Aluminum alloy die casting deformation amount detection system

Abstract

This invention discloses a deformation detection system for aluminum alloy die castings, belonging to the field of aluminum alloy die casting quality inspection technology. The system includes a data acquisition module that acquires continuous contour data of the surface of the part under test along multiple preset scanning paths, including multiple sampling points and spatial coordinates; a point cloud construction module that constructs an initial three-dimensional point cloud model based on the coordinates and performs surface fitting to obtain the fitted surface equation; a deviation marking module that calculates the normal deviation distance from each sampling point to the fitted surface and marks those exceeding a threshold as deformation candidate points; a region extraction module that determines the deformation concentration region based on the spatial distribution density of the candidate points and extracts deformation detection nodes; a curvature comparison module that obtains the curvature change value corresponding to the node and compares it with a preset range; if the value exceeds the range, a result recording module records the spatial coordinates and timestamp of the node, generating a deformation detection result. This invention achieves automated detection and precise positioning of surface deformation in die castings, improving detection efficiency and reliability.

Description

Technical Field

[0001] This invention relates to the field of quality inspection technology for aluminum alloy die castings, specifically to a deformation detection system for aluminum alloy die castings. Background Technology

[0002] During the cooling process of aluminum alloy die castings, the surface and internal structure undergo varying degrees of deformation due to factors such as thermal stress release and material shrinkage. This deformation directly affects the assembly accuracy and structural strength of the parts. Existing methods for detecting deformation in aluminum alloy die castings mainly include contact coordinate measuring machine (CMM) and standard template-based comparative inspection. CMM obtains discrete spatial coordinates by contacting the probe point by point, while the standard template comparative method uses physical or digital templates to analyze deviations in key sections of the die casting. However, these methods obtain the linear distance deviation of sparse sampling points relative to the ideal model, failing to characterize the continuous distribution of the deformation area on the curved surface. This leads to a complete reliance on interpolation to infer deformation at non-detection points, resulting in missed detections and misjudgments. Contact measurement's point placement strategy in complex curved surface areas is limited by mechanical accessibility, resulting in low efficiency of dense sampling, and the probe contact force may introduce micro-deformation, affecting measurement accuracy. The standard template method is highly dependent on the manufacturing accuracy of the template and its registration with the part under test, making it difficult to adapt to the flexible inspection needs of multiple varieties and varying batches.

[0003] The problem this solution aims to solve is how to accurately determine the location area where deformation is concentrated and extract effective detection nodes after acquiring surface contour data, and how to combine surface curvature features to quantify and determine the degree of deformation. Summary of the Invention

[0004] The aluminum alloy die casting deformation detection system disclosed in this invention aims to solve the problem that existing detection methods are difficult to accurately locate the deformation concentration area and incorporate curvature change into the deformation degree judgment criteria. It proposes a detection scheme that extracts deformation detection nodes from surface contour data according to normal deviation and density distribution, and judges deformation based on curvature change, so as to realize the detection of deformation distribution and severity on the surface of die castings.

[0005] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a deformation detection system for aluminum alloy die castings, the system comprising a data acquisition module, a point cloud construction module, a deviation marking module, a region extraction module, a curvature comparison module, and a result recording module.

[0006] The data acquisition module is used to acquire continuous surface contour data generated when the surface of the aluminum alloy die casting under test moves along multiple preset scanning paths. The continuous surface contour data includes multiple sampling points and their corresponding spatial coordinates. To improve scanning efficiency while also considering the ability to capture local details, the preferred method for generating the preset scanning paths is as follows: Based on the three-dimensional design model of the die casting, a surface curvature distribution map is extracted, and the surface is divided into high-curvature and low-curvature regions. In the high-curvature regions, a cross-grid path formed by longitudinal and transverse scanning lines is laid out to increase the scanning density. In the low-curvature regions, parallel scanning lines extending along the length of the workpiece are laid out, thereby improving the point cloud acquisition quality of deformation-sensitive areas while ensuring detection efficiency. During the acquisition process, a laser displacement sensor moves along the preset path and acquires the original distance values ​​of each sampling point according to a preset sampling frequency. Simultaneously, the real-time attitude angle and motion trajectory coordinates of the sensor are acquired. The relative coordinates of the sampling points are converted to the global coordinate system, and after stitching, high-precision continuous surface contour data covering the entire surface of the workpiece is obtained.

[0007] The point cloud construction module constructs an initial 3D point cloud model based on the spatial coordinates of multiple sampling points, and performs surface fitting on the initial 3D point cloud model to obtain the fitted surface equation. In a preferred embodiment, the module first performs noise filtering on the sampling points in the continuous surface contour data, removing outlier sampling points that deviate from the neighborhood average coordinates by more than a preset multiple. Then, it uses voxel downsampling to obtain a uniformly distributed set of downsampled points. Subsequently, it constructs a local fitting plane based on the moving least squares method, and redirects the point cloud normal vectors according to the consistency of the local plane normal vectors. An implicit surface reconstruction algorithm is then used to generate a smooth initial 3D point cloud model. Based on this, the 3D point cloud is parameterized and mapped to a 2D parameter domain, and a bicubic B-spline surface approximation algorithm is used to perform surface fitting. The obtained fitted surface equation can accurately characterize the ideal geometric shape of the die casting, providing a reliable comparison benchmark for deformation deviation analysis.

[0008] The deviation marking module calculates the normal deviation distance from each sampling point in the initial 3D point cloud model to the fitted surface based on the fitted surface equation, and marks sampling points whose normal deviation distance exceeds a preset deviation threshold as deformation candidate points. Specifically, this module calculates the nearest projection point of each sampling point on the fitted surface and obtains the Euclidean distance between the sampling point and the projection point along the normal vector direction as the normal deviation distance. This distance directly reflects the degree of local concavity or convexity of the workpiece surface relative to the ideal surface. The preset deviation threshold can be dynamically determined based on the statistical mean of the normal deviation distances of all sampling points, so that the judgment criteria are adaptively matched with the overall manufacturing level of the workpiece. Furthermore, the system can refit the reference surface using the sampling points marked as non-deformation points, and use the fitted reference surface as the comparison benchmark for subsequent deformation detection, thereby eliminating the interference of the identified deformation area on the benchmark construction and improving the stability of subsequent detection results.

[0009] The region extraction module determines the deformation concentration region based on the spatial distribution density of all deformation candidate points, and uses the deformation candidate points within the deformation concentration region as deformation detection nodes. In a preferred embodiment, this module divides the point cloud space into multiple three-dimensional voxel grids, counts the number of deformation candidate points within each voxel grid to obtain the deformation point density value, and performs neighborhood smoothing on the density value to generate a smooth density distribution map. Based on this distribution map, it determines the density peak region and clusters the nearby deformation candidate points within the region into deformation clusters. By calculating the spatial extension range of the deformation clusters, it filters out the true deformation concentration regions whose size exceeds a preset threshold, effectively filtering out isolated candidate points caused by noise or slight surface roughness. Deformation candidate points uniformly selected from the deformation concentration region at preset intervals are defined as deformation detection nodes and numbered and stored to ensure the representativeness and orderliness of deformation feature extraction, avoiding subsequent complex calculations on a large number of redundant points.

[0010] The curvature comparison module obtains the curvature change value corresponding to the deformation detection node based on its projection position on the fitted surface equation, and compares the curvature change value with a preset curvature change range. During operation, the spatial coordinates of each deformation detection node are projected onto the fitted surface along the normal vector direction. The Gaussian curvature value and principal curvature value at that location are calculated based on the projected coordinates, and the ratio of the Gaussian curvature value to the average curvature value is used as the curvature change value of that node. This curvature change value comprehensively reflects the degree of curvature and the intensity of its change in the local shape, exhibiting greater geometric sensitivity than a single normal deviation. The preset curvature change range is obtained by multiplying the theoretical curvature value of the three-dimensional design model of the workpiece under test by a preset tolerance coefficient, forming an upper and lower limit value for the design curvature, serving as a criterion for distinguishing between normal shape fluctuations and deformation defects. Through this threshold comparison based on curvature characteristics, minute curvature anomalies caused by residual stress or processing damage can be effectively identified, compensating for the problem of smooth but large-scale shape drift that cannot be detected by relying solely on distance thresholds.

[0011] When the curvature change value exceeds a preset curvature change range, the result recording module acquires the target deformation detection node that generated the abnormal curvature change value and records the spatial coordinates and timestamp of the node. The timestamp originates from the original sampling time corresponding to the node and can mark the temporal information of the deformation. Preferably, the system can further sort the deformation detection nodes according to the magnitude of the curvature change value deviating from the preset range, marking the node with the largest deviation as the main deformation node, and combining it with its adjacent deformation detection nodes to form a deformation feature point set. Based on this, a temporal spatial coordinate sequence is arranged according to the timestamps of all deformation detection nodes, the spatial displacement vector and displacement change rate between nodes at adjacent times are calculated, and cluster analysis is performed on nodes with consistent rates to divide their independent deformation evolution regions; the spatial coordinate extreme values ​​within each region are obtained to acquire deformation feature parameters such as the maximum deformation amplitude and deformation extension direction, generating a multi-dimensional deformation detection result containing spatial location, temporal evolution, and deformation amplitude. This recording method not only allows for locating the deformation position at the current moment but also traces the deformation development process, providing a continuous and quantitative basis for process adjustments.

[0012] To further ensure the reliability of deformation detection results, this invention can also assess the confidence level of the generated deformation detection results: The total number of all deformation detection nodes involved in the result generation and the number of abnormal nodes with curvature change values ​​exceeding the limit are obtained; the proportion of abnormal nodes is calculated to obtain the deformation confidence baseline coefficient; the deformation consistency coefficient is obtained based on the variance of the normal deviation distance of all deformation detection nodes; the deformation coverage coefficient is obtained based on the ratio of the spatial distribution density of deformation candidate points in the deformation concentration area to the preset standard density; and the overall verification confidence level is obtained by multiplying the above three coefficients. Quantification of confidence makes the reliability of the detection results transparent and verifiable, facilitating the decision on whether to re-inspect defective areas based on numerical values ​​in actual quality inspection scenarios, thereby effectively balancing detection efficiency and accuracy.

[0013] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0014] By calculating the normal deviation distance from each sampling point in the initial 3D point cloud model to the fitted surface, and marking sampling points whose normal deviation distance exceeds a preset deviation threshold as deformation candidate points, the quantitative extraction of minute geometric offsets is achieved. This method does not directly use the horizontal or vertical distance difference between sampling points, but instead calculates the Euclidean distance between the sampling point and the fitted surface along the surface normal vector direction. This ensures that the deviation measurement direction is consistent with the actual surface normal, accurately reflecting the degree of deformation on the 3D surface of the aluminum alloy die casting and avoiding distortion in deviation calculations caused by changes in the surface curvature of the die casting. Based on this, the spatial distribution density of deformation candidate points determines the deformation concentration region and extracts deformation detection nodes from it. This ensures that the subsequent analysis object is no longer a discrete, isolated outlier, but a representative point within a spatially clustered deformation region. This eliminates isolated deviation interference introduced by random noise or local roughness, focusing detection resources on the location where the actual deformation occurs.

[0015] The projected positions of deformation detection nodes on the fitted surface equation are obtained, and the corresponding curvature change values ​​are acquired. These curvature change values ​​are then compared with a preset curvature change range comprised of the upper and lower limits of the designed curvature. This expands the deformation judgment criterion from a single geometric distance deviation to a composite criterion that includes changes in the degree of surface curvature. When plastic deformation occurs in a localized area of ​​a die-cast part, not only does it cause a positional shift, but it also alters the surface curvature morphology of that area. By comparing the actual curvature change values ​​with the designed allowable curvature range, surface curvature anomalies that cannot be identified by distance thresholds alone can be detected, improving the accuracy of deformation detection for die-cast parts. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a schematic diagram of the deformation detection system for aluminum alloy die castings;

[0018] Figure 2 This is a flowchart of the deformation concentration region detection process based on deviation labeling and density clustering;

[0019] Figure 3 This is a flowchart comparing the range of personalized curvature variations based on projected curvature. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See Figure 1 This invention provides a deformation detection system for aluminum alloy die castings, including a data acquisition module, a point cloud construction module, a deviation marking module, a region extraction module, a curvature comparison module, and a result recording module. The data acquisition module acquires continuous surface contour data generated as the surface of the aluminum alloy die casting moves along multiple preset scanning paths. The continuous surface contour data includes multiple sampling points and their corresponding spatial coordinates. The point cloud construction module constructs an initial three-dimensional point cloud model based on the spatial coordinates of the multiple sampling points and performs surface fitting on the initial three-dimensional point cloud model to obtain the fitted surface equation. The deviation marking module calculates the normal deviation distance from each sampling point in the initial three-dimensional point cloud model to the fitted surface based on the fitted surface equation, and marks sampling points whose normal deviation distance exceeds a preset deviation threshold as deformation candidate points. The region extraction module determines the deformation concentration region based on the spatial distribution density of all deformation candidate points and uses the deformation candidate points within the deformation concentration region as deformation detection nodes. The curvature comparison module obtains the curvature change value corresponding to the deformation detection node based on the projection position of the deformation detection node on the fitted surface equation and compares the curvature change value with a preset curvature change range. When the curvature change value exceeds the preset curvature change range, the result recording module obtains the deformation detection node corresponding to the curvature change value, records the spatial coordinates and timestamp of the deformation detection node, and generates the deformation detection result based on the spatial coordinates and timestamp.

[0022] Example 1

[0023] In practical implementation, the data acquisition module acquires continuous surface contour data generated as the surface of the aluminum alloy die casting under test moves along multiple preset scanning paths. The surface of the aluminum alloy die casting under test is divided into multiple detection areas, and a corresponding preset scanning path is set for each detection area. Each preset scanning path includes a scanning start point, a scanning end point, and a scanning direction. The division of detection areas is based on the geometric characteristics of the aluminum alloy die casting under test, and is divided according to the degree of curvature change and surface connectivity, so that the surface inside each detection area is continuous and the curvature change is gentle. For each detection area, the scanning start point and scanning end point of the preset scanning path are set at the two endpoints of the boundary of the detection area, and the scanning direction is set as a vector direction from the scanning start point to the scanning end point. The laser displacement sensor is controlled to move at a constant speed from the scanning start point to the scanning end point along each preset scanning path. During the movement, the original distance values ​​of each sampling point on the surface of the aluminum alloy die casting under test are collected according to a preset sampling frequency. The preset sampling frequency is determined based on the response speed of the laser displacement sensor and the roughness requirements of the surface under test. In this embodiment, the preset sampling frequency is set to 1000 Hz.

[0024] The real-time attitude angle of the laser displacement sensor at each sampling point is acquired synchronously, and the relative coordinates of each sampling point in the sensor coordinate system are calculated based on the real-time attitude angle and the original distance value. The real-time attitude angle is acquired by an inertial measurement unit (IMU) mounted on the laser displacement sensor, which outputs the Euler angles of the laser displacement sensor in the pitch, yaw, and roll directions. The laser emission point of the laser displacement sensor is taken as the origin of the sensor coordinate system, with the X-axis along the laser emission direction, and the Y and Z axes determined according to the right-hand rule. For the i-th sampling point, the original distance value is denoted as... The pitch angle in the real-time attitude angle is denoted as Yaw angle is denoted as Roll angle is denoted as According to the principle of spherical coordinate transformation, the relative coordinates of the i-th sampling point in the sensor coordinate system are... Calculated using the following formula: , , The above calculations do not consider the effect of roll angle on coordinates because the laser spot position of the laser displacement sensor does not change with roll. In other implementations, when the laser displacement sensor has an installation offset, the roll angle needs to be incorporated into the compensation matrix to correct the coordinates.

[0025] The real-time motion trajectory coordinates of the laser displacement sensor during the scanning process are acquired, and the relative coordinates of each sampling point in the sensor coordinate system are transformed to spatial coordinates in the global coordinate system based on these real-time motion trajectory coordinates. The real-time motion trajectory coordinates are obtained through feedback from the motion controller of a multi-axis robotic arm or a coordinate measuring machine. The motion controller outputs the position coordinates and attitude matrix of the laser displacement sensor in the global coordinate system at each sampling point. The global coordinate system is established on the detection platform base, with the origin located at one corner of the platform, and the three coordinate axes parallel to the platform guide rails. The relative coordinates in the sensor coordinate system are multiplied by the attitude matrix on the left and then added to the position coordinates to obtain the spatial coordinates in the global coordinate system. The attitude matrix is ​​constructed from the pitch, yaw, and roll angles in the real-time attitude angles, and is in the form of a direction cosine matrix. After the transformation, each sampling point corresponds to a spatial coordinate in the global coordinate system. .

[0026] The spatial coordinates of all sampling points in the global coordinate system are stitched together according to the corresponding preset scan paths to obtain continuous surface contour data. During stitching, sampling points belonging to the same preset scan path are arranged in timestamp order to form a scan line point column. All scan line point columns corresponding to the preset scan paths are merged and indexed and stored according to the preset number of the scan path to form continuous surface contour data. The continuous surface contour data contains multiple sampling points and their corresponding spatial coordinates. At the same time, each sampling point also retains the corresponding original distance value, real-time attitude angle, and timestamp for subsequent traceability.

[0027] In some embodiments, the point cloud construction module constructs an initial 3D point cloud model based on the spatial coordinates of multiple sampling points, and performs surface fitting on the initial 3D point cloud model to obtain the fitted surface equation. Noise filtering is applied to the spatial coordinates of each sampling point in the continuous surface contour data to remove outlier sampling points whose spatial coordinates deviate from the average neighborhood coordinates by more than a preset neighborhood deviation multiple, resulting in a denoised effective sampling point set. The noise filtering process employs a statistical outlier removal algorithm. For each sampling point, its neighborhood is defined as the set of all other sampling points contained within a spatial sphere centered on the sampling point and with a radius of r. The radius r is set according to the average sampling interval of the continuous surface contour data, and is set to 3 times the average sampling interval. The average spatial coordinates of all sampling points within the neighborhood point set are calculated to obtain the neighborhood average coordinates. The Euclidean distance deviation between the spatial coordinates of the sampling point and the neighborhood average coordinates is calculated. The mean μ and standard deviation σ of the Euclidean distance deviations of all sampling points are statistically analyzed. The preset neighborhood bias factor k is set to 1.5. The rejection criterion is: if the Euclidean distance deviation of a sampling point is greater than μ + k·σ, then the sampling point is identified as an outlier and rejected. The expression for this criterion is:

[0028]

[0029] in, This represents the Euclidean distance deviation between the spatial coordinates of sampling point P and the average coordinates of its neighborhood. This represents the mean of the Euclidean distance deviations across all sampling points; This represents the standard deviation of the Euclidean distance deviation across all sampling points. This indicates the preset neighborhood deviation factor. . The value of 1.5 is based on the assumption that under the normal distribution, about 86.6% of the data falls within the range of mean ± 1.5 times the standard deviation. This can remove significantly deviated noise points and avoid accidentally deleting normal sampling points located at the edge of the surface.

[0030] The effective sampling point set obtained after noise filtering is downsampled using voxels according to a preset spatial grid partitioning rule to obtain a uniformly distributed downsampled point set. The preset spatial grid partitioning rule is as follows: the 3D bounding box containing the effective sampling point set is uniformly divided into cubic voxels with equal side lengths along the coordinate axes. The side length of the cube is set to twice the average sampling interval of the continuous surface contour data. Within each cubic voxel, a centroid is used to replace all effective sampling points contained within the voxel. The spatial coordinates of the centroid are the arithmetic mean of the spatial coordinates of all effective sampling points within the voxel. If a cubic voxel has no effective sampling points, no centroid is generated. The centroids generated within all voxels constitute the downsampled point set.

[0031] Based on the downsampled point set, a local fitting plane is constructed using the moving least squares method. The normal vectors of the downsampled point set are then redirected according to the consistency of the normal vectors across all local fitting planes, resulting in point cloud data with consistent normal vectors. For each point in the downsampled point set, its q nearest neighbors are selected as the local fitting point set, where q is 15. A local coordinate system is established centered on this point, and a quadratic polynomial surface is fitted using the moving least squares method, calculating the unit normal vector at that point. After calculating the normal vectors for all downsampled points, there is a problem where the normal vectors of some points point inwards and others outwards. By constructing a minimum spanning tree, the normal vector of an initial point is selected as the reference direction, and the normal direction is propagated along the minimum spanning tree, ensuring that the dot product of the normal vectors of adjacent points is non-negative, thus completing the normal vector redirection.

[0032] Based on point cloud data with consistent normal vectors, an implicit surface reconstruction algorithm is used to generate an initial 3D point cloud model. The implicit surface reconstruction algorithm is a Poisson surface reconstruction algorithm; the input is a set of downsampled points with normal vectors, and the output is an indicator function defined on an octree structure. A triangular mesh surface is extracted from the indicator function using an isosurface extraction algorithm, and this triangular mesh surface constitutes the initial 3D point cloud model. The initial 3D point cloud model is then parametrically mapped, projecting the 3D point cloud onto a 2D parameter domain. The parametric mapping uses a least-squares conformal mapping method, mapping each vertex of the triangular mesh surface to the parameter coordinates (u, v) within the planar rectangular parameter domain. In the 2D parameter domain, a bicubic B-spline surface approximation algorithm is used to fit the surface, obtaining the fitted surface equation. A bicubic B-spline surface is defined as the tensor product surface of cubic B-spline basis functions along the u and v directions on the control mesh. The vertex coordinates of the control mesh are optimized by minimizing the sum of squared distances from the surface to the sampling points. The fitted surface equation is finally expressed in parametric form. ,in The three-dimensional spatial coordinates and normal vectors of points on the surface can be calculated at any parameter coordinate.

[0033] Example 2

[0034] See Figure 2 In practice, the deviation marking module obtains the spatial coordinates of each sampling point in the initial 3D point cloud model, and substitutes these coordinates into the fitting surface equation to calculate the coordinates of the nearest projection point of each sampling point on the fitting surface. The calculation of the nearest projection point is achieved by solving an optimization problem of finding the shortest distance from the sampling point to the fitting surface. The fitting surface equation is in parametric form. , where the parameter field For the first 3D point cloud model... There are 1 sampling point, and its spatial coordinates are 1. Find the parameter coordinates on the fitted surface equation , making To the surface point The Euclidean distance is minimized. The Newton-Raphson iterative method is used to solve for the parameter coordinates, with the initial values ​​set as the center of the parameter domain. The convergence condition for the iteration is that the change in parameter coordinates between two consecutive iterations is less than 1. After convergence, the coordinates of the nearest projected point are obtained. .

[0035] Based on the spatial coordinates of each sampling point and the coordinates of the corresponding nearest projection point, calculate the normal deviation distance from each sampling point to the fitted surface. The normal deviation distance is the Euclidean distance from the sampling point along the normal vector direction of the fitted surface to the nearest projection point. For the Each sampling point, the fitted surface at the projection point The unit normal vector at point is denoted as This is obtained by normalizing the cross product of surface partial derivatives. The spatial coordinates of the current sampling point and the coordinates of the projected point form a vector. Normal deviation distance Pick ,because The direction may be the same as or opposite to the normal vector; taking the absolute value ensures that the deviation distance is non-negative. When and When the included angle is less than 90 degrees, the deviation is positive, and vice versa. However, the absolute value is used in subsequent comparisons.

[0036] The absolute values ​​of the normal deviation distances of all sampling points are counted, and the arithmetic mean is calculated to obtain the mean normal deviation distance. A preset deviation threshold is set based on the average normal deviation distance. The calculation formula is:

[0037]

[0038] in, Indicates the preset deviation threshold; This represents the mean of the absolute values ​​of the normal deviation distances of all sampling points; This represents the deviation threshold amplification factor. The value is 2.0. Deviation threshold amplification factor. The setting of 2.0 is based on the following: multiple scan analyses were performed on qualified samples from the same batch as the aluminum alloy die-casting part to be tested, and the normal deviation distance of each sampling point on the surface of the qualified samples was distributed within the mean value. Nearly, the fluctuation range generally does not exceed twice the average. Setting it to 2.0 allows it to include normal deviations caused by surface roughness while identifying deformation features that exceed the normal fluctuation range.

[0039] The absolute value of the normal deviation distance of each sampling point is sequentially compared with the preset deviation threshold. Compare them. If the absolute value of the normal deviation distance of a sampling point is greater than... Then mark the sampling point as a candidate deformation point; if the absolute value of the normal deviation distance of a sampling point is less than or equal to If the sampling point is not found, it is marked as a non-deformable point. The spatial coordinates of all non-deformable points are extracted to form a set of non-deformable points. Based on the spatial coordinates of all non-deformable points, a reference surface is refitted. The refitting process uses the same algorithm as constructing the fitted surface equation: voxel downsampling of the non-deformable point set, calculation of the normal vector using the moving least squares method and consistent retargeting, reconstruction of the triangular mesh surface using a Poisson surface, and finally approximation using a bicubic B-spline surface in the two-dimensional parameter domain to obtain the reference surface equation. The reference surface serves as a comparison benchmark for subsequent deformation detection.

[0040] In some embodiments, the region extraction module determines the deformation concentration region based on the spatial distribution density of all deformation candidate points. The spatial region containing the initial 3D point cloud model is divided into multiple 3D voxel grids. The side length of the 3D voxel grid is set to 5 times the average sampling interval of the continuous surface contour data. This multiple ensures that each non-empty voxel contains a sufficient number of deformation candidate points for density statistics. The number of deformation candidate points in each 3D voxel grid is counted to obtain the deformation point density value of each 3D voxel grid. If a 3D voxel grid does not contain any deformation candidate points, its deformation point density value is 0.

[0041] Neighborhood smoothing is applied to the density values ​​of deformation points in each 3D voxel mesh to obtain a smoothed density distribution map. The neighborhood smoothing uses a size of [missing value]. The three-dimensional mean filtering kernel takes a given three-dimensional voxel grid as its center, takes the deformation point density values ​​of its 26 spatially adjacent three-dimensional voxel grids, and together with the central three-dimensional voxel grid, calculates the arithmetic mean, and assigns this mean value to the central three-dimensional voxel grid as the smoothed deformation point density value. For voxels located at the spatial boundary of the voxel grid, only their effective neighboring voxels are used for averaging.

[0042] Density peak regions are determined based on the smoothed density distribution map. A density peak determination threshold is set, which is 70% of the density value of the maximum smoothed deformation point in the smoothed density distribution map. All 3D voxel meshes are traversed, and 3D voxel meshes with smoothed deformation point density values ​​greater than the density peak determination threshold are marked as high-density voxels. Using a 3D connected component analysis method, adjacent voxels marked as high-density voxels are merged to form several density peak regions, each density peak region being a spatially connected component.

[0043] All deformation candidate points within a density peak region are grouped into the same deformation cluster. For example, a density peak region may contain deformation candidate points scattered across multiple adjacent voxels; these candidate points are assigned to a single deformation cluster. For each deformation cluster, the spatial centroid coordinates and spatial extent are calculated. The spatial centroid coordinates are the arithmetic mean of the spatial coordinates of all deformation candidate points within the cluster. The spatial extent is determined by calculating the span of the deformation cluster along the three coordinate axes of the global coordinate system; specifically, the difference between the maximum and minimum X-coordinate values ​​of all deformation candidate points is taken as the X-axis extent, the difference between the maximum and minimum Y-coordinate values ​​as the Y-axis extent, and the difference between the maximum and minimum Z-coordinate values ​​as the Z-axis extent.

[0044] The size of a deformation cluster is determined based on its spatial extension range to see if it exceeds a preset deformation size threshold. This threshold is set to 10% of the minimum feature size in the 3D design model of the aluminum alloy die casting under test. The minimum feature size in the 3D design model refers to the minimum side length or diameter of the structural feature with the strictest dimensional tolerances in the die casting design requirements; this value is obtained from the design drawings. If the spatial extension range of a deformation cluster in any direction exceeds the preset deformation size threshold, the spatial area covered by the cluster is defined as a deformation concentration area. If the spatial extension range of a deformation cluster in all three directions is less than or equal to the preset deformation size threshold, the cluster is considered small and not considered a deformation concentration area.

[0045] Deformation candidate points uniformly selected at preset intervals within the deformation concentration area are used as deformation detection nodes. The preset interval is set to be the same as the side length of the 3D voxel mesh, i.e., 5 times the average sampling interval of the continuous surface contour data. The selection process is as follows: within the deformation concentration area, starting from the mesh node position of the voxel mesh, the deformation candidate point closest to each mesh node is selected as the deformation detection node. If there are no deformation candidate points around a mesh node, that mesh node is skipped. The spatial coordinates of the selected deformation detection nodes are numbered and stored according to the detection order. The detection order is determined based on the extension direction of the scanning path in the deformation concentration area and is arranged in a serpentine traversal manner, so that adjacent numbered deformation detection nodes are spatially adjacent or continuous along the scanning path.

[0046] Example 3

[0047] See Figure 3 In practical implementation, the curvature comparison module obtains the spatial coordinates of each deformation detection node and projects these coordinates onto the fitted surface equation along the normal vector direction of the fitted surface, thus obtaining the projected coordinates of each deformation detection node. The fitted surface equation is in parametric form. It is expressed by a bicubic B-spline surface, with a parameter domain. For the first Each deformation detection node has the following spatial coordinates: The projection process employs iterative nearest-point solving from a point to the parametric surface: initial values ​​for the projection parameters are set. The initial value is set to the coordinates of the projection parameters obtained from the previous adjacent deformation detection node. If the current deformation detection node is the first projection node within the deformation concentration region, the initial value is set to... In the first In the next iteration, the current parameters are calculated. Corresponding surface points and surface partial derivatives and Construct the error vector Solving the linear equations yields the parameter increments. Update parameters as follows The iteration termination condition is: The iteration count reaches 20 times or the number of iterations reaches millimeters. After convergence, the projected coordinates of the deformation detection nodes on the fitted surface equation are obtained. .

[0048] Based on the projected coordinates of each deformation detection node, calculate the principal curvature and Gaussian curvature values ​​of the fitted surface equation at those projected coordinates. For projected coordinates... Corresponding parameter coordinates Calculate the equation of the fitted surface The first-order partial derivative vector at the parameter coordinates and and the second-order partial derivative vector , and The unit normal vector is obtained by cross product of first-order partial derivatives and normalization. Calculate the first fundamental form coefficients of the surface: , , ; Calculate the second fundamental form coefficients of the surface: , , Mean curvature and Gaussian curvature Calculated using the following formulas respectively:

[0049]

[0050]

[0051] Principal curvature value and Satisfy the equation Solving for the given information yields the following results. and .

[0052] The average curvature value is calculated based on the principal curvature value, and the ratio of the Gaussian curvature value to the average curvature value is used as the curvature change value corresponding to the deformation detection node. Curvature change value corresponding to each deformation detection node The calculation formula is:

[0053]

[0054] In the formula, Indicates the first The curvature change value corresponding to each deformation detection node; The equation of the fitted surface is expressed as follows: Gaussian curvature values ​​at the projected coordinates of each deformation detection node; The equation of the fitted surface is represented in the th case. The average curvature value at the projected coordinates of each deformation detection node; This represents a very small constant to prevent the denominator from being zero. Values The Gaussian curvature value is taken as an absolute value because the curvature change caused by deformation can be represented by an increase in the absolute value of the Gaussian curvature in both convex and concave regions. Taking the absolute value makes the curvature change value have a symmetrical response characteristic to the two deformation directions. A minimal constant is introduced into the denominator. The purpose is to prevent division-to-zero anomalies when the average curvature is close to zero in flat regions. Value The curvature is much smaller than the average curvature of a normal aluminum alloy die-cast surface and will not have a substantial impact on the calculation results of the curvature change value.

[0055] A preset curvature variation range is obtained, which consists of the upper and lower limits of the design curvature of the aluminum alloy die casting under test. The upper and lower limits are obtained by multiplying the theoretical curvature values ​​of the 3D design model of the aluminum alloy die casting under test by preset tolerance coefficients. The 3D design model is the nominal geometric model of the aluminum alloy die casting under test constructed in CAD software. On the surface of the 3D design model, coordinates are projected relative to the deformation detection nodes. The corresponding geometric location is the sampling point. Calculate the Gaussian curvature of the 3D design model surface at that sampling point. and mean curvature And calculate the theoretical curvature change value. The preset tolerance factor includes the upper limit tolerance factor. and lower limit tolerance coefficient , The value is 1.15. The value is set to 0.85. The upper and lower tolerance coefficients are set based on the casting tolerance grades specified in the national standard GB / T6414 for aluminum alloy die castings. The deviation rate between the maximum and minimum material dimensions corresponding to CT6 grade tolerance is approximately ±15%. Therefore, the tolerance coefficient is set to fluctuate by 15% above and below the theoretical value. (Design curvature upper limit value) Design curvature lower limit .

[0056] The curvature change value corresponding to each deformation detection node is sequentially recorded. The curvature change value is compared with the preset curvature change range to determine whether it is less than the lower limit of the designed curvature or greater than the upper limit of the designed curvature. or At that time, the judgment of the first The curvature change value of each deformation detection node exceeds the preset curvature change range; when At that time, the judgment of the first The curvature change value of each deformation detection node is within the preset curvature change range.

[0057] In some embodiments, the theoretical curvature variation values ​​of the 3D design model differ for different regions of the die-cast part surface within the deformation concentration area. The theoretical curvature variation value is larger in high-curvature regions, resulting in correspondingly larger upper and lower limits of design curvature under the preset tolerance coefficient; conversely, the theoretical curvature variation value is smaller in low-curvature regions, leading to correspondingly smaller upper and lower limits of design curvature. Instead of using a globally uniform threshold, a personalized preset curvature variation range is obtained for each deformation detection node based on its projection position, thus adapting to the non-uniform curvature distribution of the aluminum alloy die-cast part surface.

[0058] Example 4

[0059] In practical implementation, when the result recording module determines that the curvature change value exceeds the preset curvature change range, it marks the curvature change value exceeding the preset curvature change range as an abnormal curvature change value and obtains the target deformation detection node that generated the abnormal curvature change value. The determination process is completed in the curvature comparison module. The result recording module receives the comparison result data output by the curvature comparison module. The comparison result data includes the number, spatial coordinates, projected coordinates, curvature change value, upper limit of design curvature, lower limit of design curvature, and a Boolean flag indicating whether the limit is exceeded for each deformation detection node. The result recording module traverses the comparison result data of all deformation detection nodes, filters out the deformation detection nodes with the Boolean flag set to true, marks their corresponding curvature change values ​​as abnormal curvature change values, and determines the deformation detection node that generated the abnormal curvature change value as the target deformation detection node.

[0060] The spatial coordinates of the target deformation detection node are obtained, and the original acquisition time corresponding to the target deformation detection node is queried in the continuous surface contour data based on the spatial coordinates. The continuous surface contour data is organized in timestamp order when generated by the data acquisition module, and each sampling point records its spatial coordinates and corresponding acquisition time. All sampling points in the continuous surface contour data are traversed to find the sampling point with the smallest Euclidean distance between its spatial coordinates and the spatial coordinates of the target deformation detection node. The acquisition time corresponding to this sampling point is the original acquisition time of the target deformation detection node. The original acquisition time is used as the timestamp of the target deformation detection node, and the timestamp is associated with and stored with the spatial coordinates of the target deformation detection node, forming a structured record containing the deformation detection node number, abnormal curvature change value, spatial coordinates, and timestamp.

[0061] Target deformation detection nodes are sorted according to the degree of deviation of their curvature change values ​​from the preset curvature change range. For a target deformation detection node, the degree of deviation is defined as the ratio of the difference between the curvature change value and the upper limit of the design curvature to the upper limit of the design curvature. When the curvature change value is greater than the upper limit of the design curvature, the degree of deviation is positive; when the curvature change value is less than the lower limit of the design curvature, the degree of deviation is defined as the ratio of the difference between the lower limit of the design curvature and the curvature change value to the lower limit of the design curvature, and is also positive. All target deformation detection nodes are sorted in descending order of their deviation values ​​to obtain a sorted list. The target deformation detection node with the largest deviation value is marked as the main deformation node according to the sorted list. If two or more target deformation detection nodes have the same and the same maximum deviation value, the target deformation detection node whose spatial coordinates are closest to the geometric center of the deformation concentration area is selected as the main deformation node.

[0062] Obtain the adjacent deformation detection nodes surrounding the main deformation node. Adjacent deformation detection nodes are identified by taking the c nearest neighbors (c = 2) from the position of the main deformation node in the deformation detection node numbering sequence, moving forward and backward from its position in the sequence. The deformation detection node numbering sequence is arranged in a serpentine traversal pattern in the region extraction module, ensuring that nodes with adjacent numbers are also spatially adjacent. The main deformation node and these four adjacent deformation detection nodes together form a deformation feature point set, containing a total of five deformation detection nodes.

[0063] In some embodiments, the result recording module generates deformation detection results based on spatial coordinates and timestamps. The spatial coordinates and timestamps of all deformation detection nodes are acquired. The range of all deformation detection nodes includes anomalous nodes marked as target deformation detection nodes and normal deformation detection nodes not marked as anomalous. The spatial coordinates are arranged temporally according to the timestamps, with the earliest timestamp placed first and the latest timestamp placed last. Deformation detection nodes with the same timestamp are arranged in ascending order of their numbers, resulting in a temporal spatial coordinate sequence of the deformation detection nodes. This temporal spatial coordinate sequence reflects the spatial positional changes of the deformation detection nodes acquired along the scanning path in the time dimension.

[0064] Calculate the spatial displacement vector between deformation detection nodes corresponding to adjacent timestamps based on the temporal spatial coordinate sequence. Let the spatial coordinates of the s-th deformation detection node in the temporal spatial coordinate sequence be... The corresponding timestamp is The spatial coordinates of the (s+1)th deformation detection node are The corresponding timestamp is Spatial displacement vector between adjacent deformation detection nodes Defined as That is, the three components of the spatial displacement vector are respectively , , The displacement rate of each deformation detection node is calculated based on the spatial displacement vector. The displacement rate of the s-th deformation detection node is... The calculation formula is:

[0065]

[0066] In the formula, This represents the displacement rate of the s-th deformation detection node in the temporal spatial coordinate sequence; This represents the spatial coordinate difference between the (s+1)th deformation detection node and the sth deformation detection node in the global coordinate system along the X-axis. This represents the spatial coordinate difference between the (s+1)th deformation detection node and the sth deformation detection node in the global coordinate system along the Y-axis. This represents the spatial coordinate difference between the (s+1)th deformation detection node and the sth deformation detection node in the global coordinate system along the Z-axis. This represents the timestamp value of the (s+1)th deformation detection node in the temporal spatial coordinate sequence. This represents the timestamp value of the s-th deformation detection node in the temporal spatial coordinate sequence. Displacement rate of change. The unit is millimeters per second.

[0067] Cluster analysis was performed on the deformation detection nodes based on their displacement change rates. A density-based clustering algorithm was used, with the following parameters: the neighborhood radius ε was set to 0.3 times the arithmetic mean of all displacement change rates, and the minimum number of neighborhood points required for core point determination was set to 3. Deformation detection nodes whose displacement change rates fell within each other's neighborhoods were grouped into the same category, with nodes in each category exhibiting similar displacement change rates. After cluster analysis, deformation detection nodes with consistent displacement change rates were grouped into the same deformation evolution region. All deformation detection nodes within a deformation evolution region not only exhibited consistent displacement change rates but also remained spatially connected through density connections established during the clustering process.

[0068] Obtain the spatial coordinate extrema of the deformation detection node within each deformation evolution region. The spatial coordinate extrema include the maximum value of the X-coordinate. Minimum value of X coordinate Maximum value of Y coordinate Minimum value of Y coordinate Maximum Z-coordinate and minimum Z coordinate The maximum deformation amplitude and deformation extension direction of each deformation evolution region are calculated based on the spatial coordinate extreme values. Maximum deformation amplitude. Take the maximum value of the extended range in the three coordinate axis directions, that is The direction of deformation extension is determined by the coordinate axis direction that achieves the maximum extension range. If the extension range is largest along the X-axis, the direction of deformation extension is the X-axis direction; if the extension range is largest along the Y-axis, the direction of deformation extension is the Y-axis direction; if the extension range is largest along the Z-axis, the direction of deformation extension is the Z-axis direction.

[0069] The maximum deformation amplitude and deformation extension direction are used as deformation characteristic parameters. For each deformation evolution region, a set of deformation characteristic parameters is generated, including the deformation evolution region number, the number of deformation detection nodes in the region, a list of deformation detection node numbers in the region, the maximum deformation amplitude, the deformation extension direction, and the average displacement change rate of all deformation detection nodes in the region. Deformation detection results are generated based on the deformation characteristic parameters of all deformation evolution regions. The deformation detection results include the name and batch number of the aluminum alloy die casting to be tested, the test date, a summary table of deformation characteristic parameters of all deformation evolution regions, a list of spatial coordinates and timestamps for each target deformation detection node, and the spatial coordinates and deviation value of the main deformation node.

[0070] Example 5

[0071] In practice, the preset scanning path is generated as follows: A 3D design model of the aluminum alloy die casting to be tested is acquired, and a surface curvature distribution map of the die casting is extracted based on the 3D design model. The 3D design model is the nominal geometric model of the die casting in CAD software, in the STEP or IGES standard exchange format. The 3D design model is imported into a geometric analysis engine, which discretizes all outer surfaces of the 3D design model, generating a mesh model composed of triangular facets. Each vertex of a triangular facet contains spatial coordinates and a vertex normal vector. Based on the mesh model, the Gaussian curvature and mean curvature of each triangular facet are calculated, and the absolute value of the Gaussian curvature or mean curvature is assigned as a curvature index to each vertex of the triangular facet. Color mapping is applied to the curvature indices of all vertices on the mesh model to generate a surface curvature distribution map. Areas with larger curvature index values ​​in the surface curvature distribution map correspond to areas with drastic changes in surface geometry, while areas with smaller curvature index values ​​correspond to areas with gentler surface geometry.

[0072] Based on the surface curvature distribution map, high-curvature and low-curvature regions on the surface of the aluminum alloy die-casting under test are identified. A curvature segmentation threshold is set, which is 1.5 times the median curvature index value of all vertices in the surface curvature distribution map. Regions containing vertices with curvature index values ​​greater than the curvature segmentation threshold are identified as high-curvature regions, and regions containing vertices with curvature index values ​​less than or equal to the curvature segmentation threshold are identified as low-curvature regions. Connectivity analysis is performed on the vertices of the identified high-curvature and low-curvature regions respectively. Spatially adjacent vertices with the same curvature category are merged into connected regions, resulting in several high-curvature region blocks and low-curvature region blocks. The scanning path density of high-curvature regions is set to be greater than that of low-curvature regions, and the scanning path spacing of high-curvature regions is set to 2 mm, while the scanning path spacing of low-curvature regions is set to 8 mm. The spacing is set based on the following: the surface curvature of high-curvature regions changes drastically, requiring denser sampling points to capture subtle deformation features; the surface of low-curvature regions is gentle, and a larger spacing is sufficient to meet the deformation detection accuracy requirements.

[0073] Multiple longitudinal and transverse scan lines are set in the high-curvature region, forming a pre-defined, intersecting grid-like scan path. Each high-curvature region is independently configured with scan lines. The smallest bounding rectangle of the high-curvature region is used as the scan line layout area. Longitudinal scan lines are evenly spaced along the length of the rectangle (2 mm spacing), while transverse scan lines are evenly spaced along the width (2 mm spacing). At their intersections, the longitudinal and transverse scan lines form an orthogonal grid. The laser displacement sensor passes through the grid intersection point twice, once along the longitudinal scan line direction and once along the transverse scan line direction. This intersecting grid-like pre-defined scan path ensures that the high-curvature region is scanned and covered in both orthogonal directions, enabling the acquisition of surface contour information from different directions.

[0074] Multiple parallel scan lines are set in the low-curvature region, extending along the length of the aluminum alloy die casting under test. The length of the die casting corresponds to the direction of the maximum dimension in the 3D design model. The axial span of the three-axis bounding box of the 3D design model is calculated, and the axis with the largest span is taken as the length direction. Parallel scan lines are evenly spaced along the length of the low-curvature region block, with a spacing of 8 mm. Each parallel scan line extends from one end of the low-curvature region block to the other, completely covering the low-curvature region block.

[0075] The intersecting grid-like preset scan paths and parallel scan lines are merged to obtain multiple preset scan paths covering the entire surface of the aluminum alloy die-casting under test. During the merging process, each longitudinal scan line, each transverse scan line, and each parallel scan line is treated as an independent preset scan path and assigned a unique number. For the intersecting grid-like preset scan paths, the scanning start and end points of the longitudinal and transverse scan lines are set according to the scanning direction: the longitudinal scan line extends from one boundary of the rectangle to the opposite boundary, and the transverse scan line extends from the other boundary of the rectangle to the opposite boundary. The scanning start and end points of the parallel scan lines are set at the two ends of the low-curvature region block along its length. The scanning direction of all preset scan paths is set from the scanning start point to the scanning end point. All preset scan paths are stored in the scan path configuration file in numerical order for use by the data acquisition module.

[0076] In some embodiments, after generating deformation detection results based on spatial coordinates and timestamps, a confidence assessment is also performed on the deformation detection results. The total number of all deformation detection nodes on which the deformation detection results are generated is obtained, denoted as [missing information]. All deformation detection nodes are those selected, numbered, and stored by the region extraction module within the deformation set region. The number of abnormal nodes whose curvature change values ​​exceed the preset curvature change range is recorded as follows: The criteria for determining abnormal nodes are the deformation detection nodes whose Boolean flag is true in the comparison result data output by the curvature comparison module, that is, deformation detection nodes whose curvature change value is less than the lower limit of the design curvature or greater than the upper limit of the design curvature.

[0077] The deformation confidence baseline coefficient is obtained by calculating the ratio of the number of abnormal nodes to the total number of deformation detection nodes. Deformation confidence baseline coefficient The calculation formula is:

[0078]

[0079] In the formula, Indicates the basic confidence coefficient of deformation; This indicates the number of abnormal nodes whose curvature changes exceed the preset curvature change range; This represents the total number of deformation detection nodes upon which the deformation detection results are based. Deformation confidence baseline coefficient. The value range is between 0 and 1. A higher value indicates a higher proportion of abnormal nodes detected among all deformation detection nodes, and a higher basic reliability of the deformation detection results.

[0080] The normal deviation distance for each deformation detection node is obtained, and the variance of the normal deviation distances for all deformation detection nodes is calculated. The deformation consistency coefficient is then derived from the variance. The normal deviation distances for each deformation detection node are extracted from the normal deviation distance data of the sampling points stored in the deviation marking module. Using the spatial coordinates of each deformation detection node as an index, the absolute value of the corresponding normal deviation distance is retrieved from the normal deviation distance data. The normal deviation distances of all deformation detection nodes constitute a dataset. Calculate the variance of this dataset. The variance value reflects the degree of dispersion of the deformation detection nodes from the reference surface. Deformation consistency coefficient. The calculation method is as follows: a variance reference value is preset. Variance reference value The value is the statistical average of the variance of the normal deviation distance of the deformation detection nodes in the inspection of qualified aluminum alloy die-casting samples of the same batch. hour, The value is 1.0; when hour, Deformation consistency coefficient The value ranges from 0 to 1. The closer the value is to 1, the more concentrated the normal deviation distribution of the deformation detection nodes is, and the better the consistency of the deformation characteristics.

[0081] The spatial distribution density of all candidate deformation points within the deformation concentration region is obtained, and the ratio of this spatial distribution density to the preset standard density is calculated to obtain the deformation coverage coefficient. (Spatial distribution density of all candidate deformation points within the deformation concentration region) Defined as the total number of deformation candidate points within the deformation concentration region divided by the volume of the deformation concentration region, in units of points per cubic millimeter. Preset standard density. Based on the surface area of ​​the 3D design model of the aluminum alloy die casting to be tested and the expected detection accuracy requirements, a deformation candidate point of 0.05 per cubic millimeter was determined. This value corresponds to a uniform distribution of 0.05 deformation candidate points per cubic millimeter, meeting the density requirement for complete capture of the deformation characteristics of the die casting surface. Deformation coverage coefficient The calculation formula is: When the calculation result is greater than 1.0, The cutoff value is 1.0. Deformation coverage factor. The value ranges from 0 to 1. The closer the value is to 1, the higher the spatial distribution density of deformation candidate points in the deformation concentration area reaches or exceeds the standard requirements, and the more sufficient the spatial coverage of deformation detection is.

[0082] Deformation confidence baseline coefficients Deformation consistency coefficient and deformation coverage coefficient Multiply by the product to obtain the overall validation confidence level of the deformation detection results. ,Right now Overall validation confidence level The value range is between 0 and 1. A value closer to 1 indicates a higher level of confidence in the deformation detection results. Overall validation confidence level. As an auxiliary parameter to the deformation detection results, it is output together with the deformation detection results to guide the subsequent adoption and application of the deformation detection results.

[0083] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. An aluminum alloy die casting deformation amount detection system characterized by comprising: include: The data acquisition module acquires continuous surface contour data generated when the surface of the aluminum alloy die casting to be tested moves along multiple preset scanning paths. The continuous surface contour data includes multiple sampling points and their corresponding spatial coordinates. The point cloud construction module constructs an initial three-dimensional point cloud model based on the spatial coordinates of multiple sampling points, and performs surface fitting on the initial three-dimensional point cloud model to obtain the fitted surface equation. The deviation marking module calculates the normal deviation distance from each sampling point in the initial three-dimensional point cloud model to the fitting surface based on the fitted surface equation, and marks the sampling points whose normal deviation distance exceeds a preset deviation threshold as deformation candidate points. The region extraction module determines the deformation concentration region based on the spatial distribution density of all deformation candidate points, and uses the deformation candidate points within the deformation concentration region as deformation detection nodes. The curvature comparison module obtains the curvature change value corresponding to the deformation detection node based on the projection position of the deformation detection node on the fitted surface equation, and compares the curvature change value with a preset curvature change range. If the curvature change value exceeds the preset curvature change range, the result recording module obtains the deformation detection node corresponding to the curvature change value, records the spatial coordinates and timestamp of the deformation detection node, and generates a deformation detection result based on the spatial coordinates and timestamp.

2. The aluminum alloy die casting deformation amount detection system according to claim 1, characterized by The acquisition of continuous surface contour data generated when the surface of the aluminum alloy die casting under test moves along multiple preset scanning paths includes: The surface of the aluminum alloy die casting to be tested is divided into multiple detection areas, and a corresponding preset scanning path is set for each detection area. Each preset scanning path includes a scanning start point, a scanning end point, and a scanning direction. The laser displacement sensor is controlled to move from the scanning start point to the scanning end point along each preset scanning path, and the original distance values ​​of each sampling point on the surface of the aluminum alloy die casting under test are collected according to the preset sampling frequency during the movement. The real-time attitude angle of the laser displacement sensor at each sampling point is acquired synchronously, and the relative coordinates of each sampling point in the sensor coordinate system are calculated based on the real-time attitude angle and the original distance value. The real-time motion trajectory coordinates of the laser displacement sensor during the scanning process are obtained, and the relative coordinates of each sampling point in the sensor coordinate system are transformed to the spatial coordinates in the global coordinate system based on the real-time motion trajectory coordinates. The spatial coordinates of all sampling points in the global coordinate system are stitched together according to the corresponding preset scanning path to obtain continuous surface contour data.

3. The aluminum alloy die casting deformation amount detection system according to claim 1, characterized by The process of constructing an initial 3D point cloud model based on the spatial coordinates of multiple sampling points, and performing surface fitting on the initial 3D point cloud model to obtain the fitted surface equation, includes: The spatial coordinates of each sampling point in the continuous surface contour data are subjected to noise filtering to remove outlier sampling points whose spatial coordinates deviate from the average coordinates of the neighborhood by more than a preset neighborhood deviation multiple, thus obtaining a set of effective sampling points after noise reduction. The effective sampling point set is downsampled using voxels according to a preset spatial grid division rule to obtain a uniformly distributed downsampled point set; Based on the downsampled point set, a local fitting plane is constructed using the moving least squares method, and the normal vector of the downsampled point set is redirected according to the consistency of the normal vectors of all local fitting planes to obtain point cloud data with consistent normal vectors. Based on the point cloud data with consistent normal vectors, an implicit surface reconstruction algorithm is used to generate an initial 3D point cloud model. The initial three-dimensional point cloud model is parametrically mapped to project the three-dimensional point cloud onto a two-dimensional parameter domain, and a bicubic B-spline surface approximation algorithm is used to fit the surface in the two-dimensional parameter domain to obtain the fitted surface equation.

4. The deformation detection system for aluminum alloy die castings according to claim 1, characterized in that, The step of calculating the normal deviation distance from each sampling point in the initial 3D point cloud model to the fitted surface based on the fitted surface equation, and marking sampling points whose normal deviation distance exceeds a preset deviation threshold as deformation candidate points, includes: Obtain the spatial coordinates of each sampling point in the initial three-dimensional point cloud model, and substitute the spatial coordinates of each sampling point into the fitting surface equation to calculate the coordinates of the nearest projection point of each sampling point on the fitting surface; Based on the spatial coordinates of each sampling point and the coordinates of the corresponding nearest projection point, the normal deviation distance from each sampling point to the fitted surface is calculated, wherein the normal deviation distance is the Euclidean distance from the sampling point along the normal vector direction of the fitted surface to the nearest projection point; The mean of the normal deviation distance of all sampling points is calculated, and a preset deviation threshold is set based on the mean. The normal deviation distance of each sampling point is compared with the preset deviation threshold in turn. Sampling points with a normal deviation distance greater than the preset deviation threshold are marked as deformation candidate points, and sampling points with a normal deviation distance less than or equal to the preset deviation threshold are marked as non-deformation points. The reference surface is refitted based on the spatial coordinates of all non-deformed points, and the reference surface is used as the comparison benchmark for subsequent deformation detection.

5. The deformation detection system for aluminum alloy die castings according to claim 1, characterized in that, The step of determining the deformation concentration region based on the spatial distribution density of all deformation candidate points, and using the deformation candidate points within the deformation concentration region as deformation detection nodes, includes: The spatial region where the initial three-dimensional point cloud model is located is divided into multiple three-dimensional voxel grids, and the number of deformation candidate points in each three-dimensional voxel grid is counted to obtain the deformation point density value of each three-dimensional voxel grid. The density values ​​of the deformation points of each three-dimensional voxel mesh are smoothed in the neighborhood to obtain a smooth density distribution map, and the density peak region is determined based on the smooth density distribution map. All candidate deformation points contained within the density peak region are divided into the same deformation cluster, and the spatial centroid coordinates and spatial extension range of each deformation cluster are calculated. Based on the spatial extension range, determine whether the size of the deformation cluster exceeds the preset deformation size threshold. If it does, then the area covered by the deformation cluster is determined as the deformation concentration area. Deformation candidate points are selected evenly at preset intervals within the deformation concentration area as deformation detection nodes, and the spatial coordinates of the deformation detection nodes are numbered and stored according to the detection order.

6. The deformation detection system for aluminum alloy die castings according to claim 1, characterized in that, The step of obtaining the curvature change value corresponding to the deformation detection node based on the projection position of the deformation detection node on the fitted surface equation, and comparing the curvature change value with a preset curvature change range, includes: Obtain the spatial coordinates of each deformation detection node, and project the spatial coordinates of each deformation detection node onto the equation of the fitted surface along the normal vector direction of the fitted surface to obtain the projected coordinates of each deformation detection node. Based on the projected coordinates of each deformation detection node, calculate the principal curvature value and Gaussian curvature value of the fitted surface equation at that projected coordinate. The average curvature value is calculated based on the principal curvature value, and the ratio of the Gaussian curvature value to the average curvature value is used as the curvature change value corresponding to the deformation detection node. Obtain a preset curvature variation range, wherein the preset curvature variation range is composed of the upper limit value of the design curvature and the lower limit value of the design curvature of the aluminum alloy die casting to be tested; The curvature change value corresponding to each deformation detection node is compared with the preset curvature change range in turn to determine whether the curvature change value is less than the lower limit of the design curvature or greater than the upper limit of the design curvature.

7. The deformation detection system for aluminum alloy die castings according to claim 6, characterized in that, The upper and lower limits of the design curvature are obtained by multiplying the theoretical curvature value of the three-dimensional design model of the aluminum alloy die casting under test by a preset tolerance coefficient.

8. The deformation detection system for aluminum alloy die castings according to claim 1, characterized in that, If the curvature change value exceeds the preset curvature change range, then the deformation detection node corresponding to the curvature change value is obtained, and the spatial coordinates and timestamp of the deformation detection node are recorded, including: When it is determined that the curvature change value exceeds the preset curvature change range, the curvature change value is marked as an abnormal curvature change value, and the target deformation detection node that generated the abnormal curvature change value is obtained. Obtain the spatial coordinates of the target deformation detection node, and query the original acquisition time corresponding to the target deformation detection node in the continuous surface contour data based on the spatial coordinates; The original acquisition time is used as the timestamp of the target deformation detection node, and the timestamp is associated with and stored with the spatial coordinates of the target deformation detection node; The target deformation detection nodes are sorted according to the degree of deviation of the curvature change value from the preset curvature change range, and the target deformation detection node with the largest deviation is marked as the main deformation node according to the sorting result. Obtain the adjacent deformation detection nodes around the main deformation node, and combine the main deformation node and the adjacent deformation detection nodes to form a deformation feature point set.

9. The deformation detection system for aluminum alloy die castings according to claim 1, characterized in that, The step of generating deformation detection results based on the spatial coordinates and timestamps includes: Obtain the spatial coordinates and timestamps of all deformation detection nodes, and arrange the spatial coordinates in time sequence according to the timestamps to obtain the temporal spatial coordinate sequence of the deformation detection nodes; The spatial displacement vector between deformation detection nodes corresponding to adjacent timestamps is calculated based on the temporal spatial coordinate sequence, and the displacement change rate of each deformation detection node is calculated based on the spatial displacement vector. Based on the displacement change rate, cluster analysis is performed on the deformation detection nodes to divide the deformation detection nodes with the same displacement change rate into the same deformation evolution region. Obtain the spatial coordinate extreme values ​​of the deformation detection nodes within each deformation evolution region, and calculate the maximum deformation amplitude and deformation extension direction of each deformation evolution region based on the spatial coordinate extreme values; The maximum deformation amplitude and the deformation extension direction are used as deformation feature parameters, and deformation detection results are generated based on the deformation feature parameters of all deformation evolution regions.

10. The deformation detection system for aluminum alloy die castings according to claim 1, characterized in that, The preset scan path is generated as follows: A three-dimensional design model of the aluminum alloy die casting to be tested is obtained, and a surface curvature distribution map of the aluminum alloy die casting to be tested is extracted based on the three-dimensional design model. Based on the surface curvature distribution map, high curvature regions and low curvature regions on the surface of the aluminum alloy die casting to be tested are identified, wherein the scanning path density of the high curvature region is greater than that of the low curvature region; Multiple longitudinal scan lines and multiple transverse scan lines are set in the high curvature region, and the longitudinal scan lines and the transverse scan lines form a preset scan path in a cross-grid pattern in the high curvature region; Multiple parallel scanning lines are set in the low curvature region and extend along the length direction of the aluminum alloy die casting to be tested. The intersecting grid-like preset scanning paths and the parallel scanning lines are combined to obtain multiple preset scanning paths that cover the entire surface of the aluminum alloy die casting to be tested.

Images