automotive parts.

[0015] Other features and advantages of the present invention will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present invention. The objectives and other advantages of the present invention are realized and attained by the structures particularly pointed out in the specification, claims and drawings.

[0016] To make the above objectives, features and advantages of the present invention more comprehensible, the following provides preferred embodiments in conjunction with the accompanying drawings for a detailed description as follows. Brief Description of the Drawings

[0017] Figure 1 Schematic diagram of the first embodiment of the crack detection method for precision automotive parts in an embodiment of the present invention; Figure 2 Schematic diagram of an embodiment of the crack detection device for precision automotive parts in an embodiment of the present invention. Detailed Embodiment

[0018] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0019] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes other steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0020] For ease of understanding of this embodiment, the following describes the specific process of the embodiment of the present invention. Please refer to Figure 1 , the first embodiment of the crack detection method for precision automotive parts in an embodiment of the present invention includes: 101. Perform ultrasonic multi-dimensional pre-scanning on the automotive parts to be tested to obtain acoustic reflection data, and identify the abnormal acoustic impedance regions in the acoustic reflection data to obtain abnormal region distribution data; The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0021] The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, robotics technology, biometric technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0022] In this embodiment, by performing a full - circle spiral ultrasonic scan on the automotive part to be measured, acoustic reflection data is obtained, and based on the material density parameter corresponding to the automotive part to be measured, the sound velocity is calculated for the reflection time information in the acoustic reflection data to obtain the acoustic impedance values of each scan point; the acoustic impedance values of each scan point are arranged in a three - dimensional space to obtain an acoustic impedance three - dimensional distribution, and the gradient difference is calculated for adjacent scan points in the acoustic impedance three - dimensional distribution, and the scan position points where the detected gradient difference exceeds the preset acoustic impedance change threshold are determined to generate a set of acoustic impedance mutation points; the spatial adjacent mutation points in the set of acoustic impedance mutation points are aggregated and connected to obtain the contour of the acoustic impedance abnormal area, and various geometric feature parameters of the contour of the acoustic impedance abnormal area are calculated and extracted to generate abnormal area distribution data.

[0023] In practical applications, first, a full - circle spiral ultrasonic scan is performed. This scan method means that the ultrasonic probe continuously moves along the surface of the automotive part to be measured in a spiral trajectory to ensure comprehensive coverage of the surface and interior of the part. Specifically, during implementation, the ultrasonic probe moves at a preset pitch and angular interval, while emitting ultrasonic pulses and receiving reflected signals. During the scan, multiple acoustic reflection parameters of each scan point are automatically recorded, including the reflection time (the time interval from the emission of the ultrasonic wave to the reception of the reflected signal), the reflection amplitude (the intensity of the reflected signal), and the three - dimensional coordinates (the precise position of the scan point in space). For example, for an automotive wheel bearing with a diameter of 80 mm, when performing a spiral scan, the probe moves at a pitch of 0.1 mm and an angular interval of 2 degrees, and the corresponding reflection time, amplitude, and coordinate information are recorded for each scan point. Next, the sound velocity is calculated. According to the material density parameter of the automotive part to be measured (such as the density of steel is 7.85 g / cm³), the propagation velocity of ultrasonic waves in this material is calculated in combination with the elastic modulus of the material. Then, the sound velocity value is multiplied by the reflection time of each scan point to obtain the ultrasonic propagation distance, and in combination with the reflection amplitude and material characteristics, the acoustic impedance values of each scan point are calculated (which reflects the product of material density and sound velocity and is an important index of the acoustic characteristics of the material). This full - circle spiral scan method ensures the comprehensiveness and continuity of the detection, avoids areas that may be missed by traditional dot - matrix scans, and at the same time, the sound velocity calculation based on the material density parameter improves the accuracy of the acoustic impedance values.

[0024] Secondly, arrange the acoustic impedance values of each obtained scanning point in space according to their corresponding three-dimensional coordinates to construct a three-dimensional distribution of acoustic impedance. This three-dimensional distribution is a spatial data structure containing X, Y, Z coordinate information and corresponding acoustic impedance values. During specific implementation, a coordinate system is established with the geometric center of the automotive part to be measured as the origin, and the acoustic impedance value of each scanning point is mapped to its corresponding spatial position to form a three-dimensional acoustic impedance distribution cloud map. Subsequently, calculate the gradient difference between adjacent scanning points in the three-dimensional distribution of acoustic impedance, that is, calculate the difference in acoustic impedance values between two adjacent points in space. Among them, the gradient difference calculation uses the nearest neighbor algorithm in three-dimensional space to find adjacent points within a preset distance range around each scanning point, and then calculate the difference in acoustic impedance values. For example, when the acoustic impedance value of a certain scanning point is 15.2 MRayl and the acoustic impedance value of its adjacent point is 12.8 MRayl, the gradient difference is 2.4 MRayl. Based on a preset acoustic impedance change threshold (such as 1.5 MRayl) set in advance, when the gradient difference exceeds this threshold, mark the corresponding scanning position point as an acoustic impedance mutation point, and thus form a set of acoustic impedance mutation points by combining all the marked mutation points (which refers to the set of spatial positions where the acoustic impedance values change significantly, and these positions usually correspond to defects, cracks or structural discontinuities inside the material). Through gradient difference calculation and threshold judgment, abnormal regions inside the material can be effectively identified.

[0025] Finally, perform an aggregation connection process on the set of acoustic impedance mutation points for spatially adjacent mutation points, that is, the process of grouping mutation points with relatively close spatial distances into the same abnormal region. During specific implementation, use a distance-based clustering algorithm, set an aggregation distance threshold (such as 0.5 mm), and when the spatial distance between two mutation points is less than this threshold, group them into the same aggregation group. Fit the boundaries of all mutation points within each aggregation group to obtain the contour of the acoustic impedance abnormal region (which refers to the spatial boundary curve or surface surrounding the abnormal mutation points). Next, calculate and extract various geometric feature parameters for each contour of the acoustic impedance abnormal region. The calculation of geometric feature parameters includes: calculating the central coordinates of the abnormal region (the average value of the coordinates of all points within the contour), the covered area (the size of the spatial region surrounded by the contour), and the abnormal intensity value (the statistical value of the acoustic impedance gradient differences of all mutation points within the contour). For example, an abnormal region contains 15 mutation points, and an elliptical contour is obtained through boundary fitting. The calculated central coordinates are (25.3, 18.7, 12.1) mm, the covered area is 3.2 mm², and the abnormal intensity value is 2.8 MRayl. Finally, generate abnormal region distribution data, which contains detailed information about all abnormal regions on the surface and inside the automotive part to be measured, recording the position, size and degree of abnormality of each abnormal region. This aggregation connection and feature extraction method can integrate discrete mutation point information into a complete description of the abnormal region, providing an accurate spatial reference for subsequent grasping sensitivity analysis and detection strategy formulation.

[0026] 102. Based on the abnormal area distribution data, calculate the influence range of the abnormal areas on the automotive parts to be measured and classify the sensitive thresholds, obtain the grasping sensitivity partition, and perform partition grasping force allocation and stress loading on the grasping sensitivity partition to obtain stress-acoustic response time series data; In this embodiment, by calculating the abnormal intensity values corresponding to each abnormal area in the abnormal area distribution data and calculating the influence radius of the abnormal intensity values in the abnormal area distribution data, the spatial influence range of each abnormal area is obtained; the surface of the automotive parts to be measured is divided into regular grids to obtain surface grid partitions, and based on the spatial influence range, the cumulative calculation of the abnormal influence intensity of each grid unit in the surface grid partitions is performed to obtain the abnormal influence intensity values of each grid unit (that is, by positioning and identifying the grid center coordinates of each grid unit in the surface grid partitions, obtaining a grid coordinate index table, and calculating the distance from the grid to the abnormal area between the grid coordinate index table and the abnormal area distribution data to obtain a spatial distance matrix; comparing the distance values in the spatial distance matrix with the spatial influence ranges of each abnormal area to obtain an abnormal comparison result, and marking the grids in the spatial distance matrix where the distance values are less than the spatial influence range of the corresponding abnormal area as affected grids to obtain an influence relationship mapping table; calculating the influence intensity of the affected grids in the influence relationship mapping table to obtain the single-point influence intensity of the grid, and based on the single-point influence intensity of the grid, performing superposition summation and numerical normalization of the influence intensities of multiple abnormal areas on each grid unit to obtain the abnormal influence intensity values of each grid unit), and extracting the acoustic impedance values corresponding to each grid unit in the abnormal area distribution data; based on the acoustic impedance values, perform sensitivity index weighted calculation on the abnormal influence intensity values to obtain the grasping sensitivity index of each grid, and based on a preset sensitive threshold, perform threshold classification on the grasping sensitivity index to obtain the grasping sensitivity partition, where the grasping sensitivity partition includes a grasping safety area, a grasping observation area, and a grasping prohibited area; based on a preset partition grasping force parameter, perform differential grasping force configuration on the grasping sensitivity partition to obtain a partition grasping force configuration plan, and perform three-stage progressive stress loading execution and stress wave change numerical detection on the partition grasping force configuration plan to obtain stress-acoustic response time series data.

[0027] In practical applications, first, calculate the abnormal intensity values corresponding to each abnormal area in the abnormal area distribution data. This abnormal intensity value is a numerical value obtained by statistically calculating the gradient differences of all acoustic impedance mutation points within the abnormal area, reflecting the severity of the abnormal area. Specifically, when implementing, perform a weighted average calculation on the acoustic impedance gradient differences of all mutation points within each abnormal area. The weight is determined according to the distance between the mutation point and the center of the abnormal area. The closer the mutation point is to the center, the greater the weight. For example, an abnormal area contains 8 mutation points, and the gradient differences are 2.1, 1.8, 2.5, 1.9, 2.3, 1.7, 2.0, and 2.2 MRayl respectively. The abnormal intensity value obtained through weighted average calculation is 2.06 MRayl. Next, calculate the influence radius. The influence radius refers to the effective distance range within which the abnormal area affects the surrounding space. The calculation of the influence radius uses an empirical formula based on the abnormal intensity value. The specific method is to multiply the abnormal intensity value by the material-related influence coefficient and then add the square root of the coverage area of the abnormal area. For steel parts, the influence coefficient is taken as 0.8. When the abnormal intensity value is 2.06 MRayl and the coverage area is 3.2 mm², the calculated influence radius is 3.43 mm. Thus, the spatial influence range of each abnormal area is obtained (referring to a spherical space area with the center of the abnormal area as the center of the sphere and the influence radius as the radius. Any position within this area will be affected by this abnormal area). This method of calculating the influence radius based on the abnormal intensity value can accurately quantify the degree of influence of the abnormal area on the surrounding space, ensuring that the spatial diffusion effect of the abnormal area is fully considered during subsequent grasping operations.

[0028] The surface of the automotive component under test is then divided into regular grids, resulting in surface mesh partitions (which uniformly divide the component surface into multiple equal-sized grid cells, such as squares or triangles). Specifically, an appropriate grid size is selected based on the component's surface geometry. For the inner ring of a precision bearing, a 0.5mm × 0.5mm square grid is selected to ensure a high enough grid density to capture subtle changes in anomaly influences, resulting in surface mesh partitions. Anomaly influence intensity accumulation is then calculated. First, the grid center coordinates of each grid cell in the surface mesh partitions are located and identified. The three-dimensional coordinates of the geometric center of each grid cell are calculated, and each grid cell is assigned a unique number to form a grid coordinate index table (a data structure containing grid numbers, center coordinates, and boundary information). Next, the grid coordinate index table and the anomaly region distribution data are used to calculate the distance from each grid cell to the center of each anomaly region. A three-dimensional Euclidean distance algorithm is used to calculate the spatial straight-line distance from each grid center to the center of each anomaly region, forming a spatial distance matrix (a two-dimensional array with rows representing grid cells, columns representing anomaly regions, and matrix elements representing corresponding spatial distance values). The distance values in the spatial distance matrix are then compared one by one with the spatial influence range of each anomaly region. When the distance from a grid cell to an anomaly region is less than the anomaly region's influence radius, the grid cell is marked as affected by the anomaly region, forming an influence relationship mapping table (which records the relationship between each anomaly region and which grid cells are affected). Finally, the influence strength of each affected grid cell in the influence relationship mapping table is calculated. This calculation uses a distance decay function, where closer distances increase the influence strength. The calculation formula is: the influence strength equals the anomaly strength value multiplied by the distance decay factor, where the distance decay factor is 1 minus the ratio of the distance value to the influence radius. For example, if a grid cell is 1.2 mm from an anomaly region, the anomaly region's influence radius is 3.43 mm, and the anomaly strength value is 2.06 MRayl, then the influence strength of the grid cell is 1.34 MRayl. For grid cells affected by multiple anomalies, the influence strengths of each anomaly region are summed and then normalized. Normalization is the process of mapping the influence strength values of all grid cells to a range between 0 and 1 to obtain the anomaly influence strength value for each grid cell. This distance-attenuation-based cumulative impact intensity calculation method accurately reflects the combined impact of multiple anomalies on various surface locations. It also extracts the acoustic impedance value corresponding to each grid cell in the anomaly distribution data. This acoustic impedance value is calculated by spatially interpolating the grid center coordinates with the ultrasonic scan data. This spatial interpolation uses an inverse distance weighting method, selecting the four closest scan points around the grid center and assigning weights based on distance, with closer distances giving greater weights. This calculates the acoustic impedance value at the grid center.

[0029] Furthermore, based on the acoustic impedance values, a sensitivity index weighted calculation is performed on the abnormal influence intensity values to obtain the grasping sensitivity index for each grid (which is a comprehensive evaluation index considering both the abnormal influence intensity and the acoustic properties of the material), that is, by multiplying the abnormal influence intensity value by the acoustic impedance sensitivity coefficient. The acoustic impedance sensitivity coefficient is determined according to the acoustic impedance value of the material. The larger the acoustic impedance value, the denser the material, and the less likely it is to cause damage during grasping, so the sensitivity coefficient is smaller; conversely, the smaller the acoustic impedance value, the more likely there are defects or looseness in the material, and the easier it is to cause damage during grasping, so the sensitivity coefficient is larger. For example, the abnormal influence intensity value of a certain grid is 0.65, the acoustic impedance value is 14.2 MRayl, and the corresponding acoustic impedance sensitivity coefficient is 1.1. The calculated grasping sensitivity index is 0.715. Then, based on the preset sensitivity thresholds (including the safety threshold of 0.3 and the danger threshold of 0.7), the grasping sensitivity index is classified by threshold to obtain the grasping sensitivity partition. When the grasping sensitivity index is less than 0.3, it is classified as the grasping safety zone, indicating that the grasping force can be normally applied to this area without causing damage to the parts; when the grasping sensitivity index is between 0.3 and 0.7, it is classified as the grasping observation zone, indicating that the grasping force needs to be carefully controlled in this area; when the grasping sensitivity index is greater than 0.7, it is classified as the grasping prohibited zone, indicating that the grasping force is strictly prohibited in this area to avoid expanding the defects. This weighted calculation method based on the acoustic impedance sensitivity coefficient can simultaneously consider the influence of abnormal distribution and material properties on grasping safety, and achieve an accurate risk assessment of the surface of the parts.

[0030] Furthermore, based on the preset partition grasping force parameters (the grasping force ranges preset according to the safety requirements of different sensitivity partitions, the grasping force parameter in the grasping safety area is 80 - 100% of the standard grasping force, the grasping force parameter in the grasping observation area is 40 - 60% of the standard grasping force, and the grasping force parameter in the grasping prohibited area is 0 - 20% of the standard grasping force), differential grasping force configuration is performed on the grasping sensitivity partitions to obtain a partition grasping force configuration scheme. The partition grasping force configuration scheme is a set of operation instructions specifying specific grasping force values for each grid unit. For example, for the precision bearing detection with a standard grasping force of 50N, the grasping force in the grasping safety area is configured to be 40 - 50N, the grasping force in the grasping observation area is configured to be 20 - 30N, and the grasping force in the grasping prohibited area is configured to be 0 - 10N. Next, three-stage progressive stress loading execution and stress wave change numerical detection are performed on the partition grasping force configuration scheme. Among them, the three-stage progressive stress loading means that the process of applying the grasping force is divided into three consecutive stages: the initial contact stage, the progressive loading stage, and the stable holding stage; in the initial contact stage, the robotic arm gripper slightly touches the surface of the component, applying 10% of the target grasping force for a duration of 0.5 seconds; in the progressive loading stage, the grasping force is gradually increased from 10% to 100% of the target grasping force at an increasing rate of 18% of the target grasping force per second, with a total duration of 5 seconds; in the stable holding stage, the 100% target grasping force is maintained unchanged, and the duration is determined according to the detection needs. During the entire stress loading process, the numerical changes of the stress wave are monitored in real time. The stress wave change numerical value refers to the measurement data of the internal stress state of the component changing with time, and the stress signal and acoustic signal are synchronously collected through the stress sensors and ultrasonic transducers arranged on the surface of the component. The stress sensors record the stress numerical changes at a sampling frequency of 1000Hz, and the ultrasonic transducers record the amplitude and phase changes of the acoustic reflection signals at the same frequency. The synchronously collected stress data and acoustic data are organized in a time series to form stress-acoustic response time series data. The stress-acoustic response time series data is a multi-dimensional time series data set containing timestamps, stress numerical values, acoustic amplitudes, and acoustic phases. This three-stage progressive stress loading method avoids impact damage to automotive components caused by sudden force application, and at the same time, by synchronously monitoring stress and acoustic responses, it can evaluate the impact of the grasping process on the component state in real time, ensuring the safety and accuracy of the detection process.

[0031] 103. Reconstruct the three-dimensional stress field of the automotive component to be tested from the stress-acoustic response time series data to obtain the stress concentration area, and perform dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration area to obtain multi-physical field response signals; In this embodiment, spatial coordinates are extracted from the stress wave values in the stress-acoustic response time series data to obtain discrete stress measurement point data. Based on a preset voxel space, stress values of internal voxel points of the automotive component are calculated from the discrete stress measurement point data to obtain a three-dimensional stress field distribution (that is, by performing voxelized mesh division on the three-dimensional space of the automotive component to obtain a voxel space coordinate system, calculating the distances from each voxel point in the voxel space coordinate system to multiple stress measurement points to obtain a voxel-measurement point distance relationship table; performing radial basis function transformation on the distance values in the voxel-measurement point distance relationship table to obtain a voxel interpolation weight matrix, and performing weighted product calculation on the voxel interpolation weight matrix and the stress values of the corresponding discrete stress measurement points in the discrete stress measurement point data to obtain the voxel stress contribution values of each voxel point; performing summation calculation on the voxel stress contribution values to obtain voxel interpolation stress values, and performing spatial mapping combination on the voxel interpolation stress values based on the voxel space coordinate system to obtain a three-dimensional stress field distribution); stress gradients are calculated for each voxel point in the three-dimensional stress field distribution, and the voxel positions corresponding to high stress gradients are identified to obtain high stress gradient regions. The high stress gradient regions are spatially clustered and adjacent high stress gradient voxels are merged and connected to obtain stress concentration regions sensitive to cracks; based on preset frequency band parameters, dual-frequency modulation excitation and multi-physical field synchronous detection are performed on the stress concentration regions to obtain synchronous response signals of multiple physical fields, and phase synchronization calculation is performed on the periods of the synchronous response signals to obtain multi-physical field response signals.

[0032] In practical applications, first, the spatial coordinates of the stress wave values in the stress-acoustic response time series data are extracted to obtain discrete stress measurement point data (a data set containing three-dimensional coordinate positions and corresponding stress values, and each measurement point data contains four parameters: X, Y, Z coordinates and stress values). The stress wave values are the stress change data collected by stress sensors distributed on the surface and inside of the component during the aforementioned three-stage progressive stress loading process. The extraction of spatial coordinates refers to identifying the sensor position coordinates corresponding to each stress value from the time series data. For example, for a precision gear component, 25 stress sensors are arranged on its surface and inside. After the extraction of spatial coordinates, 25 discrete stress measurement point data are obtained, and each measurement point contains position coordinates such as (12.5, 8.3, 15.7) mm and the corresponding stress value such as 45.6 MPa. Next, based on the preset voxel space, the stress values of the voxel points inside the automotive component are calculated for the discrete stress measurement point data. The voxel space is a spatial discretization method that divides the internal space of a three-dimensional object into regular cubic units. Each cubic unit is called a voxel, which is similar to the concept of pixels in a two-dimensional image extended to the three-dimensional space. In specific implementation, the three-dimensional space of the automotive component is divided into a voxelized grid. The voxel size is determined according to the geometric size of the component and the required analysis accuracy. For a precision bearing with a diameter of 50 mm, a voxel size of 1 mm × 1 mm × 1 mm is selected for spatial division to obtain a voxel space coordinate system (a three-dimensional coordinate grid system with the geometric center of the component as the origin and containing the coordinates of all voxel centers). Subsequently, the multi-stress measurement point distance is calculated for each voxel point in the voxel space coordinate system. The three-dimensional Euclidean distance formula is used to calculate the spatial straight-line distance from the center of each voxel to each discrete stress measurement point, forming a voxel-measurement point distance relationship table (a data matrix that records the distance values between each voxel and all stress measurement points). Next, a radial basis function transformation is performed on the distance values in the voxel-measurement point distance relationship table (a distance-based interpolation weight calculation method), that is, by using the Gaussian function as the radial basis function, the closer the measurement point is, the greater the influence weight on the voxel, and the influence weight of the farther measurement point decays exponentially, thereby obtaining a voxel interpolation weight matrix (a two-dimensional array containing the influence weight coefficients of each voxel corresponding to each measurement point). Then, a weighted product calculation is performed on the voxel interpolation weight matrix and the stress values of the corresponding discrete stress measurement points in the discrete stress measurement point data to obtain the voxel stress contribution value of each voxel point (referring to the contribution amount of a single stress measurement point to the stress state at a specific voxel position, and the calculation method is to multiply the measurement point stress value by the corresponding interpolation weight). Finally, the voxel stress contribution values are summed to obtain the voxel interpolation stress value, and based on the voxel space coordinate system, the voxel interpolation stress value is spatially mapped and combined to obtain a three-dimensional stress field distribution (a three-dimensional data field that reflects the stress states at various positions inside the component, containing the stress numerical information at each voxel position).This voxel interpolation method based on radial basis functions can accurately reconstruct the continuous stress distribution inside components, with higher accuracy and better smoothness compared to traditional linear interpolation methods, and is particularly suitable for dealing with stress field reconstruction problems under the complex geometries of automotive components.

[0033] Calculate the stress gradient of each voxel point in the three-dimensional stress field distribution and identify the voxel positions corresponding to high stress gradients to obtain high stress gradient regions (which are the set of all voxel positions with high stress gradients, and these regions usually correspond to geometric mutations, material defects, or crack initiation positions inside components). Among them, stress gradient calculation refers to calculating the change rate of stress values between adjacent voxels in space. Using the finite difference method in three-dimensional space, calculate the stress difference between each voxel point and its adjacent voxels in the X, Y, and Z directions, and then calculate the vector norm of the three-direction gradients as the stress gradient value of this voxel. For example, the stress value of a certain voxel is 52.3 MPa, the stress value of its adjacent voxel in the positive X direction is 48.7 MPa, the stress value of its adjacent voxel in the positive Y direction is 55.1 MPa, and the stress value of its adjacent voxel in the positive Z direction is 51.8 MPa. Through finite difference calculation, the stress gradient value of this voxel is 4.2 MPa / mm. The identification of voxel positions corresponding to high stress gradients is achieved by setting a stress gradient threshold. When the stress gradient value of a voxel exceeds the preset threshold, this voxel is marked as a high stress gradient voxel. For precision automotive components made of steel, the stress gradient threshold is set to 5.0 MPa / mm. The setting of this threshold is based on the theoretical analysis of the stress concentration coefficient in material mechanics. When the stress gradient exceeds this value, it indicates that there is a significant stress concentration phenomenon in this region. Next, perform spatial clustering on the high stress gradient regions and merge and connect adjacent high stress gradient voxels to obtain stress concentration regions sensitive to cracks. Among them, spatial clustering refers to the process of grouping high stress gradient voxels that are close in space into the same region. Using the density-based DBSCAN clustering algorithm, set the clustering radius to 2 mm and the minimum number of clustering points to 5 voxels. The merging and connection of adjacent high stress gradient voxels means forming a complete continuous region by connectivity analysis for the voxels within the same cluster. The connectivity judgment is based on the six-neighborhood connection criterion, that is, the directly adjacent voxels of a voxel in the X, Y, and Z directions are considered connected. And the stress concentration region is a high stress gradient region with a complete boundary formed after clustering and connection processing. These regions represent the sensitive positions inside components where cracks are most likely to initiate and propagate. This method for identifying stress concentration regions based on stress gradients can effectively locate the weak links inside components, and can more accurately reflect the crack formation mechanism compared to traditional judgment methods based on the magnitude of stress values, because the generation of cracks is mainly related to the non-uniform distribution of stress rather than the absolute stress value.

[0034] Based on preset frequency band parameters (including two parameters: fundamental frequency and modulation frequency. The fundamental frequency is selected as the characteristic frequency of the component material. For example, for precision automotive components made of steel, 2.5 MHz is selected, and the modulation frequency is selected as one-tenth of the fundamental frequency, i.e., 250 kHz), dual-frequency modulation excitation and multi-physical field synchronous detection are performed on the stress concentration area to obtain synchronous response signals of multiple physical fields. Among them, dual-frequency modulation excitation means applying two excitation signals with different frequencies simultaneously. By frequency modulation, rich spectral components are generated, which can stimulate the response characteristics of defects at different scales inside the material. Specifically, during implementation, an ultrasonic transducer, an electromagnetic induction probe, and an infrared thermal imaging sensor are placed on the surface of the identified stress concentration area. The ultrasonic transducer emits dual-frequency modulated ultrasonic signals to excite the stress concentration area, and at the same time, the three sensors synchronously record the response signals. Multi-physical field synchronous detection means simultaneously monitoring the response signals of three different physical fields: acoustic, electromagnetic, and thermal infrared. The acoustic response reflects the change in the elastic properties of the material, the electromagnetic response reflects the change in the conductivity and permeability of the material, and the thermal infrared response reflects the change in the heat conduction characteristics of the material. For example, when performing dual-frequency excitation on a stress concentration area of 2 mm × 3 mm, the ultrasonic transducer records a 20% change in the amplitude of the reflected signal, the electromagnetic induction probe records a 15-degree phase shift of the eddy current signal, and the infrared thermal imaging sensor records an abnormal temperature gradient of 0.8 °C. The synchronous response signals of multiple physical fields are a multi-dimensional data set containing timestamps and the response values of each physical field. Next, phase synchronization calculation is performed on the synchronous response signals, that is, by analyzing the phase relationship between the response signals of each physical field, the cross-correlation function is used to calculate the time delay and phase difference between the signals of different physical fields to ensure that all response signals are aligned in time, and multi-physical field response signals (including the amplitude, phase, and spectral information of each physical field) are obtained. This method of dual-frequency modulation excitation combined with multi-physical field synchronous detection can obtain richer defect characteristic information than single physical field detection. Different physical fields have different sensitivities to different types and scales of defects. Acoustic detection is sensitive to geometric defects, electromagnetic detection is sensitive to conductive defects, and thermal infrared detection is sensitive to heat transfer defects. The combination of the three can achieve a comprehensive characterization of crack characteristics.

[0035] 104. Perform multi-parameter fusion calculation on the multi-physical field response signals to obtain crack geometric parameters, perform multi-scale wavelet decomposition on the acoustic components in the multi-physical field response signals to obtain acoustic time-frequency characteristics, and splice the crack geometric parameters and acoustic time-frequency characteristics in terms of feature dimensions to obtain crack feature classification data; In this embodiment, numerical extraction is performed on the acoustic modulation depth, electromagnetic response amplitude, and thermal infrared gradient anomaly value in the multi-physical field response signal to obtain the response intensity parameters corresponding to the three types of physical fields, and wavelet decomposition of the multi-time-frequency characteristic parameters of the acoustic component in the multi-physical field response signal is performed to obtain the acoustic time-frequency characteristics, and weighted fusion calculation of each stress concentration region is performed on each of the response intensity parameters to obtain the comprehensive response intensity distribution; based on the comprehensive response intensity distribution, the positions of the response intensity peak points in each stress concentration region are identified to obtain the crack center candidate point coordinates, and based on a preset intensity boundary threshold, threshold segmentation is performed on the crack center candidate point coordinates to obtain at least one crack space boundary contour; measurement calculations of various geometric parameters are performed on each of the crack space boundary contours to obtain the crack geometric parameters.

[0036] In practical applications, first, numerical extraction is performed on the acoustic modulation depth, electromagnetic response amplitude, and thermal infrared gradient anomaly value in the multi-physical field response signal to obtain the response intensity parameters corresponding to the three types of physical fields. Among them, the acoustic modulation depth refers to the modulation degree of the acoustic signal under dual-frequency modulation excitation. The calculation method is to divide the difference between the maximum amplitude and the minimum amplitude of the acoustic response signal by the average amplitude, which reflects the degree to which the acoustic signal is affected by the modulation frequency. For example, for the acoustic response signal in a stress concentration area, the maximum amplitude is 1.8V, the minimum amplitude is 0.6V, and the average amplitude is 1.2V. The calculated acoustic modulation depth is 1.0. The electromagnetic response amplitude refers to the peak amplitude of the eddy current signal detected by the electromagnetic induction probe, and the maximum peak value is directly extracted from the time-domain waveform of the electromagnetic response signal as the electromagnetic response amplitude parameter. The thermal infrared gradient anomaly value refers to the deviation value between the temperature gradient detected by the infrared thermal imaging sensor and the temperature gradient in the normal area, which is obtained by calculating the difference between the spatial derivative of the temperature in the detection area and the spatial derivative of the temperature in the reference area. The response intensity parameters corresponding to the three types of physical fields are a numerical set including the acoustic modulation depth, electromagnetic response amplitude, and thermal infrared gradient anomaly value. Each stress concentration area corresponds to a set of response intensity parameters. Next, wavelet decomposition of multi-time-frequency characteristic parameters is performed on the acoustic component in the multi-physical field response signal, that is, continuous wavelet transform is used for time-frequency domain analysis of the acoustic signal. The Morlet wavelet is selected as the mother wavelet function, and the decomposition scale range is set from 1 to 64, corresponding to a frequency range from 39 kHz to 2.5 MHz. During the wavelet decomposition process, the energy distribution, frequency centroid, bandwidth, and peak frequency in the time-frequency domain are extracted as the acoustic time-frequency characteristic parameters, and the acoustic time-frequency characteristics (a multi-dimensional parameter vector reflecting the characteristic distribution of the acoustic signal at different times and frequencies) are obtained. Then, weighted fusion calculation is performed on each response intensity parameter for each stress concentration area, that is, a weight distribution method based on the reliability of the physical field is adopted. The weight of the acoustic modulation depth is 0.5, the weight of the electromagnetic response amplitude is 0.3, and the weight of the thermal infrared gradient anomaly value is 0.2. The weight distribution is based on the sensitivity and accuracy evaluation of each physical field for crack detection, and the comprehensive response intensity distribution is obtained (a spatial distribution data obtained by weighted summing the three types of response intensity parameters according to the weights, which reflects the comprehensive anomaly degree of each stress concentration area). This multi-physical field parameter extraction and weighted fusion method makes full use of the complementary detection capabilities of different physical fields for crack characteristics. Acoustic detection is sensitive to geometric discontinuities, electromagnetic detection is sensitive to changes in conductivity, and thermal infrared detection is sensitive to thermal conduction anomalies. The fusion of the three can improve the accuracy and reliability of crack identification.

[0037] Furthermore, based on the comprehensive response intensity distribution, the positions of the peak points of the response intensity within each stress concentration region are identified. That is, by using the local maximum detection algorithm, a spatial scan is performed on the comprehensive response intensity distribution to find the position point with the maximum response intensity value within each stress concentration region, and the coordinates of the candidate crack center points are obtained. During specific implementation, the response intensity values of all voxel points within each stress concentration region are compared, and the position with the maximum response intensity value and greater than the surrounding neighboring voxels is selected as the peak point. For example, a certain stress concentration region contains 120 voxels, and 3 peak points are identified through local maximum detection, with coordinates (23.5, 15.2, 8.7)mm, (25.1, 16.8, 9.3)mm, and (24.2, 14.6, 10.1)mm respectively. The coordinates of the candidate crack center points are the three-dimensional coordinate set of the positions of all peak points, and these positions represent the most likely center positions of the crack. Next, based on a preset intensity boundary threshold (set to 60% of the peak response intensity, and this threshold is set based on the attenuation characteristics of the crack response signal. The response intensity decays exponentially with increasing distance from the crack center, and the 60% threshold can effectively delimit the effective influence range of the crack), threshold segmentation is performed on the coordinates of the candidate crack center points. That is, starting from the candidate crack center points, the positions where the response intensity value is equal to the boundary threshold are searched outward, and these position points are connected to form a closed spatial boundary, obtaining at least one crack spatial boundary contour (which is a three-dimensional space surface surrounding the crack region, and each candidate crack center point corresponds to an independent boundary contour). For example, for a candidate crack center point with a peak response intensity of 2.4, a boundary threshold of 1.44 is set, and an ellipsoidal spatial boundary contour is obtained through threshold segmentation, with a major axis of 2.8mm, a minor axis of 1.6mm, and a thickness of 0.9mm. This method of crack location based on the peak of the response intensity and boundary extraction through threshold segmentation can accurately determine the spatial position and geometric range of the crack, and has higher three-dimensional positioning accuracy compared to traditional image processing methods, especially suitable for crack detection in the complex geometric environment inside automotive parts.

[0038] Furthermore, various geometric parameters of the spatial boundary contours of each crack are measured and calculated to obtain crack geometric parameters. The geometric parameter measurement includes three main parameters: length, depth, and width. The length refers to the maximum distance of the crack in the main propagation direction and is obtained by calculating the projected length of the contour boundary points in the main direction; the depth refers to the maximum penetration distance of the crack in the thickness direction and is obtained by calculating the maximum span of the contour in the Z-axis direction; the width refers to the maximum distance of the crack in the direction perpendicular to the main propagation direction and is obtained by calculating the maximum span of the contour in the direction perpendicular to the main direction. For example, after geometric parameter measurement, the length of a certain crack spatial boundary contour is 2.8 mm, the depth is 0.7 mm, and the width is 0.3 mm. Crack geometric parameters are quantitative indicators that describe the shape and size characteristics of cracks, and these parameters directly reflect the severity and development status of cracks. Furthermore, feature dimension splicing is performed on the crack geometric parameters and acoustic time-frequency characteristics, that is, by combining feature parameters from different sources into a unified feature vector to obtain crack feature classification data. Specifically, during implementation, the length, depth, and width in the crack geometric parameters are array-spliced with the main frequency, concentration degree, and dispersion degree in the acoustic time-frequency characteristics. Array splicing is the process of arranging six numerical values in a fixed order to form a one-dimensional array, and the arrangement order is [length, depth, width, main frequency, concentration degree, dispersion degree]. For example, the geometric parameters of a certain crack are [2.8 mm, 0.7 mm, 0.3 mm], and the acoustic time-frequency characteristics are [2.45 MHz, 0.82, 0.31]. After splicing, a six-dimensional feature vector [2.8, 0.7, 0.3, 2.45, 0.82, 0.31] is obtained. Crack feature classification data is a comprehensive feature description containing six dimensions. This data structure combines the geometric morphology characteristics and acoustic dynamic characteristics of cracks and can comprehensively describe the multi-dimensional characteristics of cracks. This feature representation method that combines geometric parameters and acoustic characteristics overcomes the limitations of single-feature description. Geometric parameters provide spatial morphology information of cracks, and acoustic characteristics provide material response information of cracks. The combination of the two can more accurately characterize the essential characteristics of cracks.

[0039] 105. Perform feature similarity matching and extended prediction calculation on the crack feature classification data to generate crack detection results.

[0040] In this embodiment, crack feature clustering is performed on the crack feature classification data to obtain crack feature category labels, and risk score calculation is performed on the crack feature category labels to obtain the crack risk levels of each category label; based on a preset crack historical case library, historical case matching is performed on the crack feature classification data to obtain a reference case set, and based on the historical extension records corresponding to the reference case set, an expansion rate calculation is performed on the crack feature category labels to generate a crack expansion prediction model; based on the time series parameters corresponding to the crack expansion prediction model, an evolution prediction calculation of the future crack size is performed on the crack risk level to generate a crack detection result.

[0041] In practical applications, crack feature clustering is performed on the crack feature classification data, that is, cracks with similar features are grouped into the same category of data. The K-means clustering algorithm is used to group six-dimensional crack feature classification data to obtain crack feature category labels. Specifically, when implementing, the number of clusters is set to 5 categories, corresponding to five typical crack types: microcracks, surface cracks, through cracks, bifurcated cracks, and composite cracks. In the clustering process, the six-dimensional feature data is first standardized, and the numerical values of each dimension such as length, depth, width, main frequency, concentration, and dispersion are mapped into the 0-1 interval to eliminate the influence of different dimensions on the clustering result. Then, the Euclidean distance is used to calculate the similarity between feature vectors. The smaller the distance, the more similar the crack features are. For example, the standardized feature vector of a certain crack is [0.28, 0.35, 0.15, 0.82, 0.67, 0.41], and after clustering analysis, it is classified into the surface crack category, and the corresponding crack feature category label is "Type-2". The crack feature category label is a classification number that identifies different crack types, and each label represents a group of cracks with similar feature patterns. Next, risk score calculation is performed on the crack feature category labels. By using a weighted scoring method based on the geometric parameters and acoustic feature parameters of the crack, with the length weight of 0.4, the depth weight of 0.3, the width weight of 0.2, and the acoustic anomaly degree weight of 0.1, the crack risk levels of each category label are obtained. The specific calculation method is to sum the standardized feature values of all cracks in each category according to the weights to obtain the comprehensive risk score of the category. According to the scoring results, the risk levels are divided. A score of 0-0.3 is low risk, 0.3-0.6 is medium risk, 0.6-0.8 is high risk, and 0.8-1.0 is extremely high risk. For example, the risk score of the surface crack category is 0.52, corresponding to the medium risk level. This risk grading method based on feature clustering can convert complex multi-dimensional feature information into an intuitive risk level representation, facilitating engineering personnel to quickly judge the danger degree of cracks and the processing priority.

[0042] Based on a preset crack history case library (i.e., a database pre-collected and organized containing crack type feature data, development processes, and final results of various precision automotive components. Each record in the case library includes the initial features of the crack, feature changes during the development process, and the final failure mode), historical case matching is performed on the crack feature classification data. That is, by using the nearest neighbor search algorithm, the Euclidean distance between the currently detected crack features and the features of all historical cases in the case library is calculated, and the 10 cases with the smallest distance are selected as reference cases to obtain a reference case set. Specifically, during implementation, the distance is calculated between the six-dimensional feature vector of the current crack and the feature vectors of each historical case in the case library. The distance threshold is set to 0.15, and cases with a distance less than this threshold are considered to be similar. For example, the distance between the feature vector [0.28, 0.35, 0.15, 0.82, 0.67, 0.41] of a detected surface crack and the historical case numbered HC-247 in the case library is 0.12, and it is selected into the reference case set. The reference case set is a combination of historical cases similar to the currently detected crack features, and the development processes of these cases provide a reference basis for predicting the future evolution of the current crack. Based on the historical extension records corresponding to the reference case set (which is a data sequence of the crack size changing with time in the reference cases, including length, depth, and width measurement values at different time points), the expansion rate of the crack feature category label is calculated. By using the linear regression analysis method, the size changes in the historical extension records are fitted, and the average expansion rates of the crack in the three directions of length, depth, and width are calculated to generate a crack expansion prediction model. For example, the average length expansion rate of the 10 cases in the reference case set is 0.08 mm / month, the depth expansion rate is 0.03 mm / month, and the width expansion rate is 0.01 mm / month. The crack expansion prediction model is a mathematical relationship describing the crack development law obtained through statistical analysis of historical data. This prediction model includes expansion rate parameters and a time-dependent function. This prediction method based on historical case matching makes full use of past detection experience and failure cases, and has higher practicality and accuracy compared to pure theoretical analysis.

[0043] Based on the time series parameters corresponding to the crack propagation prediction model (including three key parameters: prediction time span, time step, and confidence interval. The prediction time span is set to 24 months, the time step is set to 1 month, and the confidence interval is set to 95%), the evolution prediction calculation of the future crack size for the crack risk level is carried out. By using the Monte Carlo simulation method, considering the randomness and uncertainty of the propagation rate, the probability distribution of the crack size at each time node is estimated to generate the crack detection result. The specific calculation process is to superimpose random fluctuations on the basis of the average propagation rate. The random fluctuations follow a normal distribution with a standard deviation of 20% of the average propagation rate. For example, for a surface crack with an initial length of 2.8 mm, based on the propagation prediction model, the expected length after 6 months is 3.28 mm, the expected length after 12 months is 3.76 mm, and the expected length after 18 months is 4.24 mm. At the same time, the change in the risk level at each predicted time point is calculated. When the predicted size exceeds the safety threshold, the risk level is correspondingly increased. The safety threshold is determined according to the key size and safety factor of the component. For precision bearing components, the length safety threshold is set to 5.0 mm, and the depth safety threshold is set to 1.5 mm. The generated crack detection result includes four parts: the classification information of the current crack, the risk level, the predicted development trend, and the recommended treatment measures. The detection result is presented in the form of a structured report, including the crack position coordinates, geometric dimensions, category label, current risk level, the size evolution prediction curve for the next 24 months, and the corresponding risk level change trend. For example, a certain detection result shows that a surface crack located at the coordinates (25.3, 18.7, 12.1) mm is currently at a medium risk level and is expected to develop into a high risk level after 12 months. It is recommended to carry out maintenance treatment within 8 months. This detection result based on time series prediction not only provides the current state assessment, but more importantly, predicts the future development trend, enabling the maintenance decision to change from passive response to active prevention.

[0044] In the embodiments of the present invention, by performing ultrasonic multi-dimensional pre-scanning and acoustic impedance abnormal area identification on the automotive parts to be tested, the abnormal area distribution data is obtained. Then, based on the calculation of the influence range of the abnormal area and the classification of sensitive thresholds, the grasping sensitivity partition is obtained and the differential grasping force configuration is implemented. Next, the three-dimensional stress field is reconstructed through the stress-acoustic response time series data, the stress concentration area is identified, and the dual-frequency modulation excitation and multi-physical field synchronous detection are implemented. Finally, the multi-parameter fusion calculation and multi-scale wavelet decomposition are performed on the multi-physical field response signals, and the similarity matching and extended prediction are carried out in combination with the crack geometric parameters and acoustic time-frequency characteristics, and the crack detection result is output. Through the hierarchical multi-physical field fusion detection and intelligent grasping control, the problem of accurate identification of crack detection of precision automotive parts is solved. Especially in aspects such as stress concentration identification, non-destructive grasping, and synchronous detection of internal and external cracks, the complex geometric shape, material inhomogeneity, and stress distribution characteristics of precision automotive parts are fully considered, effectively improving the detection accuracy. And the grasping sensitivity partition and adaptive force control strategy are adopted, which not only realizes the non-destructive operation in the detection process but also enhances the reliability of crack identification. In addition, through the multi-physical field response fusion and predictive analysis, the crack type, geometric parameters, and development trend are accurately identified, thus realizing the high-precision intelligent detection of cracks in precision automotive parts as a whole.

[0045] The crack detection method for precision automotive parts in the embodiments of the present invention has been described above. Next, the crack detection device for precision automotive parts in the embodiments of the present invention will be described. Please refer to Figure 2 , an embodiment of the crack detection device for precision automotive parts in the embodiments of the present invention includes: A pre-scanning module 201, configured to perform ultrasonic multi-dimensional pre-scanning on the automotive parts to be tested, obtain acoustic reflection data, and identify the acoustic impedance abnormal area of the acoustic reflection data to obtain abnormal area distribution data; An intelligent grasping module 202, configured to calculate the influence range of the abnormal area of the automotive parts to be tested and classify the sensitive thresholds based on the abnormal area distribution data, obtain the grasping sensitivity partition, and perform partition grasping force allocation and stress loading on the grasping sensitivity partition to obtain stress-acoustic response time series data; A stress excitation module 203, configured to reconstruct the three-dimensional stress field of the automotive parts to be tested through the stress-acoustic response time series data, obtain the stress concentration area, and perform dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration area to obtain multi-physical field response signals; A feature fusion module 204 is configured to perform multi-parameter fusion calculation on the multi-physical-field response signals to obtain crack geometric parameters, perform multi-scale wavelet decomposition on the acoustic components in the multi-physical-field response signals to obtain acoustic time-frequency features, and splice the feature dimensions of the crack geometric parameters and the acoustic time-frequency features to obtain crack feature classification data; An intelligent diagnosis module 205 is configured to perform feature similarity matching and extended prediction calculation on the crack feature classification data to generate a crack detection result.

[0046] In an embodiment of the present invention, through hierarchical multi-physical-field fusion detection and intelligent grasping control, the problem of accurate identification of cracks in precision automotive parts is solved. Especially in aspects such as stress concentration identification, non-destructive grasping, and synchronous detection of internal and external cracks, the complex geometric shape, material non-uniformity, and stress distribution characteristics of precision automotive parts are fully considered, effectively improving the detection accuracy; and by adopting a grasping sensitivity zoning and adaptive force control strategy, both non-destructive operation during the detection process is achieved and the reliability of crack identification is enhanced; in addition, through multi-physical-field response fusion and predictive analysis, the crack type, geometric parameters, and development trend are accurately identified, thus overall realizing high-precision intelligent detection of cracks in precision automotive parts.

[0047] This application can be used in numerous general or specific computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and so on. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This application can also be practiced in a distributed computing environment where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0048] As mentioned above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for crack detection of precision automotive parts, characterized in that, The crack detection method for the precision automotive parts includes: Performing ultrasonic multi-dimensional pre-scanning on the automotive parts to be tested to obtain acoustic reflection data, and identifying the acoustic impedance abnormal regions from the acoustic reflection data to obtain the abnormal region distribution data; Based on the abnormal region distribution data, calculating the influence range of the abnormal regions on the automotive parts to be tested and classifying the sensitive thresholds to obtain the grasping sensitivity partitions, and performing partitioned grasping force allocation and stress loading on the grasping sensitivity partitions to obtain the stress-acoustic response time series data; Reconstructing the three-dimensional stress field of the automotive parts to be tested from the stress-acoustic response time series data to obtain the stress concentration regions, and performing dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration regions to obtain the multi-physical field response signals; Performing multi-parameter fusion calculation on the multi-physical field response signals to obtain the crack geometric parameters, and performing multi-scale wavelet decomposition on the acoustic components in the multi-physical field response signals to obtain the acoustic time-frequency characteristics, and splicing the crack geometric parameters and the acoustic time-frequency characteristics in terms of feature dimensions to obtain the crack feature classification data; Performing feature similarity matching and extended prediction calculation on the crack feature classification data to generate the crack detection results.

2. The crack detection method for precision automotive parts according to claim 1, characterized in that, The step of performing ultrasonic multi-dimensional pre-scanning on the automotive parts to be tested to obtain acoustic reflection data, and identifying the acoustic impedance abnormal regions from the acoustic reflection data to obtain the abnormal region distribution data includes: Performing full-circle spiral ultrasonic scanning on the automotive parts to be tested to obtain acoustic reflection data, and calculating the sound velocity from the reflection time information in the acoustic reflection data based on the material density parameters corresponding to the automotive parts to be tested to obtain the acoustic impedance values of each scanning point; Performing three-dimensional spatial arrangement on the acoustic impedance values of each scanning point to obtain the three-dimensional acoustic impedance distribution, and calculating the gradient difference between adjacent scanning points in the three-dimensional acoustic impedance distribution and detecting the scanning position points where the detection gradient difference exceeds the preset acoustic impedance change threshold to generate a set of acoustic impedance mutation points; Performing aggregation connection of spatially adjacent mutation points in the set of acoustic impedance mutation points to obtain the contour of the acoustic impedance abnormal region, and calculating and extracting various geometric feature parameters of the contour of the acoustic impedance abnormal region to generate the abnormal region distribution data.

3. The crack detection method for precision automotive parts according to claim 1, characterized in that, The step of, based on the abnormal region distribution data, calculating the influence range of the abnormal regions on the automotive parts to be tested and classifying the sensitive thresholds to obtain the grasping sensitivity partitions, and performing partitioned grasping force allocation and stress loading on the grasping sensitivity partitions to obtain the stress-acoustic response time series data includes: Calculating the abnormal intensity values corresponding to each abnormal region in the abnormal region distribution data, and calculating the influence radius of the abnormal intensity values in the abnormal region distribution data to obtain the spatial influence range of each abnormal region; Perform regular grid division on the surface of the automotive part to be tested to obtain a surface grid partition. Based on the spatial influence range, perform cumulative calculation of the abnormal influence intensity for each grid cell in the surface grid partition to obtain the abnormal influence intensity value of each grid cell, and extract the acoustic impedance value corresponding to each grid cell in the abnormal area distribution data; Based on the acoustic impedance value, perform sensitivity index weighted calculation on the abnormal influence intensity value to obtain the grasping sensitivity index of each grid, and based on a preset sensitivity threshold, perform threshold classification on the grasping sensitivity index to obtain a grasping sensitivity partition, where the grasping sensitivity partition includes a grasping safety area, a grasping observation area, and a grasping prohibited area; Based on preset partition grasping force parameters, perform differential grasping force configuration on the grasping sensitivity partition to obtain a partition grasping force configuration plan, and perform three-stage progressive stress loading execution and stress wave change numerical detection on the partition grasping force configuration plan to obtain stress-acoustic response time series data.

4. The crack detection method for precision automotive parts according to claim 3, wherein, The performing cumulative calculation of the abnormal influence intensity for each grid cell in the surface grid partition based on the spatial influence range to obtain the abnormal influence intensity value of each grid cell includes: Locate and identify the grid center coordinates for each grid cell in the surface grid partition to obtain a grid coordinate index table, and calculate the distance from the grid to the abnormal area between the grid coordinate index table and the abnormal area distribution data to obtain a spatial distance matrix; Compare the distance values in the spatial distance matrix with the spatial influence range of each abnormal area to obtain an abnormal comparison result, and mark the grids in the spatial distance matrix whose distance values are less than the spatial influence range of the corresponding abnormal area in the abnormal comparison result as affected grids to obtain an influence relationship mapping table; Calculate the influence intensity of the affected grids in the influence relationship mapping table to obtain the single-point grid influence intensity, and based on the single-point grid influence intensity, perform superposition summation and numerical normalization of the influence intensities of multiple abnormal areas on each grid cell to obtain the abnormal influence intensity value of each grid cell.

5. The crack detection method for precision automotive parts according to claim 1, wherein The performing three-dimensional stress field reconstruction of the automotive part to be tested on the stress-acoustic response time series data to obtain a stress concentration area, and performing dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration area to obtain a multi-physical field response signal includes: Extract the spatial coordinates of the stress wave values in the stress-acoustic response time series data to obtain discrete stress measurement point data, and based on a preset voxel space, calculate the stress values of the voxel points inside the automotive part for the discrete stress measurement point data to obtain a three-dimensional stress field distribution; Calculate the stress gradient for each voxel point in the three-dimensional stress field distribution and identify the voxel positions corresponding to high stress gradients to obtain a high stress gradient area, and perform spatial clustering and merging connection of adjacent high stress gradient voxels on the high stress gradient area to obtain a stress concentration area sensitive to cracks; Based on preset frequency band parameters, perform dual-frequency modulation excitation and multi-physical field synchronous detection on the stress concentration region to obtain synchronous response signals of multiple physical fields, and perform phase synchronization calculation on the periods of the synchronous response signals to obtain multi-physical field response signals.

6. The crack detection method for precision automotive parts according to claim 5, wherein, Based on the preset voxel space, calculate the stress values of the voxel points inside the automotive parts for the discrete stress measurement point data to obtain a three-dimensional stress field distribution, including: Perform voxelized mesh division on the three-dimensional space of the automotive parts to obtain a voxel space coordinate system, and calculate the distances from each voxel point in the voxel space coordinate system to multiple stress measurement points to obtain a voxel-measurement point distance relationship table; Perform radial basis function transformation on the distance values in the voxel-measurement point distance relationship table to obtain a voxel interpolation weight matrix, and perform weighted product calculation on the voxel interpolation weight matrix and the stress values of the corresponding discrete stress measurement points in the discrete stress measurement point data to obtain the voxel stress contribution values of each voxel point; Perform summation calculation on the voxel stress contribution values of each to obtain voxel interpolation stress values, and based on the voxel space coordinate system, perform spatial mapping combination on the voxel interpolation stress values to obtain a three-dimensional stress field distribution.

7. The crack detection method for precision automotive parts according to claim 1, characterized in that, Perform multi-parameter fusion calculation on the multi-physical field response signals to obtain crack geometric parameters, and perform multi-scale wavelet decomposition on the acoustic components in the multi-physical field response signals to obtain acoustic time-frequency characteristics, including: Extract the numerical values of the acoustic modulation depth, electromagnetic response amplitude, and thermal infrared gradient anomaly value in the multi-physical field response signals to obtain the response intensity parameters corresponding to the three types of physical fields, and perform wavelet decomposition on the acoustic components in the multi-physical field response signals for multi-time-frequency characteristic parameters to obtain acoustic time-frequency characteristics, and perform weighted fusion calculation on the response intensity parameters for each stress concentration region to obtain a comprehensive response intensity distribution; Based on the comprehensive response intensity distribution, identify the positions of the response intensity peak points in each stress concentration region to obtain the coordinates of the crack center candidate points, and based on a preset intensity boundary threshold, perform threshold segmentation on the coordinates of the crack center candidate points to obtain at least one crack space boundary contour; Perform measurement calculations of various geometric parameters on each crack space boundary contour to obtain crack geometric parameters.

8. The crack detection method for precision automotive parts according to claim 1, characterized in that, Perform feature similarity matching and extended prediction calculation on the crack feature classification data to generate crack detection results, including: Perform crack feature clustering on the crack feature classification data to obtain crack feature category labels, and perform risk score calculation on the crack feature category labels to obtain the crack risk levels of each category label; Based on a preset crack historical case library, perform historical case matching on the crack feature classification data to obtain a reference case set, and based on the historical extension records corresponding to the reference case set, perform extension rate calculation on the crack feature category labels to generate a crack propagation prediction model; Based on the time series parameters corresponding to the crack propagation prediction model, perform evolution prediction calculation on the crack risk levels for the future crack size to generate crack detection results.

9. A crack detection device for precision automotive parts, characterized in that, The crack detection device for precision automotive parts includes: A pre-scanning module, which is used to perform ultrasonic multi-dimensional pre-scanning on the automotive parts to be tested, obtain acoustic reflection data, and identify the abnormal acoustic impedance regions in the acoustic reflection data to obtain abnormal region distribution data; An intelligent grasping module, which is used to calculate the influence range of the abnormal regions on the automotive parts to be tested and classify the sensitive thresholds based on the abnormal region distribution data, obtain the grasping sensitivity partition, and allocate the grasping force and apply stress loading to the grasping sensitivity partition to obtain stress-acoustic response time series data; A stress excitation module, which is used to reconstruct the three-dimensional stress field of the automotive parts to be tested from the stress-acoustic response time series data, obtain the stress concentration regions, and perform dual-frequency stress modulation excitation and multi-physical field synchronous detection on the stress concentration regions to obtain multi-physical field response signals; A feature fusion module, which is used to perform multi-parameter fusion calculation on the multi-physical field response signals to obtain crack geometric parameters, perform multi-scale wavelet decomposition on the acoustic components in the multi-physical field response signals to obtain acoustic time-frequency features, and splice the feature dimensions of the crack geometric parameters and the acoustic time-frequency features to obtain crack feature classification data; An intelligent diagnosis module, which is used to perform feature similarity matching and extended prediction calculation on the crack feature classification data to generate crack detection results.

Images