Online detection method for straightness of forged round steel based on point cloud data

Abstract

The present application belongs to the technical field of industrial detection, and particularly relates to a forging round steel straightness online detection method based on point cloud data, which comprises the following steps: firstly, obtaining original point cloud and multi-dimensional physical characteristic data of the round steel in a transmission state; then, slicing the point cloud, and combining local geometric missing features and thermal deformation driving features to calculate a compensation step, and correcting a primary centroid to obtain a real three-dimensional physical axis; subsequently, projecting the axis to an orthogonal plane, and using a restriction relationship between local curvature and material bending stiffness to obtain a pseudo-boundary penalty factor, and then correcting an envelope angle and calculating a comprehensive straightness error. The present application eliminates the interference of point cloud missing, thermal eccentricity and surface microscopic defects, and improves the accuracy of the forging round steel straightness detection.

Description

Technical Field

[0001] This invention relates to the field of industrial inspection technology. More specifically, this invention relates to an online method for detecting the straightness of forged round steel bars based on point cloud data. Background Technology

[0002] Forged round steel is a key basic material in heavy machinery, automobile manufacturing and other fields. Its straightness index is directly related to the accuracy of subsequent machining and the mechanical properties of finished components. With the evolution of industrial automation, traditional offline sampling inspection methods have been gradually phased out due to their low efficiency. Machine vision technology based on 3D point cloud scanning has gradually become the mainstream evolution direction for online inspection of the straightness of round steel on the production line.

[0003] In actual high-temperature forging production lines, round steel bars have just been removed from the heat source and are in a state of high-speed continuous transmission. At this time, due to the limited visual perspective caused by the layout of the on-site inspection equipment, and the physical obstruction of the bottom of the round steel bars by the bottom conveyor rollers, the point cloud of the round steel cross-section obtained by the scanning equipment often has certain local missing points and isolated noise.

[0004] Existing 3D point cloud processing technologies generally rely on curvature fitting of pure geometry to search for the overall bending boundary of round steel when calculating straightness. However, the surface of round steel after high-temperature forging usually retains residual dead weld oxide scale, which forms tiny random protrusions on the steel surface. When existing algorithms use point cloud curvature abrupt changes to search for bending extreme points, the geometric algorithm parameters are easily intercepted by these microscopic surface oxide scale defects. This causes the system to incorrectly judge the tiny dead weld protrusions on the surface as a huge thermal bending of the entire round steel in the macroscopic structure, thus outputting distorted straightness error parameters. Once this distorted parameter is fed back, it will directly cause the subsequent automatic straightening equipment to issue incorrect pressure roller pressure commands, which may result in substandard straightening of the round steel or, in severe cases, excessive pressure from the equipment that crushes the material, causing production line abnormalities. Summary of the Invention

[0005] To address the aforementioned technical problem of insufficient accuracy in straightness detection of forged round steel, this invention provides an online method for straightness detection of forged round steel based on point cloud data, comprising: Point cloud data of forged round steel is acquired, and the point cloud data is sliced ​​to obtain multiple slice point clouds. The primary centroid coordinates of each slice point cloud are obtained. Based on the local geometric missing features and thermal deformation driving features of each slice point cloud, the cross-sectional center compensation step size of each slice point cloud is obtained, which is used to compensate and correct the primary centroid coordinates of the corresponding slice point cloud to obtain the three-dimensional physical axis of the forged round steel. Based on the geometric contour of the three-dimensional physical axis of the forged round steel, the straightness detection result of the forged round steel is obtained.

[0006] In the high-temperature forging production line scenario, this invention calculates the cross-sectional center compensation step size by combining local geometric missing features and thermal deformation driving features to correct the primary centroid point. It combines the micro-contour missing ratio with the physical thermal eccentricity for dynamic compensation, thereby restoring the true and smooth three-dimensional spatial physical axis of the forged round steel under high-speed continuous transmission, and improving the physical reference accuracy of subsequent straightness detection.

[0007] Preferably, the step of obtaining the cross-sectional center compensation step size of each slice point aggregation includes: The extreme value of thermal deformation eccentricity is determined based on the ratio of the temperature difference between the surface of the forged round steel and the environment to the thermal conduction dissipation coefficient of the forged round steel. The weight of the extreme value of thermal deformation eccentricity is determined in response to the degree of local geometric loss of each slice point cluster. The compensation step size of the cross-sectional center of each slice point cluster is obtained.

[0008] This invention establishes the extreme value of thermal deformation eccentricity by calculating the temperature difference between the surface of the forged round steel and the environment and the heat conduction dissipation coefficient, and uses the degree of local geometric loss as the basis for weight allocation, so that the calculation of the center compensation step can more realistically reflect the theoretical offset of the physical center of gravity caused by uneven heating, and eliminates the static calculation error caused by single pure geometric contour fitting when facing high temperature asymmetric thermal expansion.

[0009] Preferred options also include: Extract the normal variance of the fitting boundary of each slice point set. The ratio of the weight of the extreme value of thermal deformation eccentricity to the normal variance of the fitting boundary is the linear amplification factor. Obtain the cross-sectional center compensation step size of each slice point cluster, including obtaining the product of the extreme value of thermal deformation eccentricity and the corresponding linear amplification factor, to obtain the cross-sectional center compensation step size of each slice point cluster.

[0010] In harsh industrial environments where actual scanning presents significant noise interference, this invention utilizes the normal variance of the point cloud fitting boundary distance as the basis for adjusting the linear amplification factor. This not only ensures that the compensation algorithm takes into account the combined effects of geometric loss and thermal eccentricity, but also automatically amplifies and corrects the compensation amount based on the local variance caused by point cloud clutter. Even in scenarios with sparse local point clouds and harsh conditions, the robustness and accuracy of the axis compensation step size calculation are still guaranteed.

[0011] Preferably, the method for determining the degree of local geometric loss includes: Extract the maximum and minimum values ​​of the polar coordinate angles of each slice point cloud set, and use the difference between the maximum and minimum values ​​of the polar coordinate angles as the effective envelope angle radians of the corresponding slice point cloud set. The ratio of the effective envelope angle radian to twice the constant of pi is used as the degree of local geometric missingness of the corresponding slice point cluster.

[0012] This invention utilizes the ratio of the effective envelope angle radian extracted under polar coordinates to the complete pi constant to calculate the geometric missing ratio of the dimensionless cross section. This avoids the high computational cost of complex three-dimensional surface reconstruction and ensures the timeliness of calculation for real-time tracking and compensation of geometric missing conditions of round steel in continuous transmission on high-temperature production lines.

[0013] Preferably, obtaining the straightness test result of the forged round steel includes: The three-dimensional physical axes of the forged round steel are projected onto the first orthogonal plane and the second orthogonal plane to obtain the first plane axis point set and the second plane axis point set; Envelope extremum detection is performed on the first plane axis point set and the second plane axis point set respectively to obtain the envelope extremum of the first orthogonal plane and the second orthogonal plane; The square root of the sum of the squares of the envelope extrema of the first orthogonal plane and the second orthogonal plane is used as the straightness test result of the forged round steel.

[0014] This invention reduces the dimension of the three-dimensional physical axis and projects it onto the first and second orthogonal planes, then performs independent two-dimensional envelope extremum detection. It also merges the two-dimensional plane extremum space into a three-dimensional comprehensive macroscopic curvature through simple square root operations. This simplifies the complexity of blindly searching for the overall curvature boundary in a large point cloud and improves the online generation rate of the three-dimensional comprehensive straightness error assessment while meeting stringent detection accuracy requirements.

[0015] Preferably, performing envelope extremum detection on the first plane axis point set and the second plane axis point set respectively includes: False boundary detection is performed sequentially on each axis point of the first plane axis point set and each axis point of the second plane axis point set to obtain the false boundary penalty factor for the corresponding axis point. The envelope extrema are obtained by traversing and searching each axis point of the first plane axis point set and each axis point of the second plane axis point set. In response to the pseudo-boundary penalty factor of the corresponding axis point being greater than a set threshold, the corresponding axis point is ignored.

[0016] Preferably, the step of sequentially performing pseudo-boundary detection on each axis point of the first plane axis point set and each axis point of the second plane axis point set includes: Extract the local curvature abrupt change features of each axis point to obtain the bending stiffness of the forged round steel. Determine the pseudo-boundary penalty factor of the corresponding axis point based on the ratio of the local curvature abrupt change features of the corresponding axis point to the bending stiffness of the forged round steel.

[0017] Preferably, determining the pseudo-boundary penalty factor for the corresponding axis point based on the ratio of the local curvature abrupt change characteristic of the corresponding axis point to the bending stiffness of the forged round steel includes: The ratio of the local curvature abrupt change characteristic at the corresponding axis point to the bending stiffness of the forged round steel is used as the pseudo-boundary penalty factor.

[0018] Preferably, the traversal search of each axis point in the first and second planar axis point sets includes: Traverse the first plane axis point set and the second plane axis point set along the set direction; False boundary detection is performed on each axis point in the traversal process in turn to obtain the false boundary penalty factor for the corresponding axis point; The envelope extrema are obtained by traversing and searching each axis point of the first plane axis point set and each axis point of the second plane axis point set. In response to the pseudo-boundary penalty factor of the corresponding axis point being greater than a set threshold, the corresponding axis point is ignored.

[0019] Preferably, it further includes: performing reverse mapping on the straightness test results of the forged round steel to obtain straightening process adjustment parameters, which are used to perform online closed-loop straightening on the forged round steel.

[0020] The beneficial effects of this invention are as follows: (1) This invention integrates geometric missing rate and thermal eccentricity physical model to dynamically extract center compensation step size and restore the three-dimensional real physical axis of round steel; (2) The present invention introduces a pseudo-boundary penalty mechanism based on the true bending stiffness limit of the material, and uses physical laws to completely filter out the distortion misjudgment interference caused by the micro-protrusion of the surface dead weld oxide scale. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating the online detection method for the straightness of forged round steel based on point cloud data in this invention; Figure 2 It is a schematic diagram showing the comparison of physical axes in three-dimensional space. Detailed Implementation

[0022] This invention discloses an online method for detecting the straightness of forged round steel bars based on point cloud data, referring to... Figure 1 This includes steps S1-S4: S1: Obtain multi-dimensional data of forged round steel. The specific steps are: obtain the original three-dimensional point cloud data, surface temperature data, ambient temperature data, material process parameters, and roller conveyor speed of the forged round steel.

[0023] It should be noted that in order to achieve accurate online detection of the straightness of forged round steel and eliminate physical errors in subsequent algorithm processing, it is necessary to collect comprehensive basic data covering the surface morphology of forged round steel, external environmental conditions, the physical properties of the material itself, and the dynamic characteristics of the production line.

[0024] Specifically, the process involves acquiring the original three-dimensional point cloud data, surface temperature data, ambient temperature data, material process parameters, and roller conveyor speed of the forged round steel, including: By using multi-vision laser scanners pre-deployed around the forging production line, the forged round steel in the transmission state is dynamically scanned to obtain the original three-dimensional point cloud data of the forged round steel.

[0025] The infrared thermal imager components deployed on-site read the surface temperature data of the forged round steel and simultaneously read the reference temperature of the factory building away from the heat source radiation area. The reference temperature of the factory building is used as the ambient temperature data of the forged round steel.

[0026] The process database of the underlying production control system is retrieved to read the material process parameters corresponding to the forged round steel. The material process parameters include the heat dissipation coefficient, nominal radius, high-temperature elastic modulus, and moment of inertia of the cross section.

[0027] The system synchronously calls the dynamic frequency converter data of the underlying production control system to obtain the roller conveyor speed carrying the forged round steel.

[0028] At this point, all basic test data and physical parameters have been obtained.

[0029] S2: Obtain the three-dimensional physical axis of the forged round steel. The specific steps are as follows: Based on the spatial distance distribution of the original three-dimensional point cloud data of the forged round steel, obtain the effective outer contour point cloud of the forged round steel; spatially cut the effective outer contour point cloud of the forged round steel to obtain multiple slice point clouds; based on the local geometric missing features and thermal deformation driving features of each slice point cloud, obtain the cross-sectional center compensation step size of each slice point cloud; based on the cross-sectional center compensation step size of each slice point cloud, compensate and correct the coordinates of the primary centroid point of the corresponding slice point cloud to obtain the three-dimensional physical axis of the forged round steel.

[0030] It should be noted that the raw 3D point cloud data acquired in industrial settings is substantial and contains isolated noise points caused by metal splashes or dust, requiring noise reduction. Furthermore, poor viewing angles and obstruction from the underlying roller conveyor can cause incomplete cross-sectional profiles of the forged round steel. This, coupled with the microscopic thermal expansion and eccentricity phenomenon caused by the asymmetric heating of the forged round steel, results in the center of the directly fitted cross-section deviating significantly from the true physical axis. Therefore, this invention assesses the proportion of cross-sectional incompleteness caused by geometric defects and extracts comprehensive feature indices based on the thermal eccentricity effect driven by temperature difference. These indices are then used to calculate the comprehensive compensation step size, thereby obtaining the true and smooth spatial axis of the forged round steel.

[0031] Specifically, based on the spatial distance distribution of the original three-dimensional point cloud data of the forged round steel, the effective outer contour point cloud of the forged round steel is obtained, including: An adaptive Morton code is used to perform 3D indexing and sorting of the original 3D point cloud data of forged round steel according to the spatial bit interleaving order, thus establishing a fast spatial dictionary of the original 3D point cloud data. It should be noted that the conventional calculation method of Morton code dimensionality reduction is to convert the 3D spatial coordinate data into binary representation and arrange them bit by bit, so that physically adjacent data points are synchronously adjacent in the memory index, thereby improving the nearest neighbor search speed of the point cloud.

[0032] Based on the resolution parameters of the multi-vision laser scanner and the empirical distribution patterns of typical splash noise volumes, a preset threshold for the number of nearest neighbors to search and the standard deviation multiple is established. For example, the number of nearest neighbors to search is set to 50, and the standard deviation multiple threshold is set to 1. It should be noted that if the number of nearest neighbors to search is too large, such as 200, it will cause severe timeouts and smooth out local features on the surface of the forged round steel; if the number of nearest neighbors to search is too small, such as 5, the statistical characteristics will be unclear, and dense noise clusters cannot be identified. Similarly, if the standard deviation multiple threshold is set too high, real splash noise will be preserved; if the standard deviation multiple threshold is set too low, normal micro-protrusion textures on the surface of the forged round steel will be mistakenly deleted.

[0033] Based on the spatial fast dictionary of the original 3D point cloud data, the system searches for the nearest neighbor points of each target observation point in the original 3D point cloud data by a preset number of nearest neighbor search points, and calculates the average Euclidean distance from each target observation point to all its corresponding nearest neighbor points as the average spatial distance of each target observation point.

[0034] The mean of the average spatial distance of all target observation points is calculated as the global average spatial distance; the standard deviation of the average spatial distance of all target observation points is calculated as the global standard deviation; target observation points whose average spatial distance is greater than the product of the global average spatial distance plus a standard deviation multiple threshold and the global standard deviation are identified as free splash noise and removed; the point cloud retained after removing splash noise is taken as the effective outer contour point cloud of the forged round steel.

[0035] Preferably, the effective outer contour point cloud of the forged round steel is spatially cut to obtain multiple slice point clouds; based on the local geometric missing features and thermal deformation driving features of each slice point cloud, the cross-sectional center compensation step size of each slice point cloud is obtained, including: Based on the balance between the spatial frequency characteristics of local bending of forged round steel and the upper limit of the equipment's computing power, a fixed slicing step size is preset for point cloud segmentation. For example, the fixed slicing step size is set to 10 mm. It should be noted that if the fixed slicing step size is too large, such as 50 mm, the spatial sampling will be too sparse, missing local bending deformations of the forged round steel; if the fixed slicing step size is too small, such as 1 mm, the number of slices will increase exponentially, dragging down the real-time computing performance of online detection.

[0036] The effective outer contour point cloud of the forged round steel is cut into multiple independent slice point cloud sets along the main axis of the point cloud length with a fixed slice step distance.

[0037] For each slice point cloud, a random sampling consensus algorithm is applied to fit the circumcircle, obtaining the coordinates of the primary centroid of each slice point cloud. During this fitting process, the variance statistics of the distances from all retained points within each slice point cloud to the boundary of the fitted circumcircle are extracted, and these variance statistics are used as the normal variance of the fitted boundary of the corresponding slice point cloud.

[0038] For each slice point cloud, polar coordinates are converted, and the maximum and minimum polar angles of the point cloud within each slice point cloud are extracted. The difference between the maximum and minimum polar angles of each slice point cloud is calculated as the effective envelope angle in radians for the corresponding slice point cloud. It should be noted that the effective envelope angle in radians of each slice point cloud reflects the local geometric missing features of each slice point cloud.

[0039] The surface temperature data of the forged round steel is subtracted from the ambient temperature data of the forged round steel to obtain the temperature difference between the surface of the forged round steel and the environment.

[0040] It should be noted that in existing technologies, the centroid extraction of point clouds typically relies directly on pure geometric fitting algorithms. However, in the forging process, harsh visual perspectives can lead to incomplete point cloud contours, and pure geometric fitting can cause a false shift of the primary centroid towards the remaining side of the point cloud. Furthermore, the large temperature difference between the surface of forged round steel and the environment during high-temperature forging results in non-uniform heat conduction, causing thermal deformation and eccentricity of the steel's true physical center of gravity. Single geometric fitting ignores these errors. Therefore, this invention combines the proportion of missing contours from pure geometry with the physical quantity of thermal eccentricity caused by surface temperature differences. Within a small deformation range, the more severe the geometric loss and the larger the fitting error, the more amplified the interference of thermal eccentricity on centroid calculation.

[0041] The cross-sectional center compensation step size of any set of slice points satisfies the expression: ; In the formula, This represents the compensation step size of the cross-section center where the j-th slice points are clustered; Pi is a constant. This represents the effective envelope angle in radians of the j-th slice point cluster; This indicates the temperature difference between the surface of the forged round steel and the ambient temperature. This represents the heat dissipation coefficient of forged round steel. This represents the normal variance of the fitted boundary of the j-th slice point set; The nominal radius is used to represent the material process parameters of forged round steel.

[0042] In the formula, This represents the dimensionless percentage of missing sections quantified using the laws of pure geometric space. This represents the extreme value of thermal deformation eccentricity constructed based on the steady-state heat conduction law. Since the dimension of the heat conduction dissipation coefficient is ℃ / mm, the temperature dimension is eliminated by dividing the surface temperature difference with the heat conduction dissipation coefficient, and it is transformed into the basic dimension characterizing the radial physical eccentricity depth, millimeters. This represents the dimensionless error proportion achieved by dividing the variance of the fitted boundary normal, which has a dimension of square millimeters, by the square of the nominal radius. This represents the linear amplification factor constructed based on the error positive compensation control law.

[0043] By using the proportion of missing sections as a dimensionless proportional base, multiplying it by the extreme value of thermal deformation eccentricity with millimeter dimensions, we obtain a comprehensive compensation base. This indicates that the more severe the geometric contour incompleteness of the j-th slice point cloud, the greater the error impact exposed by the internal thermal eccentricity. Then, multiplying by a linear amplification factor containing a constant 1, we further superimpose the variance amplification effect caused by the cluttered point cloud onto the comprehensive compensation base, ultimately obtaining a cross-section center compensation step size with precisely unified millimeter dimensions. If the effective wrap angle radian approaches 2×PI, the compensation step size is set to zero.

[0044] Preferably, the coordinates of the primary centroid points of the corresponding slice point cluster are compensated and corrected based on the cross-sectional center compensation step size of each slice point cluster to obtain the three-dimensional physical axis of the forged round steel, including: It should be noted that, since the absence of local point clouds can cause the coordinates of the primary centroid points fitted by machine vision to be falsely shifted to the remaining side where the point cloud is dense, this invention uses the cross-section center compensation step size to push the coordinates of the primary centroid points in the opposite direction of the missing parts, and at the same time superimposes the thermal eccentricity reverse callback caused by the temperature gradient to restore the true physical centroid position.

[0045] Calculate the unit vector in the opposite direction of the centroid of the geometrically missing part in each slice point cloud; apply the cross-sectional center compensation step size of each slice point cloud along the unit vector in the opposite direction to the primary centroid coordinates of each slice point cloud to complete the coordinate translation correction and obtain the corrected centroid coordinates of each slice point cloud.

[0046] According to the axial spatial position order of the slice point cloud on the effective outer contour point cloud, the coordinates of the corrected centroid points corresponding to all slice point clouds are connected in sequence to obtain the three-dimensional physical axis of the forged round steel composed of line segments.

[0047] It should be noted that, as Figure 2This is a comparison diagram of the physical axes in three-dimensional space; the curves represent the true theoretical axes of the forged round steel, the dashed lines represent the fitted axes formed by connecting the primary centroids of each slice, and the broken lines represent the three-dimensional physical axes of the forged round steel obtained after compensation and correction.

[0048] Thus, the three-dimensional physical axis of the forged round steel was obtained.

[0049] S3: Obtain the minimum containment region value. The specific steps are as follows: Project the coordinates of each node of the three-dimensional physical axis of the forged round steel onto the first orthogonal plane and the second orthogonal plane respectively to obtain the ordered evaluation point set of the first plane and the ordered evaluation point set of the second plane; According to the relative constraint relationship between the local curvature change characteristics and bending stiffness characteristics of each axis node in the ordered evaluation point set of the first plane and the ordered evaluation point set of the second plane, obtain the pseudo-boundary penalty factor of each axis node in the corresponding plane respectively; Use the pseudo-boundary penalty factor of each axis node to attenuate and correct the original envelope angle of each axis node in the corresponding plane to obtain the minimum containment region value of the first orthogonal plane and the second orthogonal plane.

[0050] It should be noted that when using the envelope algorithm to search for the overall bending boundary of forged round steel, the algorithm is easily intercepted by tiny protrusions formed by residual dead weld oxide scale on the surface of the forged round steel, leading to the misjudgment of microscopic surface defects as global macroscopic thermal bending. Since forged round steel possesses ultimate bending stiffness, its actual macroscopic bending exhibits gentle, low-frequency characteristics. Therefore, this invention extracts the degree of curvature abrupt change in extreme value determination and evaluates the constraint ratio of local curvature abrupt changes relative to the material's inherent bending stiffness. By constructing a penalty factor to suppress the determination weights of high-frequency singularities that violate the material's stiffness limit, it prevents spurious distortions from occurring in the solution of the minimum containment region value.

[0051] Specifically, the coordinates of each node of the three-dimensional physical axis of the forged round steel are projected onto the first orthogonal plane and the second orthogonal plane respectively to obtain the ordered evaluation point set of the first plane and the ordered evaluation point set of the second plane, including: The coordinates of each node of the three-dimensional physical axis of the forged round steel are projected vertically onto the first orthogonal plane and the second orthogonal plane respectively, generating the first plane ordered evaluation point set corresponding to the first orthogonal plane and the second plane ordered evaluation point set corresponding to the second orthogonal plane.

[0052] Calculate the product of the difference vectors of adjacent axis nodes in the first and second ordered evaluation point sets respectively, and obtain the local curvature abrupt change characteristic value of each axis node in the first and second ordered evaluation point sets.

[0053] Preferably, based on the relative constraint relationship between the local curvature abrupt change characteristics and bending stiffness characteristics of each axis node in the first and second plane ordered evaluation point sets, the pseudo-boundary penalty factors for each axis node in the corresponding plane are obtained, including: An environmental scaling parameter is preset based on the high-temperature plastic rheological threshold of the forged round steel material, and a fixed sampling feature step size is preset based on the average spacing of discrete nodes on the three-dimensional physical axis. For example, the environmental scaling parameter is set to 0.1, and the fixed sampling feature step size is set to 5 mm. It should be noted that if the environmental scaling parameter is too large, such as 1.5, the system will over-penalize any tiny curvature fluctuations, causing the true bending points of the forged round steel to be unnecessarily suppressed; if the environmental scaling parameter is extremely small, such as 0.001, the penalty term is almost ineffective, completely failing to filter out false boundaries caused by oxide scale protrusions.

[0054] It should be noted that existing point cloud processing techniques generally rely on geometric curvature to determine inflection points when distinguishing between genuine round steel bending and pseudo-protrusions caused by residual oxide scale on the surface. However, this method, which is detached from the physical entity, easily misjudges minute residual oxide scale as macroscopic structural bending, leading to incorrect extraction of extreme values. This invention introduces the physical limit of bending resistance of materials. In actual physical laws, the bending stiffness of forged round steel is determined by the product of its cross-sectional moment of inertia and elastic modulus. Therefore, this invention uses the measured local abnormal curvature value as the numerator and adds the local abnormal curvature value to the material's true bending stiffness characteristic as the denominator to construct the pseudo-boundary penalty factor for the axial nodes.

[0055] The pseudo-boundary penalty factor for any axis node within an ordered set of evaluation points in an arbitrary plane satisfies the expression: ; In the formula, Represents the m-th ordered evaluation point set in the plane. The pseudo-boundary penalty factor for each axis node; Indicates the environment scaling parameter; Represents the m-th ordered evaluation point set in the plane. Local curvature abrupt change characteristic values ​​of each axis node; Indicates a fixed sampling feature step size; This indicates the high-temperature elastic modulus of forged round steel. The moment of inertia of the cross section of the forged round steel bar; This indicates the roller conveyor speed for forging round steel.

[0056] In the formula, This represents the physical energy intensity of a local curvature change, which is the product of the curvature change and the step size. The physical stiffness limit of forged round steel against bending deformation under high temperature motion was evaluated. The ratio of the relative constraint relationship between the local curvature mutation feature and the bending stiffness feature is formed. When an isolated sharp protrusion appears on the k-th axis node in the m-th ordered evaluation point set due to residual oxide scale, the measured curvature mutation intensity increases in a pulse-like manner, causing the value representing the bending limit of the material in the denominator to be completely submerged. The calculated ratio approaches the upper limit of 1, indicating that the point violates the physical bending limit of the material and is judged as a false boundary. Conversely, in real smooth bending deformation, the curvature mutation is extremely low, the denominator is dominated by the bending limit of the material, and the penalty factor smoothly approaches 0.

[0057] Preferably, the original envelope angle of each axis node in the corresponding plane is attenuated and corrected using the pseudo-boundary penalty factor of each axis node to obtain the minimum enclosing region value between the first orthogonal plane and the second orthogonal plane, including: It should be noted that, based on the gain regulation law in signal control systems, the general approach is to directly delete outliers, but this would disrupt the spatial continuity of the axis point cloud sequence. This invention utilizes a multiplication mechanism to convert the penalty factor into a negative feedback control gain, performing a simple proportional scaling back on spurious angles that physically exceed limits. This allows the envelope search path to smoothly avoid false boundaries and closely adhere to the forged round steel body.

[0058] In the mth orthogonal plane, a one-way search is performed on the ordered evaluation point set of the mth plane. The arctangent function is used to calculate the angle of each axis node in the ordered evaluation point set of the mth plane toward the adjacent node behind it, which is used as the original envelope angle of each axis node in the ordered evaluation point set of the mth plane.

[0059] Based on the convergence boundary conditions of the robust iterative control law, a preset envelope correction penalty weight coefficient is used. For example, the envelope correction penalty weight coefficient is set to 0.85. It should be noted that the envelope correction penalty weight coefficient is used to limit the maximum attenuation of a single angle correction, preventing excessive numerical reduction from causing the envelope line to flow backward and resulting in iterative deadlock.

[0060] The corrected envelope angle of any axis node within an ordered set of evaluation points in any plane satisfies the expression: ; In the formula, This represents the corrected envelope angle of the k-th axis node within the m-th ordered evaluation point set in the plane; This represents the original envelope angle of the k-th axis node within the m-th ordered evaluation point set in the plane; This represents the envelope correction penalty weighting coefficient; It represents the pseudo-boundary penalty factor of the k-th axis node within the m-th ordered evaluation point set of the plane.

[0061] In the formula, This represents the decay feedback term controlled by applying a basic constant weight. This represents the scaling attenuation ratio obtained by subtracting the attenuation feedback term from the base quantity 1; The physical significance of the multiplication method lies in the closed-loop gain feedback mechanism. It uses the calculated scaling and attenuation coefficient to proportionally multiply and reduce the absolute value of the original envelope angle of the k-th axis node within the m-th ordered evaluation point set on the plane, thus reducing the strain-corrected envelope angle due to system stiffness limitations. Based on the soft elimination rule of local angle maxima, if the penalty factor approaches 1, the envelope angle is forcibly reduced, successfully avoiding the forced pull and misleading effect of pseudo-boundary convex points on the entire envelope boundary ray.

[0062] By using the corrected envelope angle of each axis node in the ordered evaluation point set of the m-th plane, the corresponding original envelope angle is replaced, and the envelope extremum extension is constructed in the m-th orthogonal plane.

[0063] In the m-th orthogonal plane, extract the left envelope extreme point with the largest corrected envelope angle and the right envelope extreme point with the smallest corrected envelope angle from the ordered evaluation point set of the m-th plane.

[0064] Calculate the maximum vertical distance from the left and right envelope extreme points of the m-th orthogonal plane to the opposite side of the line connecting them. Use the maximum vertical distance of the m-th orthogonal plane as the minimum bounding region value corresponding to the m-th orthogonal plane.

[0065] Thus, the minimum containment region value of each orthogonal plane has been obtained.

[0066] S4: Complete the online detection closed-loop control of forged round steel. The specific steps are as follows: Based on the spatial geometric synthesis relationship between the minimum containment area value corresponding to the first orthogonal plane and the minimum containment area value corresponding to the second orthogonal plane, obtain the three-dimensional spatial comprehensive straightness error of the forged round steel; use the three-dimensional spatial comprehensive straightness error of the forged round steel to adjust the subsequent straightening process instructions, and complete the online detection closed-loop control of the forged round steel.

[0067] It should be noted that the independent region values ​​extracted through biorthogonal plane projection and anti-high-frequency singularity envelope search have filtered out the interference of the microscopic morphology of the forged round steel surface. By using the geometric synthesis law of spatial vectors to backtrack and fuse the minimum containment region values ​​of the first and second orthogonal planes, the true macroscopic bending state of the three-dimensional solid of the forged round steel can be reflected, and it can be directly output as a correction parameter to the control level.

[0068] Specifically, based on the spatial geometric synthesis relationship between the minimum bounding area values ​​corresponding to the first orthogonal plane and the minimum bounding area values ​​corresponding to the second orthogonal plane, the three-dimensional spatial comprehensive straightness error of the forged round steel is obtained, including: Extract the minimum bounding region value corresponding to the first orthogonal plane and the minimum bounding region value corresponding to the second orthogonal plane after the calculation is completed.

[0069] Using the composition rule of spatial geometric vectors, the square data of the minimum bounding region value corresponding to the first orthogonal plane and the square data of the minimum bounding region value corresponding to the second orthogonal plane are calculated respectively.

[0070] The square of the minimum bounding area value corresponding to the first orthogonal plane is added to the square of the minimum bounding area value corresponding to the second orthogonal plane, and the square root operation is performed to obtain the Euclidean distance scalar value. The Euclidean distance scalar value is directly used as the three-dimensional spatial comprehensive straightness error of the forged round steel.

[0071] Preferably, the three-dimensional spatial comprehensive straightness error of the forged round steel is used to adjust the instructions for the subsequent straightening process, thereby completing the online detection closed-loop control of the forged round steel, including: It should be noted that obtaining the three-dimensional spatial comprehensive straightness error only completes the state assessment stage. In order to realize the actual production value of industrial testing data, the three-dimensional spatial comprehensive straightness error must be converted into the motion control parameters of hardware equipment to achieve hardware synchronization of the production line.

[0072] The three-dimensional spatial comprehensive straightness error of the forged round steel is converted into adaptive correction and compensation parameters for the servo action of the straightening equipment according to the preset hardware mapping rule.

[0073] The adaptive correction and compensation parameters are incorporated into the process action instructions of the main controller of the downstream straightening equipment. The main controller dynamically adjusts the pressure amount of the pressure rollers and the roller ring spacing of the downstream straightening equipment, thereby completing the online detection closed-loop control of the forged round steel.

[0074] Thus, the online detection and control of straightness of forged round steel based on point cloud data was completed.

[0075] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for online detection of straightness of forged round steel based on point cloud data, characterized in that, include: Obtain point cloud data of forged round steel, slice the point cloud data to obtain multiple slice point cloud sets, and obtain the primary centroid coordinates of each slice point cloud set. Based on the local geometric missing features and thermal deformation driving features of each slice point cluster, the cross-sectional center compensation step size of each slice point cluster is obtained, which is used to compensate and correct the primary centroid coordinates of the corresponding slice point cluster, and obtain the three-dimensional physical axis of the forged round steel. Based on the geometric contour of the three-dimensional physical axis of the forged round steel, the straightness test results of the forged round steel are obtained.

2. The online detection method for straightness of forged round steel based on point cloud data according to claim 1, characterized in that, The step size for obtaining the cross-sectional center compensation of each slice point set includes: The extreme value of thermal deformation eccentricity is determined based on the ratio of the temperature difference between the surface of the forged round steel and the environment to the thermal conduction dissipation coefficient of the forged round steel. The weight of the extreme value of thermal deformation eccentricity is determined in response to the degree of local geometric loss of each slice point cluster. The compensation step size of the cross-sectional center of each slice point cluster is obtained.

3. The online detection method for the straightness of forged round steel based on point cloud data according to claim 2, characterized in that, Also includes: Extract the normal variance of the fitting boundary of each slice point set. The ratio of the weight of the extreme value of thermal deformation eccentricity to the normal variance of the fitting boundary is the linear amplification factor. Obtain the cross-sectional center compensation step size of each slice point cluster, including obtaining the product of the extreme value of thermal deformation eccentricity and the corresponding linear amplification factor, to obtain the cross-sectional center compensation step size of each slice point cluster.

4. The online detection method for straightness of forged round steel based on point cloud data according to claim 2, characterized in that, The method for determining the degree of local geometric missingness includes: Extract the maximum and minimum polar coordinate angles of the point cloud of each slice point set, and use the difference between the maximum and minimum polar coordinate angles as the effective envelope angle radians of the corresponding slice point set; The ratio of the effective envelope angle radian to twice the constant of pi is used as the degree of local geometric missingness of the corresponding slice point cluster.

5. The online detection method for the straightness of forged round steel based on point cloud data according to claim 1, characterized in that, The process of obtaining the straightness test results of the forged round steel includes: The three-dimensional physical axes of the forged round steel are projected onto the first orthogonal plane and the second orthogonal plane to obtain the first plane axis point set and the second plane axis point set; Envelope extremum detection is performed on the first plane axis point set and the second plane axis point set respectively to obtain the envelope extremum of the first orthogonal plane and the second orthogonal plane; The square root of the sum of the squares of the envelope extrema of the first orthogonal plane and the second orthogonal plane is used as the straightness test result of the forged round steel.

6. The online detection method for straightness of forged round steel based on point cloud data according to claim 5, characterized in that, The process of performing envelope extremum detection on the first plane axis point set and the second plane axis point set respectively includes: False boundary detection is performed sequentially on each axis point of the first plane axis point set and each axis point of the second plane axis point set to obtain the false boundary penalty factor for the corresponding axis point. The envelope extrema are obtained by traversing and searching each axis point of the first plane axis point set and each axis point of the second plane axis point set. In response to the pseudo-boundary penalty factor of the corresponding axis point being greater than a set threshold, the corresponding axis point is ignored.

7. The online detection method for straightness of forged round steel based on point cloud data according to claim 6, characterized in that, The step of sequentially performing pseudo-boundary detection on each axis point of the first plane axis point set and each axis point of the second plane axis point set includes: Extract the local curvature abrupt change features of each axis point to obtain the bending stiffness of the forged round steel. Determine the pseudo-boundary penalty factor of the corresponding axis point based on the ratio of the local curvature abrupt change features of the corresponding axis point to the bending stiffness of the forged round steel.

8. The online detection method for straightness of forged round steel based on point cloud data according to claim 7, characterized in that, The determination of the pseudo-boundary penalty factor for the corresponding axis point based on the ratio of the local curvature abrupt change characteristics of the corresponding axis point to the bending stiffness of the forged round steel includes: The ratio of the local curvature abrupt change characteristic at the corresponding axis point to the bending stiffness of the forged round steel is used as the pseudo-boundary penalty factor.

9. The online detection method for straightness of forged round steel based on point cloud data according to claim 6, characterized in that, The first and second plane axis point sets are traversed and searched for each axis point, including: Traverse the first plane axis point set and the second plane axis point set along the set direction; False boundary detection is performed on each axis point in the traversal process in turn to obtain the false boundary penalty factor for the corresponding axis point; The envelope extrema are obtained by traversing and searching each axis point of the first plane axis point set and each axis point of the second plane axis point set. In response to the pseudo-boundary penalty factor of the corresponding axis point being greater than a set threshold, the corresponding axis point is ignored.

10. The online detection method for the straightness of forged round steel based on point cloud data according to claim 1, characterized in that, Also includes: The straightness test results of forged round steel are reverse mapped to obtain straightening process adjustment parameters, which are used to perform online closed-loop straightening of forged round steel.

Images