Steel ring quality detection working method capable of rapidly judging defect spots

Abstract

The invention provides a steel ring quality detection working method capable of quickly judging defect spots, which comprises the following steps of: S1, scanning a complete steel ring by using a contour scanner, processing a scanned image by using a remote intelligent terminal after the scanning is completed, and exporting a file in an STL (Standard Template Library) format; s2, using analysis software PolyWorks to import the CAD model and the triangulation model obtained by scanning into the software, and performing coordinate positioning; s3, selecting a coordinate system according to the design requirement of the steel ring, creating a coordinate axis, setting attributes, determining an original point and a reference element of a workpiece, establishing a feature coordinate system, and mapping the feature coordinate system to a workpiece coordinate system; and S4, executing a steel ring detection generation instruction, and storing and sharing the output feature points and surfaces through a text display format file.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of hub detection, and in particular to a steel ring quality detection method for quickly judging defect points. BACKGROUND

[0002] In the field of industrial production and manufacturing, steel rings, as a common and key component, are widely used in many industries such as automobiles and machinery, and their quality directly affects the performance and safety of the entire product. Therefore, it is crucial to accurately and efficiently detect various parameters of the steel ring. The traditional method of detecting steel rings mainly relies on manual measurement using a caliper and other measuring tools. This method not only has low efficiency, consumes a large amount of manpower and time, but also has difficulty in ensuring measurement accuracy and is easily affected by human factors, resulting in large errors in measurement results. In particular, for the key parameter of the out-of-roundness of the steel ring, the traditional manual detection method is almost unable to accurately obtain it, making it difficult to meet the strict requirements of modern industrial production for product quality control.

[0003] With the continuous development of technology, although some detection equipment and methods have appeared on the market, there are still many shortcomings. For example, some scanning detection techniques require pre-pasting of marker points on the surface of the steel ring, which not only increases the complexity of the operation process and reduces the detection efficiency, but also may affect the scanning accuracy due to inaccurate pasting positions of the marker points. Moreover, some detection techniques take a long time to complete a complete scan, which cannot meet the demand for rapid detection in large-scale production. In the prior art, for example, CN117058151A, a hub detection method and system based on image analysis, in this scheme, the hub qualification evaluation is carried out through complex image operation, which requires a large amount of system overhead and operation process, the process is complex and redundant, and is not conducive to batch detection operation in large-scale industrial production process, which urgently needs technical personnel in the field to solve the corresponding technical problems. SUMMARY

[0004] The present application aims to at least solve the technical problems existing in the prior art, and particularly innovatively proposes a steel ring quality detection method for quickly judging defect points.

[0005] In order to achieve the above-mentioned purpose of the present application, the present application provides a steel ring quality detection method for quickly judging defect points, comprising the following steps:

[0006] S1, using a contour scanner to scan a complete steel ring, and after scanning, processing the scanned image by a remote intelligent terminal and exporting an STL format file;

[0007] S2, using analysis software PolyWorks to import the CAD model and the triangulation model obtained by scanning into the software, and performing coordinate positioning;

[0008] S3, select a coordinate system according to the design requirements of the rim, create coordinate axes and set attributes, determine the workpiece origin and reference elements, establish a feature coordinate system and map it to the workpiece coordinate system;

[0009] S4, execute the rim detection generation instruction, and output the feature points and surfaces through a text display format file for saving and sharing.

[0010] The above technical solution is preferably, the S2 includes:

[0011] S2-1, create x, y, z three coordinate axes, and set the length, unit and direction;

[0012] S2-2, scan the rim point cloud data P={p i |p i ∈R 3}, extract the dominant direction through covariance matrix analysis.

[0013]

[0014] Wherein n is a positive integer, p i is any point cloud data, is the mean value of the point cloud data, and T is the transpose,

[0015] First, establish the rim coordinate axis parameters (o, u x , u y , u z ), and locate the origin.

[0016] The above technical solution is preferably, the S2 further includes:

[0017] S2-3, combine the geometric characteristics of the rim with the physical center of gravity to locate the origin o,

[0018] Calculate the geometric center of gravity of the rim

[0019] Calculate the physical center of gravity of the rim Where ρ is the density of the rim,

[0020] Calculate the positioning origin of the fusion strategy

[0021] Where ω geo is the geometric center of gravity adjustment weight of the rim, ω phy is the physical center of gravity adjustment weight of the rim.

[0022] The above technical solution is preferably, the S2 further includes:

[0023] S2-4, construct a dynamic feature coordinate system mapping engine, automatically generate a transformation matrix of the feature coordinate system to the rim workpiece coordinate system by calculating the relative position relationship between features in real time,

[0024] Set the steel ring scanning direction as If the angle between the coordinate system of the steel ring workpiece and the z-axis is θ, then the transformation matrix is:

[0025]

[0026] The transformation matrix U is used in the ICP fitting process and corrects the orientation of the steel ring scanning data.

[0027] In a preferred embodiment of the above technical solution, step S3 includes:

[0028] S3-1, Input model voxelization V during CNN initialization CAD With scanning voxelization V Scan Output the initial transformation matrix U init ;

[0029] Develop an enhanced fitting engine based on the Iterative Closest Point (ICP) algorithm, and introduce normal vector weights and curvature weights to improve the fitting accuracy of thin-walled or high-curvature steel rings.

[0030] The objective function of ICP is:

[0031]

[0032] This formula calculates and achieves high-precision matching between scanned data and CAD models, and after optimization, outputs a transformation matrix U. opt The U opt From the initial U init Through gradual optimization, high-precision alignment of steel ring scanning data with CAD digital model was ultimately achieved.

[0033] In a preferred embodiment of the above technical solution, step S3 further includes:

[0034] S3-2, calculating the point cloud data of the steel ring scan. With digital model point cloud data Local deviations in steel rim data Perform standardized calculations:

[0035] Where μ d The average value of the local deviations represents the deviation d between all steel ring scanning points and the CAD module points. i The arithmetic mean reflects the central tendency of the overall deviation, σ. d The local deviation standard deviation is used to measure the total d. i Relative to the average value μ d The degree of dispersion reflects the range of fluctuation of the deviation;

[0036] Set threshold Z thres =3, thus marking Zi Z thres The point set S flow is recommended for point pair calibration.

[0037] The preferred technical solution is that the S3 further comprises:

[0038] S3-3, by constructing a graph G, the local defect point pairs are associated as a global matching network, solving the problem that the ICP algorithm is easy to fall into local optimization.

[0039] The defect point set S flow ={p i |Z i > Z thres}

[0040] Construct a node set V, and pair each steel ring scanning defect point flow in S with the corresponding nearest neighbor point in the CAD model to form a node Construct an edge set E, and add edges according to the compatibility between nodes, and the edge weight is the matching quality of the point pair (v ij ,v kl );

[0041] Construct a graph G=(V,E), where the node V is the steel ring scanning point and the model point pair, and the edge weight is

[0042] Where θ ij is the angle between the normal vectors, κ i and κ j are the curvatures, α is the normal consistency weight coefficient, which controls the proportion of the cosine value of the normal angle in the edge weight, the greater the value, the more important the normal direction consistency on the matching quality, β is the curvature similarity weight coefficient, which controls the proportion of the curvature difference item in the edge weight, the greater the value, the more important the curvature similarity on the matching quality, γ is the curvature difference attenuation coefficient, which adjusts the punishment degree of the curvature difference |κ i -κ j | in the edge weight, |κ i -κ j | by calculating the curvature difference between the steel ring scanning defect point and the corresponding nearest neighbor point in the CAD model, to ensure that the matching point pair has similar local geometric features in the steel ring.

[0043] Use the Hungarian algorithm to solve the maximum weight matching, input the calibration point pair into the ICP module, and update U opt .

[0044] Preferably, the S3 further comprises:

[0045] S3-4, mapping the deviation data to the voxel grid, and the color value C(x, y, z) is calculated by bilinear interpolation:

[0046] Wherein dmin is the global minimum deviation value dmax is the global maximum deviation value, the extreme value extraction is performed on the heat map data, d i is the local deviation value of the steel ring scanning point and the CAD model point.

[0047] Preferably, the S4 comprises:

[0048] For the deviation sequence {d t}, the window W size (such as the pattern hole area W) is defined, and the extreme value detection formula

[0049] If marked as out-of-tolerance, the text data report is output according to the extreme value data, and the qualified judgment data is generated in combination with the steel ring design tolerance standard threshold.

[0050] As described above, due to the adoption of the above technical solutions, the present application has the following beneficial effects:

[0051] The steel ring quality detection method realizes significant improvement in detection accuracy, efficiency and applicability through multi-dimensional technical innovation, and has the following beneficial effects: the profile scanner and the remote intelligent terminal are used to work cooperatively to realize an automatic processing flow from data acquisition to STL file generation, thereby greatly shortening the pre-processing time. The dynamic feature coordinate system mapping engine is combined to correct the scanning data direction in real time, reduce the need for manual intervention, and improve the detection efficiency. The weighted positioning strategy of geometric gravity center and physical gravity center is innovatively integrated, the dominant direction is analyzed through the covariance matrix, the limitations of single gravity center positioning are overcome, and the origin positioning accuracy is significantly improved, especially in complex curved steel rings, thereby providing a reliable reference for subsequent matching. The normal vector weight (cos theta) and the curvature weight are introduced in the standard ICP target function, which is adapted to the geometric characteristics of thin-walled or large-curvature steel rings, effectively solves the problem that the traditional algorithm is easily trapped in local optimum in complex curved surfaces, and realizes high-precision alignment (deviation control reaches microns) of the scanning data and the CAD model. The flaw point matching graph G is constructed, the node compatibility edge weight is designed, the Hungarian algorithm is used to solve the maximum weight matching, the local flaws are associated as a global network, the local convergence defect of the traditional ICP is avoided, and the flaw recognition rate of the complex structure steel ring is improved. The extreme value extraction algorithm is combined to intuitively display the local deviation distribution. The detection result quantization and rapid qualification are realized, and the detection needs of different specifications of products are adapted. It is convenient to integrate into an automatic production line, and reduces the technical transformation cost.

[0052] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and the attendant drawings or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0053] The foregoing and / or additional aspects and advantages of the present application are achieved by providing a method for detecting the quality of a steel ring, comprising the steps of:

[0054] Figure 1 is a schematic diagram of the present application;

[0055] Figure 2 is a schematic diagram of the STL export file of the present application;

[0056] Figure 3 is a schematic diagram of the initial scanning of the present application;

[0057] Figure 4 is a schematic diagram of the side plane scanning of the present application;

[0058] Figure 5 is a schematic diagram of the edge scanning of the present application;

[0059] Figure 6 is a schematic diagram of the comparison of the size of the steel ring workpiece of the present application;

[0060] Figure 7 is another schematic diagram of the comparison of the size of the steel ring workpiece of the present application;

[0061] Figure 8 is a generated text file of the present application. DETAILED DESCRIPTION

[0062] Embodiments of the present application are described in detail below with reference to the attached drawings, which show by way of example, embodiments in which the same or similar elements or elements having the same or similar functions are denoted by the same reference numerals throughout the drawings. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present application, and should not be construed as limiting the present application.

[0063] As shown in Figures 1 to 8 , the present application discloses a method for quickly judging the quality of a steel ring, comprising the following steps:

[0064] S1, using a contour scanner to scan a complete steel ring, the innovation point is that no mark point is needed in the scanning process, and it can be completed in 2 minutes, with a precision of 0.05mm. After scanning, the scanning image is processed by a remote intelligent terminal, and an STL format file is exported. The remote intelligent terminal referred to here is preferably a notebook computer and a data server connected to a remote network.

[0065] S2, use analysis software PolyWorks to import CAD models (such as.STEP,.igs format) and scanned triangulation models (.stl format) into the software, and perform coordinate positioning;

[0066] S3, select the coordinate system according to the design requirements of the rim, create coordinate axes and set properties, determine the origin of the workpiece and the reference element, establish the feature coordinate system and map it to the workpiece coordinate system;

[0067] S4, execute the rim detection generation instruction, and output the feature points and surfaces in table form to present the measurement results, and save and share the text display format file through text display. The text display methods include word, Excel, PDF, WPS, etc.

[0068] Create x, y, z three coordinate axes, and set length, unit and direction properties. Determine the origin position of the rim workpiece, and the reference elements related to the coordinate axes, such as holes and bosses, and add them to the rim workpiece as the origin. Measure the features through convolutional neural network, select the execution deviation vector instruction for the inner diameter, outer diameter and pattern hole features of the rim, then get the minimum deviation and maximum deviation, and view the maximum and minimum values of the inner diameter, outer diameter and pattern hole curved surface analysis results. Select the feature object through CAD model alignment, get the rim features through detection instruction, match the rim properties, and judge the extreme value data by identifying the minimum difference and maximum deviation value of the rim feature error.

[0069] The S2 includes:

[0070] 1. Create x, y, z three coordinate axes, and set length, unit and direction properties

[0071] Develop an adaptive coordinate axis generation algorithm based on the geometric features of the rim, automatically determine the coordinate axis direction by analyzing the symmetry axis (such as the center axis of the rim) and the maximum curvature direction of the scanned data. For example, use principal component analysis (PCA) to reduce the dimensionality of point cloud data and extract the dominant direction as the coordinate axis reference.

[0072] Scan the point cloud data P = {p i |p i ∈R 3} of the rim, and extract the dominant direction through covariance matrix analysis.

[0073]

[0074] Decompose the covariance matrix, take the eigenvector corresponding to the maximum eigenvalue as the z-direction of the coordinate axis (the center axis of the rim), and take the x-direction as the second largest eigenvector. Construct a right-handed coordinate system, where n is a positive integer, p i is any point cloud data, T is the transpose, and μ is the mean of the point cloud data.

[0075] A dynamic unit length calibration mechanism is introduced to automatically adjust the coordinate axis scale interval by combining the design size (such as the nominal diameter) of the rim with the scanning data range, ensuring that the display precision matches the measurement range.

[0076] First, establish the rim coordinate axis parameters (o, u x ,u y ,u z ), and perform origin positioning.

[0077] Combine the geometric characteristics of the rim with the physical barycenter to position the origin o,

[0078] Calculate the geometric barycenter of the rim

[0079] Then calculate the physical barycenter of the rim where ρ is the density of the rim,

[0080] Calculate the positioning origin of the fusion strategy

[0081] where ω geo is the geometric barycenter adjustment weight of the rim, ω phy is the physical barycenter adjustment weight of the rim, and the optimal empirical values are 0.6 and 0.4, respectively.

[0082] By integrating inertial measurement unit (IMU) sensor data, the coordinate axis direction offset caused by device vibration during scanning is corrected, improving the stability of the coordinate system.

[0083] 2. Determine the origin position of the rim workpiece and the reference elements (such as holes, bosses) related to the coordinate axis, and add them to the workpiece as the origin,

[0084] Design a multi-modal reference element fusion algorithm to simultaneously extract the geometric features (such as hole edge profiles, boss vertexes) and physical features (such as barycenter simulation results) on the surface of the rim. For example, by fitting the hole center points using the least squares method and combining the mass distribution model to calculate the theoretical barycenter, the origin position is cross-verified.

[0085] Develop an intelligent origin positioning module for non-symmetrical structure rims, using the shortest path algorithm in graph theory to find the optimal origin candidate point in complex surfaces (such as rims with uneven pattern hole distribution), ensuring global consistency of the coordinate system.

[0086] Introduce laser radar point cloud density analysis to preferentially select high-density areas (such as planes) as reference elements, improving the anti-noise ability of origin positioning.

[0087] 3. For the features on the rim workpiece, establish their feature coordinate system using the corresponding reference elements and map them into the rim workpiece coordinate system

[0088] A dynamic feature coordinate system mapping engine is constructed to automatically generate the transformation matrix from the feature coordinate system to the rim workpiece coordinate system by real-time calculation of the relative position relationship between features (such as the angle between the normal vector of a plane and the axial direction of a cylinder).

[0089] The scanning direction of the rim is set to be The angle between the z-axis of the rim workpiece coordinate system and the scanning direction is θ, and the transformation matrix is:

[0090]

[0091] The transformation matrix U is used in the ICP fitting process, and the direction of the rim scanning data is corrected.

[0092] A feature constraint propagation algorithm is developed to use the coordinate system parameters of key features (such as the main plane) as constraint conditions to force the association of secondary features (such as holes on the plane), ensuring that the position relationship between features meets the design tolerance.

[0093] A GPU-accelerated coordinate transformation calculation module is integrated to achieve millisecond-level mapping parameter updates and support real-time dynamic adjustment.

[0094] The above technical solution is preferably, the S3 includes:

[0095] Select the best fitting measurement object, align the scanned data with the rim-entered numerical model data, and if the effect is not good, transfer the point pair detection to the global best fitting state; select the corresponding measurement feature creation option through the instruction, select the measurement feature through the directory tree, view the maximum and minimum values of the surface analysis result, and view the extreme value of the feature object.

[0096] 4. Select the best fitting measurement object, use the scanned data to align the numerical model through the overall best fitting method

[0097] An overall fitting initialization module is designed to use a convolutional neural network (CNN) to predict the initial alignment parameters (such as rotation angle, translation) of the scanned data and the CAD numerical model. The network input is the low-resolution voxel representation of the numerical model and the scanned data, and the output is the initial transformation matrix.

[0098] During the CNN initialization process, the numerical model is voxelized as V CAD , and the scanned data is voxelized as V Scan , and the initial transformation matrix U init is output.

[0099] Develop a reinforced fitting engine based on the Iterative Closest Point (ICP) algorithm, introduce normal vector weight and curvature weight, improve the fitting accuracy of thin-walled or large-curvature steel rings. For example, give higher weight to the edge area of the steel ring to avoid the deviation of the plane area fitting dominating the global result.

[0100] The ICP objective function is:

[0101]

[0102] The formula calculates the high-precision matching of the scan data and the CAD model, and outputs the transformation matrix U after optimization opt , U opt is obtained from the initial U init , which is gradually optimized to achieve high-precision alignment of the steel ring scan data and the CAD model.

[0103] The normal vector term punishes the difference in surface direction, ensuring that the steel ring scan data and the CAD model have the same orientation in the local area; the curvature term punishes the difference in the bending degree of geometric features, improving the fitting ability of complex surfaces (such as steel ring pattern holes).

[0104] Through dynamic weighting (ωn, ωc), the algorithm balances between global alignment (normal vector dominant) and local details (curvature dominant), and finally achieves sub-millimeter (0.03mm) registration accuracy.

[0105] where ω m is the normal vector weight, preferably 0.7, ω c is the curvature weight, preferably 0.3, is the unit normal vector of the CAD model at the i-th point cloud data, used to measure the consistency of the surface direction, is the unit normal vector of the steel ring scan data after coordinate transformation U at the i-th point cloud data, used to compare with the CAD model normal vector to optimize the registration accuracy, is the average curvature value of the CAD model at the i-th point cloud data, used to constrain the matching of local feature sets of the steel ring workpiece, is the curvature value of the steel ring scan data point cloud after coordinate transformation U at the i-th point cloud data, compared with the curvature value of the CAD model to optimize the fitting effect of thin-walled or complex surfaces,

[0106] Integrate a multi-resolution fitting strategy, first align the overall contour with low precision and high speed, then refine the local features with high precision, balance efficiency and accuracy.

[0107] 5. If the overall best fitting has feature defects, change the alignment mode to point-to-point, select the correct point of the feature defect for calibration, and detect the overall best fitting state

[0108] Develop intelligent flaw point recognition algorithm, by comparing the local curvature of scanning data and digital model, the difference of normal vector, automatically mark the area whose deviation exceeds the threshold value (such as burr, depression). Use Z-score standardization method to calculate the deviation value of each point, and screen out abnormal points.

[0109] By calculating the local deviation of the scanning point cloud data of the steel ring and the digital model point cloud data of the steel ring data Standardization calculation:

[0110] Where μ d is the average value of local deviation, which represents the arithmetic mean of the deviation d i of all steel ring scanning points and CAD model points, reflecting the concentration trend of overall deviation, and σ d is the standard deviation of local deviation, which is used to measure the dispersion degree of all d i relative to the average value μ d , reflecting the fluctuation range of deviation

[0111] Set the threshold value Z thres = 3, so that the points with Z i > Z thres are marked as flaws, and the flaw point set S flow is recommended for point pair calibration.

[0112] Design point pair calibration recommendation system, based on the maximum weight matching algorithm in graph theory, recommend the optimal calibration point pair between scanning data and digital model, reduce the subjectivity of manual selection. The recommendation includes point position normal vector consistency, curvature similarity, etc.

[0113] Introduce constraint satisfaction problem (CSP) solver, after selecting the calibration point pair, automatically generate the fitting parameters that meet all the constraints, avoid local optimal solution.

[0114] 6. Select the object to be measured (such as plane, cylinder), select the corresponding measurement steel ring feature creation option, create elements > points first, then create elements > straight lines.

[0115] Develop feature automatic recognition and recommendation system, based on the pre-trained graph neural network (GNN) of steel ring CAD model, automatically recognize the key features to be measured (such as pattern hole, cylindrical surface) and recommend the creation order. The network input is the B-rep representation of the CAD model, and the output is the feature type and dependency relationship.

[0116] By constructing graph G, the local flaw point pair is associated as a global matching network, solving the problem that ICP algorithm is easy to fall into local optimal solution.

[0117] Flaw point set Sflow ={p i |Z i >Z thres}

[0118] Construct a node set V, and then... flow Each steel ring scans for defects. Nearest neighbor corresponding to the CAD model Pairing is performed to form nodes. Construct an edge set E, adding edges based on the compatibility between nodes, with edge weights equal to the weights of the node pairs (v). ij ,v kl The quality of the match.

[0119] Construct a graph G = (V, E), where node V is a pair of scanning points on the steel ring and points on the digital model, and the edge weights are...

[0120] Where θ ij κ is the angle between the normal vectors. i and κ j For curvature, calculate the matching pairs. α is the normal vector consistency weight coefficient, controlling the proportion of the cosine of the angle between normal vectors in the edge weights. The larger the value, the more important the consistency of normal vector direction is to the matching quality. β is the curvature similarity weight coefficient, controlling the curvature difference term. The larger the value of γ in the edge weight, the more important the influence of curvature similarity on matching quality. γ is the curvature difference attenuation coefficient, which adjusts the curvature difference |κ. i -κ j |The penalty strength in the edge weight,|κ i -κ j |Calculate the defects in the steel rim scan Nearest neighbor corresponding to the CAD model The curvature differences ensure that the matching point pairs have similar local geometric features in the steel ring.

[0121] The Hungarian algorithm is used to solve for the maximum weight matching. The calibration point pairs are input into the ICP module to update U. opt .

[0122] Design a rule-based feature creation engine, integrate an expert system knowledge base, and define priority rules for feature creation (e.g., planes first, then holes). For example, for measuring the inner diameter of a steel ring, it is mandatory to create the inner diameter edge points first, and then fit the cylinder axis.

[0123] An interactive guided interface is introduced to highlight candidate areas and prompt users to make selections when there is ambiguity in feature creation (such as overlapping planes), balancing automation and flexibility.

[0124] 7. Select the measurement feature through the directory tree, execute the deviation vector instruction for the inner diameter, outer diameter, and pattern hole features of the rim, and then obtain the minimum deviation and the maximum deviation. Check the maximum and minimum values of the inner diameter, outer diameter, and pattern hole curved surface analysis results

[0125] Develop a 3D deviation heat map generation algorithm to map the minimum / maximum deviation data to the surface of the rim 3D model. Use voxelization technology to convert point cloud data into a regular grid, and then generate a continuous color gradient (red-yellow-green) through a color interpolation algorithm (such as bilinear interpolation) to visually display the deviation degree in each area.

[0126] Design a deviation vector visualization engine to support dynamic profile analysis. Users can select to cut the model radially or axially along the rim to view the internal deviation distribution, which helps locate the out-of-roundness or uneven thickness issues.

[0127] Map the deviation data to the voxel grid, and the color value C(x, y, z) is calculated by bilinear interpolation:

[0128] Where dmin is the global minimum deviation value, i.e., the minimum distance between all rim scan points and the CAD model, and dmax is the global maximum deviation value, i.e., the maximum distance between all rim scan points and the CAD model. The color gradient is from red (Cmax) to yellow to green (Cmin), and the extreme value of the heat map data is extracted.

[0129] d i is the local deviation value between the rim scan point and the CAD model point,

[0130] Integrate the statistical process control (SPC) module to automatically calculate the mean and standard deviation of the deviation data and generate control charts to monitor the stability of the rim quality in real time.

[0131] 8. Select the feature object in the directory tree, obtain the rim feature by detecting the instruction, match the rim attributes, and determine the extreme value data by identifying the minimum and maximum deviation values of the rim feature error.

[0132] For the deviation sequence {d t}, define the window W size (such as the pattern hole area W), and the extreme value detection formula is

[0133] If is marked as out-of-tolerance, output the data report based on the extreme value data, and generate qualified judgment data based on the rim design tolerance standard threshold.

[0134] Develop a dynamic threshold self-adaptive analysis algorithm to automatically set the deviation allowed range based on the rim design tolerance. For example, for the inner diameter out-of-roundness, the threshold is set to ±0.05mm of the design diameter, and the out-of-tolerance area is automatically marked in red.

[0135] The extreme value data intelligent extraction module is designed to detect the minimum / maximum value in the deviation data stream in real time by using a sliding window algorithm, and record the spatial position. The window size can be dynamically adjusted according to the characteristic size of the steel ring (such as reducing the window in the pattern hole area).

[0136] The natural language generation (NLG) technology is integrated to automatically convert the extreme value data, the out-of-tolerance position and the qualified judgment conclusion into a structured report text, support multi-language output (such as Chinese / English), and improve the readability of the report.

[0137] The technical scheme of the present application forms a set of high-precision, high-automation and visual steel ring quality detection technical scheme through the whole process of coordinate system establishment, data alignment, feature measurement and deviation analysis, solves the core pain points of low efficiency, poor precision and inability to quantify the out-of-roundness of traditional manual detection, shortens the detection cycle from 30 minutes to 5 minutes, stabilizes the precision within 0.05 mm, and significantly improves the steel ring quality control capability in industrial production.

[0138] Although the embodiments of the present application have been shown and described, those skilled in the art can understand that various changes, modifications, replacements and variations can be made to these embodiments without departing from the principles and purposes of the present application, and the scope of the present application is defined by the claims and their equivalents.

Claims

1. A steel ring quality detection work method for quickly determining a defect point, characterized by, It comprises the following steps: S1, scanning the complete steel ring using a profile scanner, and after scanning, processing the scanned image by a remote intelligent terminal to export an STL format file; S2, using analysis software PolyWorks to import the CAD model and the triangulated model obtained by scanning into the software to perform coordinate positioning; S3, selecting a coordinate system according to the design requirements of the steel ring, creating coordinate axes and setting attributes, determining the workpiece origin and reference elements, establishing a feature coordinate system and mapping it to the workpiece coordinate system; S4, executing the steel ring detection generation instruction, and saving and sharing the output feature points and surfaces through a text display format file.

2. The steel ring quality inspection work method for quickly determining a defective point according to claim 1, characterized by, The S2 comprises: S2-1, creating x, y and z coordinate axes and setting the length, unit and direction; S2-2, scan the ring point cloud data P = {p i |p i ∈R 3}, extract the dominant direction by covariance matrix analysis. where n is a positive integer, p i is a point cloud data, is a point cloud data mean, T is a transpose, First, the parameters of the coordinate axis of the steel ring (o, u x ,u y ,u z ) are established to perform origin positioning.

3. The steel ring quality inspection work method for quickly determining a defective point according to claim 2, characterized in that, The S2 further comprises: S2-3, positioning the origin o in combination with the geometric characteristics and physical center of gravity of the steel ring, Calculating the geometric center of a rim Recalculating the physical center of gravity of the rim where p is the density of the rim, calculating the positioning origin of the fusion strategy where ω geo is the rim geometry adjustment weight, ω phy is the rim physical center of gravity adjustment weight.

4. The steel ring quality inspection work method for quickly determining a defective point according to claim 2, characterized in that, The S2 further comprises: S2-4, constructing a dynamic feature coordinate system mapping engine, automatically generating a transformation matrix of the feature coordinate system to the steel ring workpiece coordinate system by real-time calculation of the relative position relationship between features, The scanning direction of the steel ring is set as The angle between the z-axis of the steel ring workpiece coordinate system and the z-axis of the world coordinate system is θ, and the transformation matrix is: wherein the transformation matrix U is used for the ICP fitting process, and the direction of the steel ring scanning data is corrected.

5. The steel ring quality inspection work method for quickly determining a defective point according to claim 1, characterized in that, The S3 comprises: S3-1, input number model voxelization V in CNN initialization process CAD with scanning voxelization V Scan , output initial transformation matrix U init ; Developing a reinforced fitting engine based on the iterative closest point (ICP) algorithm, introducing normal vector weight and curvature weight to improve the fitting accuracy of thin-walled or large-curvature steel rings; The ICP objective function is: The formula calculates the high-precision matching of the scanning data and the CAD model, and outputs the transformation matrix U after optimization opt The U opt From the initial U init Step-by-step optimization, and finally realize the high-precision alignment of the rim scanning data and the CAD model.

6. The steel ring quality inspection work method for quickly determining a defective point according to claim 5, characterized by, The S3 further comprises: S3-2, by calculating the rim scan point cloud data with digital point cloud data of the rim data local deviation standardized calculation: where μ d is the average of the local deviations, d i is the arithmetic mean of all the steel ring scan point deviations from the CAD modulus points, d d is the standard deviation of the local deviations, d i is the dispersion of the deviations from the average μ d reflects the range of fluctuations of the deviations; Setting a threshold value Z thres = 3, so that the mark Z i > Z thres The point is a defect, and the defect point set S flow The point-to-point calibration recommendation is performed.

7. The steel ring quality inspection work method for quickly determining a defective point according to claim 6, characterized by, The S3 further comprises: S3-3, by constructing a graph G, the local defect point pairs are associated as a global matching network to solve the problem that the ICP algorithm is easy to fall into local optimization. Set of defect points S flow = {p i | Z i > Z thres} Construct node set V, and S flow scan each ring for defect points find the nearest neighbor in the CAD model pair them up to form nodes Construct edge set E, and add edges according to compatibility between nodes, with edge weights being the matching quality of the point pair (v ij ,v kl ) Construct a graph G=(V,E), the node V is the steel ring scanning point and the number model point pair, and the edge weight is where θ ij is the angle between the normal vectors, κ i and κ j are the curvatures, α is the normal consistency weight coefficient, controlling the proportion of the cosine value of the angle between the normal vectors in the edge weight, the larger the value, the more important the normal direction consistency to the matching quality, β is the curvature similarity weight coefficient, controlling the proportion of the curvature difference term in the edge weight, the larger the value, the more important the curvature similarity to the matching quality, γ is the curvature difference attenuation coefficient, adjusting the punishment degree of the curvature difference |κ i - κ j | in the edge weight, |κ i - κ j | by calculating the curvature difference between the scan defect point of the rim and the corresponding nearest neighbor point in the CAD model ensures the similarity of the local geometric features of the matching point pair in the rim.​ The Hungarian algorithm is used to solve the maximum weight matching. The pairs of landmarks are input into the ICP module, which updates U opt .

8. The rapid judgment of the defect point of the steel ring quality detection work method according to claim 7, characterized in that, The S3 further comprises: S3-4, mapping the deviation data to the voxel grid, and the color value C(x,y,z) is calculated by bilinear interpolation: where dmin is the global minimum deviation value dmax is the global maximum deviation value, the heat map data is subjected to extreme value extraction, d i is the local deviation value of the steel ring scanning point and the CAD model point.

9. The rapid judgment of the defect point of the steel ring quality detection work method according to claim 1, characterized in that, The S4 comprises: For the bias sequence {d t}, define the window W size (e.g., the pattern hole area W), the extreme value detection formula If If marked as out of tolerance, output a text data report based on the extreme value data, combined with the rim design tolerance standard threshold, to generate qualified judgment data.

Images