Intelligent drawing method and system for parting surface of mold combined with appearance design of injection molded part

Abstract

The application provides a mold parting surface intelligent drawing method and system combined with the appearance design of an injection molded part, relates to the technical field of injection mold design, and first acquires an injection molded part appearance data set containing appearance curved surface units and appearance continuity transition regions; then performs parting line direction deduction processing on the injection molded part appearance data set to generate an initial parting line spatial position sequence; then performs parting surface starting boundary anchoring processing based on the initial parting line spatial position sequence to obtain a parting surface starting boundary curve set carrying starting boundary identifiers; then calls the parting surface starting boundary curve set and the initial parting line spatial position sequence to perform parting surface spatial extension generation processing to obtain a parting surface preliminary geometric model; and finally performs intersection avoidance pruning processing on the preliminary geometric model to obtain a target mold parting surface geometric model. The application can improve the accuracy and efficiency of mold parting surface drawing and improve the appearance quality of injection molded parts.

Description

Technical Field

[0001] This application relates to the field of injection mold design technology, and more specifically, to a method and system for intelligent drawing of mold parting surfaces that combines injection molded part appearance design. Background Technology

[0002] In the field of injection mold design and manufacturing, the drawing of the mold parting surface is a crucial step, as its quality directly affects the quality of the injection molded parts and the manufacturing efficiency of the mold. Traditional methods for drawing mold parting surfaces mainly rely on the experience of engineers and manual operation. Engineers need to carefully observe the appearance of the injection molded parts, rely on their professional knowledge and practical experience to conceive the general direction and position of the parting surface in their minds, and then draw it using drawing software.

[0003] However, the above methods have many drawbacks. On the one hand, the experience and skill levels of different engineers vary, resulting in inconsistent quality of the parting surfaces and making it difficult to guarantee their rationality and accuracy. On the other hand, for complex injection molded parts, such as those with multiple curved surface units and complex continuous transition areas, manually drawing parting surfaces is not only time-consuming and labor-intensive but also prone to errors, requiring repeated modifications and adjustments, which greatly reduces the efficiency of mold design. Furthermore, traditional methods fail to fully consider the spatial relationship between the parting surface and the continuous transition areas of the appearance, easily leading to interference between the parting surface and the transition areas, affecting the appearance quality of the injection molded parts. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method and system for intelligent drawing of mold parting surfaces that combines injection molded part appearance design.

[0005] According to a first aspect of this application, a method for intelligently drawing mold parting surfaces in conjunction with the appearance design of injection molded parts is provided, the method comprising: Obtain a set of appearance data for injection molded parts, wherein the set of appearance data includes appearance surface units and appearance continuity transition areas between appearance surface units; The parting line direction deduction process is performed on the injection molded part appearance data set, and an initial parting line spatial position sequence is generated based on the extreme value trajectory of the curvature change of the appearance surface unit and the width center trajectory of the appearance continuous transition region. Based on the initial parting line spatial position sequence, the appearance surface unit is subjected to parting surface starting boundary anchoring processing, and the parting surface starting boundary curve is determined on the appearance surface unit to obtain a set of parting surface starting boundary curves carrying the starting boundary identifier. The parting surface spatial extension generation process is performed by calling the set of parting surface initial boundary curves carrying the initial boundary identifier and the initial parting line spatial position sequence. Starting from the parting surface initial boundary curve, the process extends along the initial parting line spatial position sequence into the internal solid region of the injection molded part to generate parting surface spatial sheet units, thereby obtaining a preliminary geometric model of the parting surface composed of parting surface spatial sheet units. The parting surface preliminary geometric model is subjected to intersection avoidance trimming processing between the parting surface and the appearance continuity transition area. The part of the parting surface spatial sheet unit that intrudes into the appearance continuity transition area is cut off to obtain the target mold parting surface geometric model.

[0006] According to a second aspect of this application, a mold parting surface intelligent drawing system that incorporates the appearance design of an injection molded part is provided. The mold parting surface intelligent drawing system that incorporates the appearance design of an injection molded part includes a machine-readable storage medium and a processor. The machine-readable storage medium stores machine-executable instructions. When the processor executes the machine-executable instructions, the mold parting surface intelligent drawing system that incorporates the appearance design of an injection molded part implements the aforementioned mold parting surface intelligent drawing method that incorporates the appearance design of an injection molded part.

[0007] Based on any of the above aspects, the technical effect of this application is as follows: By acquiring a set of injection molded part appearance data containing surface units and continuous transition regions, parting line direction deduction processing is performed on the injection molded part appearance data set. Based on the extreme value trajectory of curvature change of the surface units and the width center trajectory of the continuous transition regions, an initial parting line spatial position sequence is generated. This determines the direction of the parting line, making its setting more consistent with the appearance characteristics and mechanical requirements of the injection molded part. Based on the initial parting line spatial position sequence, parting surface initial boundary anchoring processing is performed to determine the initial boundary curve of the parting surface, providing a clear starting position for parting surface generation and ensuring the accuracy and stability of parting surface generation. Qualitatively, the process involves calling the set of initial boundary curves of the parting surface carrying the initial boundary identifier and the spatial position sequence of the initial parting line to perform the spatial extension generation process of the parting surface. Starting from the initial boundary curve, the process extends into the internal solid area of ​​the injection molded part to generate spatial sheet units of the parting surface, thus obtaining the preliminary geometric model of the parting surface. This achieves automated generation of the parting surface and improves generation efficiency. Finally, the preliminary geometric model of the parting surface is subjected to intersection avoidance trimming to remove the parts that intrude into the continuous transition area of ​​the appearance, thus obtaining the geometric model of the target mold parting surface. This effectively avoids interference between the parting surface and the transition area of ​​the appearance, improving the appearance quality of the injection molded part and the manufacturing precision of the mold. Attached Figure Description

[0008] Figure 1 A flowchart illustrating the intelligent drawing method for mold parting surface combined with injection molded part appearance design provided in this application embodiment is shown. Figure 2 This illustration shows a schematic diagram of the component structure of an intelligent drawing system for mold parting surfaces that integrates the above-described intelligent drawing method for mold parting surface design, provided in an embodiment of this application, for implementing the intelligent drawing method for mold parting surface design. Detailed Implementation

[0009] Figure 1 This paper illustrates a flowchart of the intelligent drawing method and system for mold parting surfaces combined with injection molded part appearance design, as provided in an embodiment of this application. The detailed steps include: Step S110: Obtain the appearance data set of the injection molded part, wherein the appearance data set includes appearance surface units and appearance continuity transition areas between appearance surface units.

[0010] Step S111: Obtain the original three-dimensional appearance surface model of the injection molded product, perform appearance surface unit segmentation processing on the original three-dimensional appearance surface model, and divide the original three-dimensional appearance surface model into multiple non-overlapping appearance surface units according to the abrupt boundary of the surface normal vector and the zero intersection point of the surface Gaussian curvature in the original three-dimensional appearance surface model. Each appearance surface unit corresponds to a continuous appearance area of ​​the injection molded product.

[0011] In this embodiment, the original 3D appearance surface model is stored in a triangular mesh data structure, containing a set of mesh vertices V and a set of triangular faces F. Each vertex carries spatial coordinate components, and each triangular face carries three vertex indices and a face normal vector component. The vertex normal vector at each mesh vertex is calculated as follows: For any vertex Va, all triangular faces containing Va are traversed, and the face normal vectors of these triangular faces are summed and divided by the number of triangular faces to obtain the vertex normal vector of Va. Each edge of each triangular face is traversed. For edge Eab connecting vertices Va and Vb, the vertex normal vectors of Va and Vb are obtained. Simultaneously, the face normal vectors of the left and right triangular faces Fr of edge Eab are obtained. The cosine of the angle between the two vertex normal vectors and the cosine of the angle between the two face normal vectors are calculated. When any cosine of the angle is less than a preset abrupt change threshold, edge Eab is marked as a normal vector abrupt change boundary. Calculate the Gaussian curvature at each vertex: For vertex Vc, obtain all triangular faces within its one-ring neighborhood, solve for the eigenvalues ​​of the coefficient matrices of the first and second fundamental forms at Vc, obtain the principal curvature maxima and minima, and multiply the principal curvature maxima and minima to obtain the Gaussian curvature. Traverse all vertices and find the zero-crossing points where the Gaussian curvature changes from positive to negative or from negative to positive, and connect adjacent zero-crossing points to form a zero-crossing curve. Use the normal vector abrupt change boundary and the zero-crossing curve as the segmentation boundary, and use a region growing algorithm to segment the triangular mesh. Starting from a seed triangular face, expand to adjacent triangular faces along a path that does not cross the segmentation boundary until all triangular faces are classified, and each classified region forms an appearance surface unit.

[0012] Step S112: Perform appearance continuity transition region identification processing on two adjacent appearance surface units, extract the common boundary line between the two adjacent appearance surface units, extend a preset transition zone width into the appearance surface units on both sides with the common boundary line as the baseline, collect the normal vector angle sequence and the principal curvature ratio sequence of the surface points within the transition zone width, mark the transition zone region where the normal vector angle sequence changes continuously and the principal curvature ratio sequence changes monotonically as the appearance continuity transition region, and obtain the corresponding appearance continuity transition region for each adjacent appearance surface unit.

[0013] In this embodiment, for any two adjacent surface elements, a common boundary line is obtained between them, which is composed of a series of shared mesh edges. Using this common boundary line as a baseline, a preset transition band width is extended into the interiors of the first and second surface elements, forming a strip-shaped region covering both sides of the common boundary line. Multiple sampling points are set within this strip-shaped region according to a uniform sampling step size. For each sampling point, the normal vector component, principal curvature maxima, and principal curvature minima are calculated. Along a direction perpendicular to the common boundary line, the sampling points are sorted according to their distance from the common boundary line to obtain a sequence of normal vector angles, where each element is the angle between the normal vector of the sampling point and the normal vector of the corresponding point on the common boundary line. Simultaneously, a sequence of principal curvature ratios is obtained, where each element is the ratio of the principal curvature minima to the principal curvature maxima. When the difference between adjacent elements in the normal vector angle sequence is less than the preset change threshold, and the principal curvature ratio sequence shows a monotonically increasing or monotonically decreasing trend, the strip region is marked as an appearance continuity transition region, and each transition region is associated with and stored with its two corresponding appearance surface unit identifiers.

[0014] Step S113: Perform surface geometric attribute quantization processing on each appearance surface unit, calculate the normal vector coordinate components, principal curvature maxima and principal curvature minima at each sampling point position on each appearance surface unit, and generate the surface geometric attribute field of the appearance surface unit. The surface geometric attribute field includes the normal vector coordinate component values, principal curvature maxima values ​​and principal curvature minima values ​​of each sampling point.

[0015] In this embodiment, for each surface element, a uniform parameterization method is used to generate an M-row, N-column sampling grid within the surface parameter domain, resulting in multiple sampling points. For each sampling point, its position coordinates in three-dimensional space are calculated, and then the first and second fundamental form coefficients of the surface at that point are calculated. The principal curvature maxima and minima are obtained by solving the characteristic equation, and the normal vector components at that sampling point are also calculated. The normal vector components, principal curvature maxima, and principal curvature minima of each sampling point are organized into a six-dimensional feature vector, containing the three components of the normal vector, the principal curvature maxima, the principal curvature minima, and the identifier of the surface element to which the sampling point belongs. The six-dimensional feature vectors of all sampling points are arranged in the row and column order of the sampling grid to form the surface geometric property field of the surface element.

[0016] Step S114: Perform transition region width distribution acquisition processing on each appearance continuity transition region. Set multiple width measurement sampling lines along the direction perpendicular to the common boundary line within the appearance continuity transition region. Calculate the surface arc length from the common boundary line to the boundary of the transition region on each width measurement sampling line as the transition region width value at the corresponding sampling position, and obtain a transition region width distribution set composed of multiple transition region width values.

[0017] In this embodiment, for each transition region with a continuous appearance, its common boundary line is first determined. Multiple sampling positions are set at equal arc length intervals along the extension direction of the common boundary line. At each sampling position, a plane perpendicular to the common boundary line is drawn through the sampling point. This plane intersects the transition region to obtain a width measurement sampling line. Along this sampling line, starting from the sampling point on the common boundary line, the line extends to both sides along the curved surface, calculating the arc length of the curved surface reaching the boundary of the transition region, obtaining the left and right width values ​​respectively. The arithmetic mean of the left and right width values ​​is taken as the transition region width value at that sampling position. The transition region width values ​​corresponding to multiple sampling positions are arranged in the sampling order to obtain a set of transition region width distributions.

[0018] Step S115: Perform transition region curvature change acquisition processing on each appearance continuity transition region. Set multiple curvature sampling trajectories in the appearance continuity transition region along the direction parallel to the common boundary line. Record the normal vector angle change rate sequence and the principal curvature ratio change rate sequence on each curvature sampling trajectory to generate a transition region curvature change set containing the normal vector angle change rate sequence and the principal curvature ratio change rate sequence.

[0019] In this embodiment, for each transition region with a continuous appearance, multiple curvature sampling trajectories are set at equal intervals along a direction parallel to the common boundary line. Each trajectory intersects the transition region to form a curve. Multiple sampling points are set on this curve at equal arc length intervals. For each sampling point, the angle between the normal vector at that point and the normal vector of the corresponding projection point on the common boundary line is calculated, along with the ratio of the principal curvature minimum to the principal curvature maximum at that point. Along the extension direction of the sampling trajectory, the ratio of the change in the normal vector angle between adjacent sampling points to the arc length increment is calculated, resulting in a sequence of normal vector angle change rates; the ratio of the change in the principal curvature ratio between adjacent sampling points to the arc length increment is also calculated, resulting in a sequence of principal curvature ratio change rates. The two sequences corresponding to each curvature sampling trajectory are stored together to form a set of curvature changes in the transition region.

[0020] Step S116: The surface geometric attribute field of the appearance surface unit, the width distribution set of the transition region, and the curvature change set of the transition region are associated and stored to obtain a set of injection molded part appearance data containing the appearance surface unit identifier, the adjacent relationship identifier, and the transition region identifier.

[0021] In this embodiment, a data structure for establishing the injection molded part appearance data set is constructed, comprising an appearance surface unit dictionary, an adjacency relation dictionary, and a transition region dictionary. The key of the appearance surface unit dictionary is the appearance surface unit identifier, and the value is the surface geometric attribute field matrix of that unit. The key of the adjacency relation dictionary is the identifier of adjacent appearance surface unit pairs, and the value is the transition region identifier corresponding to that pair. The key of the transition region dictionary is the transition region identifier, and the value is the width distribution set and curvature change set corresponding to that transition region. Each surface geometric attribute field generated in step S113 is stored in the appearance surface unit dictionary, and the width distribution set and curvature change set generated in steps S114 and S115 are stored in the transition region dictionary. An adjacency relation dictionary is then established to associate these elements, resulting in a complete injection molded part appearance data set.

[0022] Step S120: Perform parting line direction deduction processing on the injection molded part appearance data set, and generate an initial parting line spatial position sequence based on the curvature change extreme value trajectory of the appearance surface unit and the width center trajectory of the appearance continuity transition region.

[0023] Step S121: Extract the surface geometric attribute field of each surface unit in the injection molded part appearance data set. Search for the spatial coordinates of surface points where the principal curvature maxima and minima are local maxima and local minima in the surface geometric attribute field. Connect the spatial coordinates of all the searched surface points to form a curve trajectory, generating the curvature change extreme value trajectory corresponding to each surface unit. The curvature change extreme value trajectory is composed of multiple extreme point spatial coordinates arranged in order of geodesic distance on the surface.

[0024] In this embodiment, the surface geometric attribute field of each appearance surface unit is traversed. This surface geometric attribute field is an M-row N-column grid matrix, and each grid unit stores a six-dimensional feature vector. A sliding window method is used to search for extreme points: for each grid point, the principal curvature maxima and minima of that point are obtained, along with the principal curvature maxima and minima of all grid points within its eight neighboring areas. When the principal curvature maxima of a point is greater than the principal curvature maxima of all points within its eight neighboring areas, and the principal curvature minima of a point is less than the principal curvature minima of all points within its eight neighboring areas, that grid point is marked as an extreme point and its spatial coordinates are recorded. All extreme points are sorted according to their geodesic distance on the surface. Starting from any extreme point, the unconnected extreme points with the nearest geodesic distance are connected sequentially to form a curve trajectory, thus obtaining the curvature change extreme value trajectory of the appearance surface unit.

[0025] Step S122: Extract the width distribution set of each transition region in the appearance continuity transition area of ​​the injection molded part appearance data set. Calculate the spatial coordinates of the midpoint of the width value of the transition region on each width measurement sampling line in the appearance continuity transition area. Connect all the spatial coordinates of the midpoints sequentially along the extension direction of the common boundary line to generate the width center trajectory corresponding to each appearance continuity transition area. The width center trajectory is composed of multiple midpoint spatial coordinates arranged in the order of the sampling lines.

[0026] In this embodiment, for each transition region, its width distribution set is extracted. This width distribution set contains the transition region width values ​​at multiple sampling locations, with each sampling location corresponding to a sampling point on the common boundary line. For each sampling location, the spatial coordinates of the sampling point and the left and right width values ​​at that point are obtained. Along the common boundary line in the normal plane at the sampling point, starting from the sampling point, half the left width value is moved to the left along the curved surface to obtain the first midpoint coordinates, and half the right width value is moved to the right along the curved surface to obtain the second midpoint coordinates. The arithmetic mean of the first and second midpoint coordinates is calculated as the midpoint spatial coordinates of the sampling location. Multiple midpoint spatial coordinates are connected sequentially according to the sampling order to form the width center trajectory of the transition region.

[0027] Step S123: Perform trajectory intersection point solving processing on the width center trajectory of the shared appearance continuity transition region between adjacent appearance surface units and the curvature change extreme value trajectory of each of the two adjacent appearance surface units. Calculate the spatial straight-line distance between the endpoint spatial coordinates of the width center trajectory and the endpoint spatial coordinates of each curvature change extreme value trajectory. When the spatial straight-line distance is less than a preset distance threshold, connect the endpoint of the width center trajectory with the endpoint of the curvature change extreme value trajectory to obtain the parting line backbone trajectory network containing trajectory intersection points.

[0028] In this embodiment, for adjacent surface unit pairs, the first curvature change extreme value trajectory of the first unit, the second curvature change extreme value trajectory of the second unit, and the width center trajectory of the shared transition region between them are obtained. The coordinates of the two endpoints of the width center trajectory, the two endpoints of the first curvature change extreme value trajectory, and the two endpoints of the second curvature change extreme value trajectory are extracted. The spatial straight-line distance between the first endpoint of the width center trajectory and each endpoint of the first curvature change extreme value trajectory is calculated. When this distance is less than a preset threshold, the two endpoints are connected to form an intersection point; similarly, the connection between the second endpoint of the width center trajectory and the second curvature change extreme value trajectory is processed. All connected trajectory segments are merged according to the intersection point to form a fractal line backbone trajectory network.

[0029] Step S124: Perform trajectory smoothing optimization processing on the main trajectory network of the fractal line. Calculate the angle between the tangent vectors of adjacent trajectory segments at each trajectory intersection point in the main trajectory network of the fractal line. When the angle between the tangent vectors is greater than a preset angle threshold, insert a Bezier curve transition segment near the trajectory intersection point to replace the original broken line segment, thereby obtaining the smoothed main trajectory network of the fractal line.

[0030] In this embodiment, each trajectory intersection point in the fractal line backbone trajectory network is traversed to obtain all adjacent trajectory segments at that point. For any two adjacent trajectory segments, the first tangent vector of the first segment at the intersection point is calculated, the second tangent vector of the second segment at the intersection point is calculated, and the angle between the two tangent vectors is calculated. When the angle is greater than a preset smoothing angle threshold, control points are taken at a preset distance along the direction of the first segment, centered on the intersection point, and control points at the same distance along the direction of the second segment are taken. Using the intersection point as the starting point and the two control points as control points, a Bézier curve transition segment is generated. The original polyline segment is replaced with this Bézier curve segment to obtain the smoothed fractal line backbone trajectory network.

[0031] Step S125: Perform trajectory spatial sorting processing on the smoothed parting line backbone trajectory network. Calculate the spatial projection length of each trajectory segment with the demolding direction of the injection molded part as the reference direction. Sort the trajectory segments in descending order of spatial projection length, and extract the spatial coordinates of each trajectory sampling point along the sorted trajectory segments to generate an initial parting line spatial position sequence according to the priority of the demolding direction.

[0032] In this embodiment, the demolding direction vector of the injection molded part is obtained. Each trajectory segment in the smoothed trajectory network, consisting of multiple trajectory sampling points, is traversed. The spatial coordinates of each sampling point are projected along the demolding direction vector, and the difference between the projected length of the last sampling point and the projected length of the first sampling point on the trajectory segment is calculated as the spatial projection length of that segment. All trajectory segments are sorted in descending order of projection length. The spatial coordinates of all sampling points on each trajectory segment are extracted sequentially according to the sorted order, forming a spatial coordinate sequence arranged according to the priority of the demolding direction, which serves as the initial parting line spatial position sequence.

[0033] Step S130: Based on the initial parting line spatial position sequence, perform parting surface starting boundary anchoring processing on the appearance surface unit, determine the parting surface starting boundary curve on the appearance surface unit, and obtain a set of parting surface starting boundary curves carrying the starting boundary identifier.

[0034] Step S131: Extract the spatial coordinates of the first trajectory sampling point from the initial parting line spatial position sequence as the starting anchor point spatial coordinates. Generate a starting boundary search circle on the appearance surface unit to which the starting anchor point spatial coordinates belong, with the starting anchor point spatial coordinates as the center, in the tangent plane of the appearance surface unit. The starting boundary search circle intersects with the surface of the appearance surface unit to form a starting boundary closed curve.

[0035] In this embodiment, the spatial coordinates of the first trajectory sampling point are read from the initial parting line spatial position sequence as the starting anchor point. The appearance surface unit to which the anchor point belongs is determined, and the equation of the tangent plane and the normal vector of the unit at the anchor point are obtained. Multiple circumferential points are generated at equal angular intervals on a circle with the anchor point as the center and a preset radius as the center. The coordinates of each circumferential point are offset along two orthogonal directions in the tangent plane. For each circumferential point, a ray is drawn along the normal vector direction, and the intersection point of the ray with the appearance surface unit is calculated. All intersection points are connected to form a closed curve to obtain the initial boundary closed curve.

[0036] Step S132: Traverse each trajectory sampling point sequentially along the initial parting line spatial position sequence, perform the initial boundary curve extension direction determination process at the spatial coordinates of each trajectory sampling point, calculate the spatial vector between the current trajectory sampling point and the next trajectory sampling point as the direction vector of the current trajectory segment, and project the direction vector onto the tangential plane of the appearance surface unit to which the current trajectory sampling point belongs to obtain the tangential extension direction of the initial boundary curve at the current trajectory sampling point.

[0037] In this embodiment, for the i-th trajectory sampling point Pi in the initial parting line spatial position sequence, the next trajectory sampling point P(i+1) is obtained, and the spatial vector from Pi to P(i+1) is calculated. The normal vector of the appearance surface unit to which Pi belongs at Pi is obtained, and the projection component of the spatial vector on the tangent plane is calculated. The normalized projection component is used as the tangential extension direction vector of the starting boundary curve at Pi.

[0038] Step S133: Perform dynamic adjustment of the width of the starting boundary curve at each trajectory sampling point, obtain the principal curvature maxima and minima of the current trajectory sampling point in the surface geometric attribute field of the appearance surface unit to which the current trajectory sampling point belongs, calculate the ratio of the principal curvature maxima to the principal curvature minima as the surface curvature degree index, query the preset width mapping relationship according to the surface curvature degree index, and determine the local width value of the starting boundary curve at the current trajectory sampling point, wherein the width mapping relationship is configured such that the larger the surface curvature degree index, the smaller the determined local width value.

[0039] In this embodiment, for each trajectory sampling point Pi, the principal curvature maxima and minima at Pi are extracted from the surface geometric attribute field of the corresponding appearance surface unit, and their ratio is calculated as an index of surface curvature. A preset width mapping function is used, which is in the form of exponential decay. When the curvature index increases, the local width value decreases exponentially. The local width value at Pi is calculated based on this mapping relationship.

[0040] Step S134: At each trajectory sampling point, generate a local segment of the starting boundary curve based on the starting boundary search circle, the tangential extension direction, and the local width value. With the current trajectory sampling point as the center, extend the local width value to both sides along the direction perpendicular to the tangential extension direction to obtain two boundary control points on the surface of the appearance surface unit. Connect the current trajectory sampling point and the two boundary control points with a spline curve to generate a local segment of the starting boundary curve at the current trajectory sampling point.

[0041] In this embodiment, for each trajectory sampling point Pi, the tangential extension direction vector at that point is obtained, and two orthogonal direction vectors in the normal plane perpendicular to that direction are calculated. A first boundary control point is obtained by moving Pi along one of these directions by half the local width value, and a second boundary control point is obtained by moving it the same distance in the opposite direction. These two control points are then projected onto the surface of the appearance surface unit. Using Pi as the midpoint and the two control points as the two side control points, a curve segment passing through these three points is generated using quadratic spline interpolation, serving as a local segment of the initial boundary curve at Pi.

[0042] Step S135: Separately stitch together the local segments of the starting boundary curve generated at all trajectory sampling points in the order of the trajectory sampling points, and perform tangential continuity adjustment at the splicing points of adjacent local segments to make the tangent vector directions of the common endpoints of adjacent local segments consistent, thereby generating a continuous parting surface starting boundary curve.

[0043] In this embodiment, all local segments are arranged in order of trajectory sampling points. For two adjacent local segments at a common endpoint, the first tangent vector of the first segment at the endpoint and the second tangent vector of the second segment at the endpoint are calculated, and the angle between them is calculated. When the angle is not zero, the position of the control point of the second segment near the endpoint is adjusted so that the tangent vector of the second segment at the endpoint is rotated to be consistent with the direction of the first tangent vector. After adjusting all adjacent splicing points, all local segments are connected into a continuous spline curve to obtain the starting boundary curve of the parting surface.

[0044] Step S136: Perform start boundary identifier addition processing on the start boundary curve of the parting surface. Add the unit identifier of the appearance surface unit and the sequential number of the trajectory sampling point to each curve segment of the start boundary curve of the parting surface to obtain a set of start boundary curves of the parting surface carrying start boundary identifiers.

[0045] In this embodiment, the initial boundary curve of the parting surface is segmented according to the trajectory sampling points, and the curve segment corresponding to each trajectory sampling point is an independent curve segment. For the i-th curve segment, the element identifier of the appearance surface unit to which the segment belongs and the sequential number of the trajectory sampling point in the initial parting line spatial position sequence are obtained. The two are combined into an initial boundary identifier string and appended to the corresponding curve segment. All curve segments carrying the identifier constitute the set of initial boundary curves of the parting surface.

[0046] Step S140: Call the set of parting surface initial boundary curves carrying the initial boundary identifier and the initial parting line spatial position sequence to perform parting surface spatial extension generation process. Starting from the parting surface initial boundary curve, extend along the initial parting line spatial position sequence to the internal solid area of ​​the injection molded part to generate parting surface spatial sheet units, and obtain the preliminary geometric model of the parting surface composed of parting surface spatial sheet units.

[0047] Step S141: Extract the set of spatial coordinate points of each curve segment of the parting surface starting boundary curve from the set of parting surface starting boundary curves carrying the starting boundary identifier, and use the set of spatial coordinate points of each curve segment as the starting contour line of the parting surface spatial extension to obtain the set of starting contour lines. Each starting contour line corresponds to a parting surface starting boundary curve segment on an appearance surface unit.

[0048] In this embodiment, each curve segment in the set of initial boundary curves of the parting surface is traversed, and multiple sampling points are extracted for each curve segment at equal arc length intervals. Each sampling point carries spatial coordinates and arc length parameters. The spatial coordinates of all sampling points are arranged in order of arc length parameters to form a set of spatial coordinate points. This set of points is used as an initial contour line, and all initial contour lines constitute the set of initial contour lines.

[0049] Step S142: Perform initial contour discretization processing on each initial contour line in the initial contour line set, extract contour sampling points at equal intervals along the arc length direction of each initial contour line, and obtain the contour sampling point sequence corresponding to each initial contour line. Each contour sampling point carries the spatial coordinates of the contour sampling point and the arc length parameter on the initial contour line.

[0050] In this embodiment, for each starting contour line, the total arc length of the contour line is calculated, the number of samples is set, and the sampling interval is calculated. Starting from the starting point of the contour line, one contour sampling point is extracted at each sampling interval, for a total of a set number of sampling points. Each sampling point carries spatial coordinates and arc length parameters, and all sampling points are arranged in ascending order of arc length parameters to form a contour sampling point sequence.

[0051] Step S143: Perform parting surface extension direction vector calculation processing on each contour sampling point, obtain the surface unit normal vector of the appearance surface unit to which the contour sampling point belongs at the position of the contour sampling point, obtain the trajectory direction vector of the contour sampling point at the corresponding position in the initial parting line spatial position sequence, and calculate the vector product of the surface unit normal vector and the trajectory direction vector as the initial extension direction vector of the parting surface at the contour sampling point.

[0052] In this embodiment, for each contour sampling point, its corresponding surface unit is determined, and the surface unit normal vector of that unit at that point is obtained. The trajectory sampling point corresponding to the contour sampling point is found in the initial parting line spatial position sequence, and the trajectory direction vector at that trajectory sampling point is obtained. The three-dimensional cross product of the surface unit normal vector and the trajectory direction vector is calculated, and the result is normalized to obtain the initial extension direction vector.

[0053] Step S144: Perform an extension path generation process for each contour sampling point. Starting from the contour sampling point, generate an extension ray along the initial extension direction vector towards the internal solid area of ​​the injection molded part. Calculate the first intersection point between the extension ray and the boundary of the internal solid area of ​​the injection molded part. Record the spatial line segment from the starting point to the first intersection point as the extension path line segment of the contour sampling point, and obtain the extension path line segment corresponding to each contour sampling point.

[0054] In this embodiment, for each contour sampling point, a ray is generated along the initial extension direction vector starting from that point to obtain the triangular mesh surface representation of the boundary of the internal solid region of the injection molded part. The intersection point of the ray and the boundary surface is calculated, and the orthogonal point with the smallest parameter value is found as the first intersection point. The spatial line segment from the contour sampling point to this intersection point is recorded as the extension path line segment of the contour sampling point.

[0055] Step S145: Perform extended path surface skinning on the extended path segments corresponding to all contour sampling points. Align the corresponding positions of the extended path segments of adjacent contour sampling points according to the arc length parameter and connect them into quadrilateral mesh units. Splice all quadrilateral mesh units to generate a continuous spatial surface sheet, which serves as the parting surface spatial sheet unit obtained from the initial contour line.

[0056] In this embodiment, for a sequence of contour sampling points corresponding to the same starting contour line, each sampling point corresponds to an extended path segment. Each extended path segment is divided into multiple segments according to its arc length, resulting in multiple division points. For two adjacent contour sampling points, the division points with the same segment index on their extended path lines are connected, and adjacent division points on the same extended path line are also connected to form quadrilateral mesh units. All adjacent contour sampling points and all segments are traversed to generate all quadrilateral mesh units, which are then spliced ​​together in their original positions to form a continuous spatial curved surface sheet, serving as the parting surface spatial sheet unit extended from the starting contour line.

[0057] Step S146: Assemble all the parting surface spatial sheet elements corresponding to the initial contour lines according to the arrangement order of the initial contour lines on the initial boundary curve of the parting surface, merge the overlapping mesh nodes at the common boundary of adjacent parting surface spatial sheet elements, and generate a preliminary geometric model of the parting surface composed of multiple seamlessly connected parting surface spatial sheet elements.

[0058] In this embodiment, all parting surface spatial sheet elements are sorted according to the arrangement order of their corresponding initial contour lines on the original parting surface initial boundary curve. For the common boundary of two adjacent sheet elements, each mesh node on the common boundary is traversed, and corresponding nodes with the same spatial coordinates are found in adjacent sheet elements. The two nodes are merged into one node, and the node index of all mesh elements containing these nodes is updated. After all common boundary nodes are merged, all sheet elements are connected into a whole to form the preliminary geometric model of the parting surface.

[0059] Step S150: Perform intersection avoidance trimming on the parting surface and the appearance continuity transition area of ​​the preliminary geometric model of the parting surface, and cut off the part of the parting surface spatial sheet unit that invades the appearance continuity transition area to obtain the geometric model of the target mold parting surface.

[0060] Step S151: Obtain the spatial boundary description of each appearance continuity transition region in the injection molded part appearance data set. The spatial boundary description includes the spatial coordinate point sequence of the boundary contour line of each appearance continuity transition region on the appearance surface of the injection molded part.

[0061] In this embodiment, the spatial boundary description of each transition region is extracted from the injection molded part appearance data set. Each transition region has a first boundary contour line near the first appearance surface unit and a second boundary contour line near the second appearance surface unit. Spatial coordinate points are extracted for each boundary contour line at equal arc length intervals to form a spatial coordinate point sequence. The point sequences of the two boundary contour lines are associated and stored as the spatial boundary description of the transition region.

[0062] Step S152: Traverse each parting surface spatial sheet element in the preliminary geometric model of the parting surface, extract the spatial coordinates of all mesh nodes of the parting surface spatial sheet element, compare the spatial coordinates of each mesh node with the spatial boundary description of each appearance continuity transition area, and determine whether each mesh node is located within the internal space of any appearance continuity transition area.

[0063] In this embodiment, for each parting surface spatial sheet element, the coordinates of all its mesh nodes are obtained. For each mesh node, the spatial boundary descriptions of all transition regions are traversed. For each transition region, its closed boundary is constructed, and a ray casting method is used to determine whether a node is located inside the closed boundary: a ray is drawn from the node in any direction, and the number of intersections between the ray and the closed boundary is calculated. If the number of intersections is odd, the node is located inside; if it is even, it is located outside. When a node is located inside the closed boundary of any transition region, the node is marked as an intruding node.

[0064] Step S153: When at least one mesh node in any parting surface space sheet element is located within the internal space of the appearance continuity transition region, mark the parting surface space sheet element as a parting surface space sheet element to be trimmed.

[0065] In this embodiment, for each parting surface space sheet element, the intrusion markers of all its mesh nodes are checked. If at least one node is marked as an intrusion node, the sheet element is marked as a parting surface space sheet element to be trimmed and added to the trimming list; if there are no intrusion nodes, the sheet element is left unchanged.

[0066] Step S154: Perform intrusion region boundary extraction processing on each parting surface spatial sheet element to be trimmed, calculate the intersection coordinates of all mesh edges of the parting surface spatial sheet element to be trimmed with the spatial boundary contour line of the appearance continuity transition area, arrange all intersection coordinates in the order of their appearance on the spatial boundary contour line to form an intrusion region boundary curve, and obtain the set of intrusion region boundary curves.

[0067] In this embodiment, for each piece of body to be trimmed, all its mesh edges are obtained. For each mesh edge, the point sequence of the spatial boundary contour lines of all transition regions is traversed, and a line segment intersection algorithm is used to determine whether the mesh edge intersects with the contour line. If an intersection point exists, the coordinates of the intersection point and the identifier of the corresponding contour line are recorded. All intersection points are sorted according to their arc length parameter on their respective contour lines, and then connected sequentially to form a closed boundary curve of the intrusion region. All curves constitute the set of boundary curves of the intrusion region.

[0068] Step S155: Perform the intrusion region internal mesh cell deletion process on each parting surface spatial sheet unit to be trimmed. Divide the parting surface spatial sheet unit to be trimmed into a set of mesh cells in the retained region and a set of mesh cells in the region to be deleted, using the boundary curve of the intrusion region as the dividing line. Delete all mesh cells in the set of mesh cells in the region to be deleted from the parting surface spatial sheet unit to be trimmed.

[0069] Step S1551: Obtain each intrusion region boundary curve from the set of intrusion region boundary curves of the spatial sheet unit to be trimmed, and extract the spatial coordinate point sequence of each intrusion region boundary curve and the spatial coordinate point sequence of the spatial boundary contour of the appearance continuity transition area corresponding to the intrusion region boundary curve.

[0070] In this embodiment, for each piece of sheet to be trimmed, a set of boundary curves of its intrusion region is obtained. For each boundary curve of the intrusion region in the set, the spatial coordinates of all intersection points on the curve are extracted and connected in order to form a point sequence. At the same time, the spatial boundary contour line point sequence of the transition region corresponding to the curve is obtained.

[0071] Step S1552: Perform mesh cell classification traversal processing on the spatial sheet element of the parting surface to be trimmed, sequentially obtain each mesh element in the spatial sheet element of the parting surface to be trimmed, extract the spatial coordinates of all mesh nodes of each mesh element, and determine the spatial position relationship between the spatial coordinates of each mesh node and the boundary curve of the intrusion area.

[0072] In this embodiment, each mesh cell in the unit to be trimmed is traversed to obtain the coordinates of all mesh nodes in that cell. For each node, the positional relationship with the boundary curve of the intrusion area is determined using ray casting: a ray is drawn from the node as the starting point, and the number of intersections with the boundary curve is calculated. An odd number indicates that the node is located inside the boundary curve, and an even number indicates that it is located outside. At the same time, the positional relationship between the node and the spatial boundary contour line of the transition area is determined to determine whether the node is located inside or outside the transition area.

[0073] Step S1553: When the spatial coordinates of all grid nodes of a grid cell are located on the same side of the boundary curve of the intrusion area and that side is in a direction away from the center of the appearance continuity transition area, mark the grid cell as a reserved area grid cell and add it to the reserved area grid cell set.

[0074] In this embodiment, for a grid cell, the position determination results of all its nodes are checked. If all nodes are located on the same side of the boundary curve of the intrusion area, and that side corresponds to the direction away from the center of the transition area, then the grid cell is determined to be completely in the retainable area, and it is marked as a retainable area grid cell and added to the retainable area grid cell set.

[0075] Step S1554: When the spatial coordinates of at least one grid node in a grid cell are located on the side of the boundary curve of the intrusion area that is close to the center of the appearance continuity transition area, mark the grid cell as a candidate grid cell to be deleted and add it to the candidate set to be deleted.

[0076] In this embodiment, for a grid cell, if at least one node is located on the side of the boundary curve of the intrusion area that is close to the center of the transition area, it is determined that the grid cell may be partially or entirely in the area to be deleted, and it is marked as a candidate grid cell to be deleted and added to the candidate set to be deleted.

[0077] Step S1555: Perform intersection judgment processing between each candidate mesh cell to be deleted and the boundary curve of the intrusion region in the candidate set to be deleted. Calculate the number of intersection points between each mesh edge of the candidate mesh cell to be deleted and the boundary curve of the intrusion region. When the number of intersection points is greater than zero, extract the coordinates of the intersection points located on the boundary curve of the intrusion region and the coordinates of the mesh nodes located on the side of the boundary curve of the intrusion region closer to the center of the appearance continuity transition area from the candidate mesh cell to be deleted.

[0078] In this embodiment, for each candidate grid cell in the candidate set to be deleted, all its grid edges are traversed. For each grid edge, the intersection point with the boundary curve of the intrusion region is calculated. The grid edge is represented as a parametric equation, and the boundary curve of the intrusion region is represented as a set of piecewise linear segments. The system of equations is solved to obtain the intersection point parameters. When all the intersection point parameters are within the valid range, the intersection point coordinates are recorded. At the same time, the coordinates of all grid nodes located on the side of the boundary curve of the intrusion region closer to the center of the transition region are recorded.

[0079] Step S1556: Based on the extracted intersection coordinates and grid node coordinates, perform grid cell subdivision processing on the candidate grid cells to be deleted. Using the original grid nodes and newly added intersection coordinates of the candidate grid cells to be deleted as vertices, the candidate grid cells to be deleted are re-divided into multiple sub-grid cells. Among them, the sub-grid cells located on the side of the boundary curve of the intrusion area that is closer to the center of the appearance continuity transition area are marked as the grid cells to be deleted and added to the set of grid cells to be deleted. The sub-grid cells located on the side of the boundary curve of the intrusion area that is farther away from the center of the appearance continuity transition area are marked as the grid cells to be retained and added to the set of grid cells to be retained.

[0080] In this embodiment, for each candidate mesh element, its original set of mesh nodes and set of intersection coordinates are obtained and merged into a new set of vertices. Based on the candidate mesh element type, a constrained triangulation method is used to re-divide the region using the boundary curve of the intrusion area as the dividing boundary. For each newly generated sub-mesh element, the positional relationship between its center point and the boundary curve is determined. If the center point is located closer to the center of the transition region, it is marked as a mesh element to be deleted and added to the set of mesh elements to be deleted; if it is located further away, it is marked as a mesh element to be retained and added to the set of mesh elements to be retained.

[0081] Step S1557: When all the mesh nodes of the candidate mesh cell to be deleted are located on the side of the boundary curve of the intrusion area that is close to the center of the appearance continuity transition area and no mesh edge intersects with the boundary curve of the intrusion area, mark the candidate mesh cell to be deleted as a mesh cell to be deleted and add it to the set of mesh cells to be deleted.

[0082] In this embodiment, for a candidate mesh cell, if all its mesh nodes are located on the side of the boundary curve of the intrusion area close to the center of the transition area, and no mesh edge is detected to intersect with the boundary curve in step S1555, then the mesh cell is determined to be completely located inside the area to be deleted, and it is directly marked as a mesh cell of the area to be deleted and added to the set of mesh cells of the area to be deleted.

[0083] Step S1558: Delete all mesh cells from the set of mesh cells in the region to be deleted from the parting surface space sheet element to be trimmed, and retain all mesh cells in the set of mesh cells in the region to be retained as the trimmed parting surface space sheet element.

[0084] In this embodiment, for each sheet element to be trimmed, all mesh elements in the set of mesh elements in the region to be deleted are removed from its mesh data structure, while all mesh elements in the set of mesh elements in the region to be retained are kept. Mesh nodes isolated due to the removal of mesh elements are deleted, and the connectivity of the remaining mesh elements is updated to obtain the trimmed parting surface space sheet element.

[0085] Step S156: Merge all trimmed parting surface space sheet elements with the parting surface space sheet elements not marked as to be trimmed to generate a target mold parting surface geometry model that does not contain any intrusive parts of the appearance continuity transition region.

[0086] In this embodiment, all trimmed parting surface spatial sheet elements and all sheet elements not marked as to be trimmed are collected. The sheet elements are merged according to their original positions, and node merging is performed at the common boundary of adjacent sheet elements to eliminate duplicate mesh nodes and mesh edges, resulting in a complete target mold parting surface geometric model. In this target mold parting surface geometric model, all parting surface spatial sheet elements do not spatially intrude into any appearance continuity transition area.

[0087] Furthermore, the method may also include: step S210: performing gap filling processing on the parting surface boundary and appearance continuity transition area of ​​the target mold parting surface geometric model, and extracting the spatial coordinate point sequence of each trimmed boundary curve in the target mold parting surface geometric model and the spatial coordinate point sequence of the spatial boundary contour line of the appearance continuity transition area corresponding to the trimmed boundary curve.

[0088] In this embodiment, there may be a gap between the trimmed boundary curve of the target mold parting surface geometry and the spatial boundary contour of the appearance continuity transition area. The spatial coordinate point sequence Qa=[Q1, Q2, ..., Qn] of each trimmed boundary curve Qx in the target mold parting surface geometry is extracted. The spatial coordinate point sequence Hb=[H1, H2, ..., Hm] of the spatial boundary contour of the appearance continuity transition area corresponding to this trimmed boundary curve is obtained from the injection molded part appearance data set.

[0089] Step S220: Calculate the shortest spatial distance between each trimming boundary point on the trimming boundary curve and the spatial boundary contour line of the appearance continuity transition area. When the shortest spatial distance is greater than zero, generate a filled connection line between the trimming boundary point and the nearest point on the spatial boundary contour line.

[0090] In this embodiment, for each trimmed boundary point Qj on the trimmed boundary curve Qx, each point Hk in the spatial coordinate point sequence Hb of the spatial boundary contour line of the appearance continuity transition region is traversed, and the Euclidean distance D_jk = sqrt((Xqj-Xhk)^2+(Yqj-Yhk)^2+(Zqj-Zhk)^2) between Qj and Hk is calculated. The minimum value Dm_j among all D_jk is taken as the shortest spatial distance between Qj and the spatial boundary contour line. When Dm_j is greater than zero, the corresponding point Hm that makes D_jk minimum is found, and a straight line segment is generated between Qj and Hm as the fill connection line Lj.

[0091] Step S230: Construct a gap-filled surface sheet based on all fill connection lines, using the trimmed boundary curve as the first boundary, the spatial boundary contour line of the appearance continuity transition area as the second boundary, and all fill connection lines as guide lines, and construct a gap-filled surface sheet connecting the trimmed boundary curve and the spatial boundary contour line through the lofted surface generation method.

[0092] In this embodiment, the trimmed boundary curve Qx is used as the first boundary curve of the lofted surface, the spatial boundary contour line Hb of the appearance continuity transition area is used as the second boundary curve of the lofted surface, and all the filling connection lines Lj generated in step S220 are used as the guide lines of the lofted surface. The lofted surface generation method is as follows: the first and second boundary curves are parameterized according to arc length to establish a parameter correspondence; linear interpolation is performed between each pair of corresponding parameter points along the guide line direction to generate a series of intermediate cross-section curves; and all intermediate cross-section curves are skinned to generate a continuous surface sheet, which is the gap-filled surface sheet Ft.

[0093] Step S240: Merge the gap-filling curved surface sheet with the target mold parting surface geometric model, and merge the overlapping mesh nodes on the boundary of the gap-filling curved surface sheet and the trimmed boundary curve of the target mold parting surface geometric model to generate the target mold parting surface geometric model after gap filling.

[0094] In this embodiment, a gap-filling surface sheet Ft is added to the target mold parting surface geometry model. The boundary of Ft perfectly coincides with the trimmed boundary curve Qx of the target mold parting surface geometry model. For each trimmed boundary point Qj on Qx, the corresponding boundary point on Ft is found, and these two points are merged into a single mesh node. The node index of all mesh cells containing this node is then updated. After merging all boundary points, the gap-filling surface sheet is seamlessly connected to the target mold parting surface geometry model, generating the gap-filled target mold parting surface geometry model.

[0095] Step S250: Perform contact boundary smoothing processing on the target mold parting surface geometry model after gap filling and the continuous transition area of ​​appearance. Extract all boundary lines in the target mold parting surface geometry model after gap filling that are in contact with the continuous transition area of ​​appearance. Calculate the angle between the surface tangent vector of the target mold parting surface geometry model and the surface tangent vector of the continuous transition area of ​​appearance at each boundary point on each boundary line. When the angle is greater than a preset smoothing angle threshold, rotate and adjust the surface normal vector of the target mold parting surface geometry model at the boundary point so that the angle between the surface tangent vector of the adjusted target mold parting surface geometry model and the surface tangent vector of the continuous transition area of ​​appearance is reduced to below the smoothing angle threshold. Obtain the target mold parting surface geometry model after contact boundary smoothing and output it as the final mold parting surface geometry model.

[0096] In this embodiment, all boundary lines Cb in contact with the appearance continuity transition area in the geometric model of the target mold parting surface after gap filling are extracted. For each contact boundary line, multiple boundary points Bi are sampled on the line at equal arc length intervals. For each boundary point Bi, the first surface tangent vector T1 of the target mold parting surface geometric model at Bi is calculated, the second surface tangent vector T2 of the appearance continuity transition area at Bi is calculated, and the angle θi between T1 and T2 is calculated as θi = arccos((T1·T2) / (|T1| |T2|)). A preset smoothing angle threshold θt is used. When θi is greater than θt, the surface normal vector N1 of the target mold parting surface geometry model at point Bi is rotated and adjusted: the rotation axis A = T1 × T2 is calculated, the rotation angle Δθ = θi - θt is calculated, and N1 is rotated around A by Δθ to obtain the adjusted normal vector N2. The positions of the mesh nodes around Bi are recalculated based on the adjusted normal vector, causing the surface tangent vector to change accordingly. After adjusting all boundary points, the target mold parting surface geometry model with smoothed contact boundaries is obtained. This model is output as the final mold parting surface geometry model.

[0097] Step S310: Obtain all parting surface spatial sheet elements in the final mold parting surface geometric model, perform parting surface thickness direction extension processing on each parting surface spatial sheet element, extend the preset thickness extension distance along the surface normal vector direction of the mesh node to the internal solid area of ​​the injection molded part with each mesh node as the reference point, and generate the thickness extension endpoint spatial coordinates corresponding to each mesh node.

[0098] In this embodiment, all parting surface spatial sheet elements in the final mold parting surface geometric model are obtained. For each mesh node Nd in any parting surface spatial sheet element, its spatial coordinates are (Xd, Yd, Zd). The surface unit normal vector (Nx, Ny, Nz) at that mesh node is obtained from the surface geometric attribute field generated in step S113. The thickness extension distance parameter H is set, which is dynamically determined according to the overall size ratio of the injection molded part. The thickness extension endpoint spatial coordinates Pe = (Xd + Nx) are obtained by moving a distance H from the mesh node Nd along the surface unit normal vector direction. H, Yd+Ny H, Zd+Nz H).

[0099] Step S320: Connect each mesh node with its corresponding thickness extension endpoint spatial coordinates to form a thickness direction line segment. Construct a thickness-extended solid mesh unit based on the mesh topology relationship of all thickness direction line segments and the original parting surface spatial sheet unit. Each thickness-extended solid mesh unit is formed by a mesh unit of the original parting surface spatial sheet unit and its corresponding mesh unit obtained by projection along the thickness direction.

[0100] In this embodiment, for each original mesh unit Ce in the original parting surface space sheet unit, the original mesh unit is composed of multiple original mesh nodes. The spatial coordinates of the thickness extension endpoint corresponding to each original mesh node are obtained, and these coordinates are connected according to the connection order of the original mesh nodes to form a corresponding projected mesh unit Cp. Each original mesh node of the original mesh unit Ce is connected to the corresponding projected node in the projected mesh unit Cp using straight line segments to form a side quadrilateral. The original mesh unit Ce, the projected mesh unit Cp, and all the side quadrilaterals together form a hexahedral solid mesh unit. The corresponding solid mesh unit is generated by traversing all the original mesh units.

[0101] Step S330: Perform solid mesh unit combination processing on all thickness-expanded solid mesh units, merge the common surfaces of adjacent thickness-expanded solid mesh units, delete duplicate mesh nodes on the common surfaces, and generate a continuous solid geometric model composed of solid mesh units.

[0102] In this embodiment, all solid mesh elements generated in step S320 are collected. For any two adjacent solid mesh elements, they share a common surface. All mesh nodes on this common surface are identified, and these mesh nodes are stored in one copy in each of the two solid mesh elements. The two copies of duplicate mesh nodes are merged into one node, and the node indices of all mesh elements containing these nodes in both solid mesh elements are updated. At the same time, duplicate mesh edges on the common surface are deleted. The common surfaces of all adjacent solid mesh elements are traversed to complete the merging process of all duplicate nodes and duplicate edges, generating a continuous solid geometry model Md composed of seamlessly connected solid mesh elements.

[0103] Step S340: Perform interference check processing on the parting surface entity and the internal structure of the injection molded part on the continuous solid geometry model, obtain the boundary description of the internal solid region of the injection molded part, and calculate the positional relationship between the spatial coordinates of each mesh node of the continuous solid geometry model and the boundary description of the internal solid region of the injection molded part.

[0104] In this embodiment, the boundary description Bd of the internal solid region of the injection molded part is obtained and stored in the form of a closed triangular mesh surface. For each mesh node Nv in the continuous solid geometry model Md, the ray casting method is used to determine whether the mesh node is located within the internal solid region of the injection molded part: a ray is drawn from Nv in any direction, and the number of intersections between the ray and the boundary triangular mesh surface of the internal solid region of the injection molded part is calculated. If the number of intersections is odd, Nv is located in the internal region; if it is even, it is located in the external region. The position determination result of each mesh node is recorded.

[0105] Step S350: When the spatial coordinates of any mesh node in the continuous solid geometry model are outside the boundary description of the solid region inside the injection molded part, the mesh node is backed back by a preset back distance in the opposite direction of its thickness direction line segment until the spatial coordinates of the mesh node return to the boundary description of the solid region inside the injection molded part, thus obtaining the interference-corrected continuous solid geometry model.

[0106] In this embodiment, for the mesh node Ne determined in step S340 to be located outside the boundary description of the solid region inside the injection molded part, the thickness direction line segment Lt of the mesh node is obtained. This thickness direction line segment connects the original mesh node No and the thickness extension endpoint Pp. The point Pn closest to Ne in the boundary description Bd of the solid region inside the injection molded part is obtained, and the spatial distance Dp from Ne to Pn is calculated. Along the opposite direction of the thickness direction line segment Lt, i.e., from Pp to No, the node is moved by a distance Dp+Ds, where Ds is a small safety distance, to obtain the corrected node position Nc. The coordinates of Ne are updated to Nc. This process is repeated until the spatial coordinates of all mesh nodes are within the boundary description of the solid region inside the injection molded part. After completing the correction of all interference nodes, the interference-corrected continuous solid geometry model Mf is obtained.

[0107] Step S360: Output the interference-corrected continuous solid geometry model as a mold parting surface structure model with solid thickness.

[0108] In this embodiment, the interference-corrected continuous solid geometry model Mf obtained in step S350 is used as the final output. This continuous solid geometry model is a mold parting surface structure model with solid thickness, which can be directly used in the subsequent mold processing and manufacturing process.

[0109] For example, after step S150, the following step may be included: Step S410: Obtain the geometric model of the target mold parting surface and the appearance data set of the injection molded part, and call the pre-generated parting surface rationality discrimination adversarial network to perform rationality discrimination processing on the geometric model of the target mold parting surface. The parting surface rationality discrimination adversarial network includes a generator sub-network and a discriminator sub-network. The generator sub-network is used to generate candidate parting surface geometric models, and the discriminator sub-network is used to output the rationality score of the parting surface geometric model.

[0110] In this embodiment, the target mold parting surface geometric model Mo is obtained in step S150, and the injection molded part appearance data set Ds is obtained in step S116. A pre-generated adversarial network for parting surface rationality judgment is constructed, comprising a generator sub-network G and a discriminator sub-network D. The generator sub-network G employs a graph convolutional network architecture, with inputs of random noise vectors and appearance constraints, and outputs mesh data of candidate parting surface geometric models. The discriminator sub-network D employs a three-dimensional convolutional neural network architecture, with inputs of mesh data of the parting surface geometric model, and outputs a scalar value Sc representing the rationality score of the input model.

[0111] Step S420: Input the geometric model of the target mold parting surface into the discriminator sub-network, and extract the local geometric features of each parting surface spatial sheet unit in the geometric model of the target mold parting surface through the multi-layer convolutional layer of the discriminator sub-network to obtain a first set of local geometric features. The first set of local geometric features includes the curvature distribution features of the mesh nodes and the variation features of the mesh unit normal vector of each parting surface spatial sheet unit.

[0112] In this embodiment, the geometric model Mo of the target mold parting surface is converted into a three-dimensional voxel representation to obtain a size L. W The voxel grid Vg of H is input into the first convolutional layer C1 of the discriminator subnetwork D. The first convolutional layer C1 contains 64 voxel grids of size 3. 3 A convolutional kernel of size 3, stride of 1, padding of 1, and output size L. W H The first feature map F1 is 64. The first feature map F1 is then passed through the second convolutional layer C2, containing 128 3's. 3 3 convolutional kernels, the third convolutional layer C3 contains 256 3-core kernels. 3 The third convolutional kernel and the fourth convolutional layer C4 contain 512 3-core kernels. 3 Three convolutional kernels are used, with each convolutional layer followed by a batch normalization layer and an activation function. Feature vectors corresponding to the regions of each fractal surface sheet unit are extracted from the feature map F4 output from the fourth convolutional layer C4, resulting in the first local geometric feature set Lf. Each local geometric feature includes a sub-feature Kf of the grid node curvature distribution and a sub-feature Nf of the grid unit normal vector change.

[0113] Step S430: The first local geometric feature set is subjected to global feature aggregation processing through the global pooling layer of the discriminator sub-network to generate a global geometric feature vector of the target mold parting surface geometric model. The global geometric feature vector is used to characterize the overall surface continuity of the target mold parting surface geometric model and the degree of avoidance of the transition area with the appearance continuity.

[0114] In this embodiment, the first local geometric feature set Lf obtained in step S420 is input into the global average pooling layer Pv of the discriminator sub-network D. This pooling layer averages the feature values ​​of all spatial locations on each feature channel to obtain a global feature vector Gv of length 512. This global feature vector Gv is then input into a fully connected layer Fc1, which contains 256 neurons and outputs a 256-dimensional first intermediate feature vector M1. The first intermediate feature vector M1 then passes through a second fully connected layer Fc2, which contains 128 neurons and outputs a 128-dimensional second intermediate feature vector M2. In the second intermediate feature vector M2, the first 64 dimensions are used to characterize the degree of surface continuity Cs, and the last 64 dimensions are used to characterize the degree of avoidance integrity Cr of the transition region with the appearance continuity. This 128-dimensional vector is the global geometric feature vector Gf of the target mold parting surface geometric model.

[0115] Step S440: Input the global geometric feature vector into the fully connected classification layer of the discriminator sub-network, calculate the rationality score of the target mold parting surface geometric model belonging to the rational parting surface category, and obtain the initial rationality score value.

[0116] In this embodiment, the 128-dimensional global geometric feature vector Gf generated in step S430 is input into the fully connected classification layer Fc3 of the discriminator sub-network D. This classification layer contains one output neuron and uses the Sigmoid activation function. The calculation process is as follows: the global geometric feature vector Gf is multiplied by the weight matrix Wc of the classification layer, a bias term Bc is added, and then mapped to a value between 0 and 1 using the Sigmoid function. The output value is the rationality score S0 of the target mold parting surface geometric model belonging to the reasonable parting surface category. The closer S0 is to 1, the more reasonable the model is; the closer it is to 0, the less reasonable the model is.

[0117] Step S450: Compare the initial rationality score with a preset rationality score threshold. When the initial rationality score is less than the rationality score threshold, mark the target mold parting surface geometric model as a parting surface geometric model to be optimized.

[0118] In this embodiment, a preset rationality score threshold St is set, typically to 0.8. The initial rationality score S0 calculated in step S440 is compared with St. If S0 is greater than or equal to St, the target mold parting surface geometry model is considered reasonable and requires no optimization. If S0 is less than St, the target mold parting surface geometry model is marked as the parting surface geometry model Mopt to be optimized, and proceeds to the subsequent optimization process.

[0119] Step S460: Perform local perturbation optimization processing on the parting surface geometric model to be optimized for the parting surface spatial sheet element, obtain the spatial coordinates of the mesh nodes of each parting surface spatial sheet element in the parting surface geometric model to be optimized, and superimpose a random perturbation vector on each mesh node spatial coordinate. The direction of the random perturbation vector is the direction of the surface normal vector of the mesh node, and the length of the random perturbation vector is dynamically determined according to the area of ​​the parting surface spatial sheet element to which the mesh node belongs, thereby generating a set of perturbed mesh node spatial coordinates.

[0120] In this embodiment, for each spatial sheet element in the parting surface geometric model Mopt to be optimized, the area Ap of that sheet element is obtained. For each mesh node Nu in that sheet element, the surface normal vector Nv at that node is obtained. A random number Rr is generated, which follows a normal distribution with a mean of 0 and a variance of σ, where σ is inversely proportional to the square root of Ap, meaning that the larger the area of ​​the sheet element, the smaller the perturbation amplitude allowed. The perturbation length Ld = Rr is calculated. Lb, where Lb is the basic perturbation length parameter. Calculate the perturbation vector Vp = Nv. Ld. Superimpose Vp onto the original spatial coordinates of Nu to obtain the perturbed node coordinates Nu_new. After performing the above operation on all mesh nodes, the set of perturbed mesh node spatial coordinates Ns is obtained.

[0121] Step S470: Reconstruct the mesh connection relationship of the parting surface spatial sheet unit according to the disturbed mesh node spatial coordinate set, and generate the optimized parting surface geometric model.

[0122] In this embodiment, the original mesh topology of the parting surface spatial sheet element remains unchanged, and the coordinates of each mesh node are replaced with the perturbed coordinates generated in step S460. For each mesh element, the corresponding node coordinates are obtained from the perturbed mesh node spatial coordinate set Ns according to its node index list, and the geometry of the mesh element is reconstructed. All mesh elements are traversed to complete the mesh reconstruction of the entire parting surface spatial sheet element. All reconstructed parting surface spatial sheet elements are combined according to their original connection relationships to generate the optimized parting surface geometric model Mnew.

[0123] Step S480: Input the optimized parting surface geometric model into the discriminator sub-network again to calculate the rationality score and obtain the optimized rationality score value. When the optimized rationality score value is still less than the rationality score threshold, repeat the local perturbation optimization process and rationality score calculation operation until the optimized rationality score value reaches or exceeds the rationality score threshold to obtain the final optimized parting surface geometric model.

[0124] In this embodiment, the optimized parting surface geometry model Mnew generated in step S470 is input again into the discriminator sub-network D, and the operations of steps S420 to S440 are repeated to calculate the optimized rationality score Sn. Snew is compared with St. If Snew is still less than St, the current optimized parting surface geometry model Mnew is used as the new parting surface geometry model to be optimized, and the local perturbation optimization processing of steps S460 and S470 is repeated to generate the optimized parting surface geometry model for the next iteration, and the rationality score is recalculated. This iterative process is repeated until the rationality score obtained in a certain iteration reaches or exceeds St. At this point, the parting surface geometry model is used as the final optimized parting surface geometry model Mfin.

[0125] Step S490: Compare the final optimized parting surface geometric model with the historical generation results of the generator sub-network, extract the positions and changes of the mesh nodes in the final optimized parting surface geometric model that have changed relative to the target mold parting surface geometric model, and generate a parting surface optimization adjustment record set. The parting surface optimization adjustment record set includes the original spatial coordinates, optimized spatial coordinates and adjustment direction vector of each adjusted mesh node.

[0126] In this embodiment, the set of mesh node coordinates Nf for the final optimized parting surface geometric model Mfin is obtained, and the set of mesh node coordinates No for the original target mold parting surface geometric model Mo is obtained. For each mesh node, the spatial coordinate difference vector Va = Nf - No for the corresponding nodes in the two sets is calculated. When the magnitude of Va is greater than a preset adjustment threshold Ta, the node is recorded as an adjustment node. For each adjustment node, its original spatial coordinates, optimized spatial coordinates, and adjustment direction vector are recorded, i.e., the result after Va normalization. The recorded information of all adjustment nodes constitutes the parting surface optimization adjustment record set Ra.

[0127] Step S4100: Update the network weight parameters of the generator subnetwork according to the set of records for optimizing the fractal surface, so that the generator subnetwork will prioritize generating fractal surface geometric models with high rationality scores when generating candidate fractal surface geometric models in the future, and obtain the updated fractal surface rationality discrimination adversarial network.

[0128] In this embodiment, the set of fractal optimization adjustment records Ra is used as positive samples to update the network weight parameters of the generator sub-network G. The difference loss Lg between the candidate fractal geometry model generated by the generator sub-network G under the current parameters and the positive samples is calculated. This loss function includes adversarial loss and reconstruction loss. The Adam optimizer is used, with a learning rate of 0.0002, an exponential decay rate of 0.5 for the first moment estimation, and an exponential decay rate of 0.999 for the second moment estimation. Gradient descent is applied to update the weight parameters of the generator sub-network G. Simultaneously, the parameters of the discriminator sub-network D are updated to better distinguish between the optimized reasonable fractal geometry model and the candidate model generated by the generator. Multiple training iterations are performed until the average reasonableness score of the candidate fractal geometry model generated by the generator sub-network G reaches a preset target value. After training, the updated fractal reasonableness discriminative adversarial network Net_new is obtained.

[0129] For example, after step S150, the following step may be included: Step S510: Obtain the spatial boundary description of the appearance continuity transition area in the geometric model of the target mold parting surface and the appearance data set of the injection molded part, and construct an interaction relationship diagram between the parting surface and the transition area. The interaction relationship diagram between the parting surface and the transition area uses each parting surface spatial sheet unit and each appearance continuity transition area in the geometric model of the target mold parting surface as graph nodes, and the spatial adjacency relationship between the parting surface spatial sheet unit and the appearance continuity transition area as graph edges.

[0130] In this embodiment, the geometric model Mo of the target mold parting surface is obtained from step S150, and the spatial boundary description of the appearance continuity transition region in the injection molded part appearance data set is obtained from step S116. An interaction graph Grap between the parting surface and the transition region is constructed, with each parting surface spatial sheet unit Pi and each appearance continuity transition region Tj in the geometric model of the target mold parting surface as graph nodes, and the spatial adjacency relationship between the parting surface spatial sheet unit and the appearance continuity transition region as graph edges. When the spatial distance between the boundary of the parting surface spatial sheet unit Pi and the boundary of the appearance continuity transition region Tj is less than a preset adjacency threshold, a graph edge Eij is established between Pi and Tj.

[0131] Step S520: Perform element geometric feature encoding processing on each parting surface spatial sheet element, extract the area value, average curvature value of the mesh node, and dispersion value of the mesh element normal vector of the parting surface spatial sheet element, and concatenate the area value, average curvature value, and dispersion value into a three-dimensional element feature vector as the initial element feature encoding of the parting surface spatial sheet element.

[0132] In this embodiment, for each parting surface spatial sheet element Pi, the area value Ai of the sheet element is calculated, the average curvature Kavg of all mesh nodes of the sheet element is calculated, and the dispersion value Vd of the normal vectors of all mesh elements of the sheet element is calculated. This dispersion value is the average of the variance of the angle between the normal vector of each mesh element and the average normal vector. The three values ​​Ai, Kavg, and Vd are concatenated into a three-dimensional element feature vector Fp=[Ai, Kavg, Vd], which serves as the initial element feature encoding for the parting surface spatial sheet element.

[0133] Step S530: Perform transition region geometric feature encoding processing on each appearance continuity transition region, extract the average width distribution value, length value and cumulative value of centerline curvature change of the appearance continuity transition region, and concatenate the average width distribution value, length value and cumulative value of curvature change into a three-dimensional transition region feature vector as the initial transition region feature encoding of the appearance continuity transition region.

[0134] In this embodiment, for each appearance continuity transition region Tj, the average width distribution Wavg is calculated from the width distribution set generated in step S114, the length value Lj of the transition region is calculated, and the cumulative curvature change value Cc of the centerline of the transition region is calculated. This cumulative curvature change value is the integral of the curvature at each point along the centerline. The three values ​​Wavg, Lj, and Cc are concatenated into a three-dimensional transition region feature vector Ft=[Wavg, Lj, Cc], which serves as the initial transition region feature encoding for the appearance continuity transition region.

[0135] Step S540: The initial unit feature encoding and the initial transition region feature encoding are respectively used as the node feature vectors of the corresponding graph nodes. The adjacency matrix of the interaction graph between the fractal surface and the transition region is used as the graph edge connection relationship. The pre-trained fractal surface spatial layout graph convolutional network is called to iteratively update the node feature vectors. The fractal surface spatial layout graph convolutional network contains multiple graph convolutional layers. Each graph convolutional layer aggregates and updates the feature vector of the current graph node according to the feature vectors of the adjacent graph nodes and the weights of the graph edges, generating the enhanced unit feature vector of each fractal surface spatial sheet unit and the enhanced transition region feature vector of each appearance continuity transition region.

[0136] In this embodiment, the initial unit feature encoding Fp generated in step S520 and the initial transition region feature encoding Ft generated in step S530 are used as the node feature vectors of the corresponding graph nodes, and the adjacency matrix Adj of the interaction graph between the fractal surface and the transition region Grapp is used as the graph edge connection relationship. The pre-trained fractal surface spatial layout graph convolutional network Gcn is called to iteratively update the node feature vectors. This graph convolutional network contains three graph convolutional layers. For each node, the first graph convolutional layer aggregates the feature vectors of its neighboring nodes, sums them with its own feature vector in a weighted manner, and then passes them through an activation function to obtain the first layer output features. The second graph convolutional layer takes the first layer output as input and repeats the aggregation operation to obtain the second layer output features. The third graph convolutional layer obtains the final output. After three layers of graph convolution, each fractal surface spatial sheet unit Pi outputs an enhanced unit feature vector Fpe, and each appearance continuity transition region Tj outputs an enhanced transition region feature vector Fte.

[0137] Step S550: Calculate the spatial matching score between each parting surface spatial sheet unit and the adjacent appearance continuity transition region based on the enhanced unit feature vector and the enhanced transition region feature vector. The spatial matching score is the cosine similarity value between the enhanced unit feature vector and the enhanced transition region feature vector.

[0138] In this embodiment, for each parting surface spatial sheet unit Pi and its adjacent appearance continuity transition region Tj, the enhanced unit feature vector Fpe of Pi and the enhanced transition region feature vector Fte of Tj are obtained, and the cosine similarity value Sij between them is calculated as Sij = (Fpe·Fte) / (|Fpe| |Fte|), the cosine similarity value is the spatial matching score between Pi and Tj.

[0139] Step S560: When the spatial matching score between any parting surface spatial sheet unit and its adjacent appearance continuity transition area is lower than the preset matching threshold, the combination of the parting surface spatial sheet unit and the appearance continuity transition area is marked as a combination pair to be adjusted.

[0140] In this embodiment, a preset matching threshold Mt is typically set to 0.7. All parting surface spatial sheet units Pi and their adjacent appearance continuity transition regions Tj are traversed, and the spatial matching score Sij calculated in step S550 is compared with Mt. When Sij is lower than Mt, (Pi, Tj) is marked as a pair to be adjusted and added to the adjustment list Ladj.

[0141] Step S570: Perform boundary fine-tuning processing on the parting surface spatial sheet unit for each pair of combinations to be adjusted, extract the boundary curve of the parting surface spatial sheet unit in the pair of combinations to be adjusted that is close to the appearance continuity transition region, and shift the boundary curve away from the appearance continuity transition region by a translation distance. The magnitude of the translation distance is linearly determined based on the difference between the spatial matching degree score and the matching degree threshold, and the fine-tuned boundary curve is obtained.

[0142] In this embodiment, for each pair of combinations (Pi, Tj) to be adjusted in the list Ladj, the boundary curve Bc of the parting surface spatial sheet unit Pi, which is close to the appearance continuity transition region Tj, is extracted. The difference Dm = Mt - Sij between the spatial matching score Sij and the matching threshold Mt is calculated. The translation distance Lm = k is determined. Dm, where k is a preset scaling factor. The boundary curve Bc is shifted a distance Lm away from Tj to obtain the fine-tuned boundary curve Bnew.

[0143] Step S580: Regenerate the mesh cells of the parting surface spatial sheet unit according to the fine-tuned boundary curve, keeping the part of the parting surface spatial sheet unit away from the appearance continuity transition area unchanged, and only replacing the boundary area mesh near the appearance continuity transition area to obtain the fine-tuned parting surface spatial sheet unit.

[0144] In this embodiment, the boundary region mesh of the parting surface spatial sheet element Pi near Tj is regenerated based on the fine-tuned boundary curve Bnew. The mesh nodes and mesh elements of the portion of Pi away from Tj remain unchanged. Using Bnew as the new boundary, the boundary region is re-triangulated to generate new mesh element connection relationships, resulting in the fine-tuned parting surface spatial sheet element Pi_new.

[0145] Step S590: Merge all finely tuned parting surface spatial sheet elements with the original parting surface spatial sheet elements in the non-adjustable combination pair to generate the final mold parting surface geometric model optimized based on graph convolutional network.

[0146] In this embodiment, all fine-tuned parting surface space sheet elements Pi_new and all original parting surface space sheet elements that do not appear in the adjustment list Ladj are collected. The sheet elements are merged according to their original positions, and node merging is performed at the common boundary of adjacent sheet elements to eliminate duplicate mesh nodes and mesh edges, thus obtaining the final mold parting surface geometric model Mfinal optimized based on graph convolutional network.

[0147] For example, before step S120, the following step may be included: Step S610: Constructing an injection molded part appearance topology graph structure, wherein each appearance surface unit is a graph node unit, and the appearance continuity transition region between every two adjacent appearance surface units is a graph connection edge. Each graph node unit carries the surface geometric attribute field feature vector of the appearance surface unit, and each graph connection edge carries the transition region width distribution set feature vector and the transition region curvature change set feature vector of the appearance continuity transition region.

[0148] In this embodiment, all appearance surface units and appearance continuity transition regions are extracted from the injection molded part appearance data set generated in step S116. A topological graph structure Gtop is constructed, with each appearance surface unit Ui as a graph node and the appearance continuity transition region Tij between every two adjacent appearance surface units as a graph edge connecting node Ui and node Uj. For each graph node Ui, a feature vector Fui is extracted from the surface geometric attribute field generated in step S113. This feature vector consists of the statistical values ​​of the normal vector components, principal curvature maxima, and principal curvature minima of all sampling points on the appearance surface unit. For each graph edge Tij, a width distribution feature vector Wij is extracted from the transition region width distribution set generated in step S114, and a curvature change feature vector Cij is extracted from the transition region curvature change set generated in step S115. Wij and Cij are concatenated to form the feature vector Eij of this graph edge.

[0149] Step S620: Input the topological graph structure of the injection molded part appearance into a pre-trained graph neural network for parting line candidate regions. The graph neural network for parting line candidate regions includes a graph convolutional encoder module and a graph attention decoder module. The graph convolutional encoder module performs multi-layer graph convolution operations on each graph node unit and its adjacent graph node units, aggregating the multi-order neighborhood information of each graph node unit to generate the node encoding feature vector of each graph node unit. The graph attention decoder module performs attention weighted fusion processing on the node encoding feature vectors of the graph node units at both ends of each graph connection edge to generate the edge weight prediction value of each graph connection edge. The edge weight prediction value represents the suitability of the appearance continuity transition region corresponding to the graph connection edge as the location through which the parting line passes.

[0150] In this embodiment, the injection molded part appearance topology graph structure Gtop constructed in step S610 is input into the pre-trained parting line candidate region graph neural network Net_pline, which includes a graph convolutional encoder module Enc and a graph attention decoder module Dec. The graph convolutional encoder module Enc contains three graph convolutional layers. The first graph convolutional layer aggregates the feature vectors Fuji of its neighboring nodes Uj for each node Ui, and then weights and sums them with its own feature vector Fui before passing them through the ReLU activation function to obtain the first layer encoded feature H1i. The second graph convolutional layer takes H1i as input and aggregates the second-order neighbor information to obtain the second layer encoded feature H2i. The third graph convolutional layer obtains the final node encoded feature vector Zui. The graph attention decoder module Dec takes each graph edge Tij and obtains the node encoding feature vectors Zui and Zuj of the two endpoints Ui and Uj of that edge. It then calculates the attention coefficient αij = softmax(LeakyReLU(a^T[WZui||WZuj])), where a is the learnable attention vector, W is the learnable weight matrix, and || denotes the concatenation operation. The attention coefficient αij is then weighted and fused with the transformation features of Zui and Zuj, and outputs the edge weight prediction value Pij through a fully connected layer. This prediction value indicates the suitability of the transition region Tij as a fractal line; the closer Pij is to 1, the more suitable it is.

[0151] Step S630: Sort all graph connection edges in the topology graph structure of the injection molded part according to the predicted edge weight values ​​of each graph connection edge from high to low, select the graph connection edges with the highest sorting position as the parting line candidate connection edge set, and extract the width center trajectory of the appearance continuity transition area corresponding to each graph connection edge in the parting line candidate connection edge set as the parting line candidate trajectory segment.

[0152] In this embodiment, the predicted edge weights Pij of all graph edges Tij obtained in step S620 are sorted from high to low, and the top K graph edges are selected as the set of candidate connecting edges for the fractal line, E_cand. For each graph edge Tij in E_cand, the width center trajectory Wcenter_ij corresponding to the transition region is extracted from the width center trajectory generated in step S122, and this width center trajectory is used as the candidate trajectory segment Seg_cand for the fractal line. All candidate trajectory segments constitute the set of candidate trajectory segments for the fractal line, Seg_set.

[0153] Step S640: Perform fractal candidate trajectory segment connection optimization processing on the set of fractal candidate trajectory segments to construct a trajectory segment connection graph. Each node in the trajectory segment connection graph corresponds to a fractal candidate trajectory segment, and the edge between any two nodes represents the spatial distance between the endpoints of the two fractal candidate trajectory segments. By solving the minimum spanning tree algorithm of the trajectory segment connection graph, the fractal candidate trajectory segments are connected into a continuous fractal backbone trajectory to obtain the graph-optimized fractal backbone trajectory.

[0154] In this embodiment, a trajectory segment connection graph Gseg is constructed, where each node corresponds to a segment Seg_i in the candidate trajectory segment set Seg_set. For any two segments Seg_i and Seg_j, the spatial distances between all endpoints of Seg_i and all endpoints of Seg_j are calculated, and the minimum value is taken as Dij. An edge is established between node i and node j, with the weight of the edge being Dij. The Kruskal algorithm is used to solve for the minimum spanning tree of the graph: all edges are sorted in ascending order of weight, and edges are taken out one by one. If the two nodes connected by the edge are not in the same connected component, the edge is added to the minimum spanning tree until all nodes are in the same connected component. The trajectory segments corresponding to the edges connected in the minimum spanning tree are arranged in connection order, and adjacent segments are connected at the nearest endpoint to form a continuous fractal main trajectory T_main.

[0155] Step S650: The graph-optimized parting line backbone trajectory is fused with the curvature change extreme value trajectory of each appearance surface unit. At the intersection of the graph-optimized parting line backbone trajectory and the curvature change extreme value trajectory, the end branch of the curvature change extreme value trajectory is connected to the parting line backbone trajectory to generate the fused parting line trajectory network.

[0156] In this embodiment, the fractal backbone trajectory T_main generated in step S640 is obtained, and the curvature variation extreme value trajectory T_curv of each appearance surface unit generated in step S121 is obtained. The intersection point of T_main and each T_curv is calculated, and the intersection point position is P_int. For each intersection point P_int, the endpoint P_end on T_curv that is closest to P_int is found, and P_end is connected to P_int, so that the curvature variation extreme value trajectory is connected to the backbone trajectory as a branch. All trajectories connected to the branch are merged to generate the fused fractal trajectory network T_net.

[0157] Step S660: Input the fused fractal trajectory network into the fractal spatial position regression network. The fractal spatial position regression network includes a spatial coordinate encoding layer and a sequence generation layer. The spatial coordinate encoding layer encodes the spatial coordinates of each trajectory point in the fused fractal trajectory network into a high-dimensional spatial coordinate embedding vector. The sequence generation layer performs autoregressive sequence generation processing on the high-dimensional spatial coordinate embedding vector and outputs the spatial coordinate adjustment offset of each trajectory point. The original spatial coordinates of each trajectory point are added to the corresponding spatial coordinate adjustment offset to obtain the optimized spatial coordinates of the trajectory point.

[0158] In this embodiment, the fused fractal trajectory network T_net is input into the fractal spatial position regression network Net_reg, which includes a spatial coordinate encoding layer Enc_pos and a sequence generation layer LSTM. The spatial coordinate encoding layer Enc_pos encodes the spatial coordinates (Xk, Yk, Zk) of each trajectory point Pk in T_net. First, the coordinate values ​​are normalized to the interval [-1, 1], and then mapped to a 256-dimensional high-dimensional spatial coordinate embedding vector Ek through a multilayer perceptron. The sequence generation layer LSTM uses a two-layer LSTM structure with a hidden layer dimension of 512. The embedding vector sequence [E1, E2, ..., En] is sequentially input into the LSTM, and a hidden state Hk is output at each time step. Hk is then passed through a fully connected layer to output a three-dimensional adjustment offset (ΔXk, ΔYk, ΔZk). Add the original spatial coordinates (Xk, Yk, Zk) to the adjusted offset (ΔXk, ΔYk, ΔZk) to obtain the optimized trajectory point spatial coordinates (Xk', Yk', Zk').

[0159] Step S670: Arrange all optimized trajectory point spatial coordinates into a sequence according to the connection order of trajectory points in the fused fractal trajectory network to generate the optimized initial fractal spatial position sequence.

[0160] In this embodiment, the optimized spatial coordinates (Xk', Yk', Zk') of all trajectory points obtained in step S660 are arranged according to the connection order of the trajectory points in the fused fractal trajectory network T_net, forming a spatial coordinate sequence Seq_opt. This sequence is used as the optimized initial fractal spatial position sequence, replacing the initial fractal spatial position sequence generated in step S125, for subsequent fractal surface generation processing.

[0161] For example, after step S150, the following step may be included: Step S710: The geometric model of the target mold parting surface is divided into multiple parting surface local blocks. Each parting surface local block corresponds to a parting surface spatial sheet unit generated by the expansion of the parting surface starting boundary curve on an appearance surface unit. The block boundary tension field calculation process is performed on each parting surface local block. The spatial coordinate sequence of all boundary grid edges of the parting surface local block is extracted. The direction of the connection line between the grid nodes at both ends of each boundary grid edge is calculated as the boundary tension direction. The spatial length of each boundary grid edge is calculated as the boundary tension amplitude. The boundary tension direction and boundary tension amplitude of all boundary grid edges are combined to form the boundary tension field of the parting surface local block.

[0162] In this embodiment, the geometric model Mtar of the target mold parting surface is obtained from step S150. This model is then divided into multiple local blocks Bk of the parting surface according to the appearance surface units corresponding to the initial boundary curve of the parting surface. Each block corresponds to a spatial sheet unit of the parting surface. For each block Bk, all its boundary mesh edges Eb are extracted. For each boundary mesh edge Eb_i, the mesh nodes Na and Nb at both ends of the edge are connected. The direction vector Dbi from Na to Nb is calculated as the boundary tension direction, and the spatial length Lbi between Na and Nb is calculated as the boundary tension amplitude. The boundary tension directions Dbi and boundary tension amplitudes Lbi of all boundary mesh edges are combined to form the boundary tension field Fb_k of the block.

[0163] Step S720: Call the pre-trained injection flow simulation proxy neural network to perform flow front propulsion resistance prediction processing on each parting surface local block. The injection flow simulation proxy neural network consists of multi-layer graph convolutional layers and global pooling layers. Input the spatial coordinate sequence of grid nodes and the grid cell connection relationship sequence of each parting surface local block into the injection flow simulation proxy neural network. Extract the local geometric features of each grid node through multi-layer graph convolutional layers. Aggregate the local geometric features of all grid nodes into a fixed-length block flow resistance feature vector through global pooling layers. Input the block flow resistance feature vector into a fully connected regression layer to output the predicted value of the flow front propulsion resistance corresponding to the parting surface local block.

[0164] In this embodiment, a pre-trained injection molding flow simulation proxy neural network, Net_flow, is invoked. For each local block Bk of the parting surface, its grid node spatial coordinate sequence Nseq and grid cell connection relationship sequence Cseq are obtained. Nseq and Cseq are input into Net_flow, which contains three graph convolutional layers Gconv and a global pooling layer Gpool. The first graph convolutional layer aggregates the coordinate features of its neighboring nodes for each grid node, outputting the first layer of node features; the second graph convolutional layer outputs the second layer of node features; and the third graph convolutional layer outputs the local geometric feature vector Fnode of the node. The global pooling layer Gpool averages the Fnode of all nodes to obtain the block flow resistance feature vector Fr. Fr is input into a fully connected regression layer Fc_reg, which outputs the predicted value Rflow of the flow front propulsion resistance for the block.

[0165] Step S730: Construct the deformation driving force field of the local block of the parting surface based on the boundary tension field and the predicted value of the flow front propulsion resistance of each local block. Multiply the boundary tension direction on each boundary grid edge by the reciprocal of the predicted value of the flow front propulsion resistance to obtain the deformation driving force vector of each boundary grid edge. Aggregate the deformation driving force vectors of all boundary grid edges according to the grid nodes. At each grid node, sum the deformation driving force vectors of all boundary grid edges sharing that grid node to obtain the deformation driving force vector of each grid node.

[0166] In this embodiment, for each local block Bk of the parting surface, the boundary tension direction Dbi and boundary tension amplitude Lbi of each boundary grid edge Eb_i in its boundary tension field Fb_k are obtained, and the predicted value Rflow of the flow front propulsion drag output in step S720 is obtained. The deformation driving force vector Vd_i=Dbi of each boundary grid edge Eb_i is calculated. (Lbi / Rflow). For each mesh node Nq within the block, find all boundary mesh edges that share this node, and sum the deformation driving force vectors Vd_i corresponding to these edges to obtain the deformation driving force vector Vq of this mesh node.

[0167] Step S740: Input the deformation driving force vector of all grid nodes of each parting surface local block into a spatial deformation prediction deep neural network. The spatial deformation prediction deep neural network consists of multiple residual convolutional layers. Each residual convolutional layer performs convolution operation on the input spatial coordinate lattice and deformation driving force vector field to output an intermediate deformation displacement field. The last residual convolutional layer outputs the final deformation displacement field. The final deformation displacement field is superimposed on the original grid node spatial coordinates of the parting surface local block to obtain the deformed parting surface local block.

[0168] In this embodiment, for each local block Bk of the fractal surface, the spatial coordinate lattice Pmat of all its mesh nodes and the deformation driving force vector field Vmat are input into the spatial deformation prediction deep neural network Net_def. Net_def consists of four residual convolutional layers, each containing two convolutional kernels with a size of 3. A convolutional layer of 3 is used, with each convolutional layer followed by a batch normalization layer and a ReLU activation function. The input is added to the original input after passing through two convolutional layers to obtain the output of the residual block. The first residual convolutional layer outputs the first intermediate displacement field D1, the second residual convolutional layer outputs the second intermediate displacement field D2, the third residual convolutional layer outputs the third intermediate displacement field D3, and the fourth residual convolutional layer outputs the final deformed displacement field Df. Df is superimposed onto the original mesh node spatial coordinates Pmat to obtain the deformed mesh node coordinates Pdef, thus obtaining the deformed parting surface local block Bk_def.

[0169] Step S750: All the deformed parting surface local blocks are spliced ​​together according to the connection order of the original parting surface starting boundary curve. Boundary alignment and stitching are performed at the common boundary of adjacent deformed parting surface local blocks. The spatial distance between corresponding mesh nodes on the common boundary is calculated. When the spatial distance is greater than zero, a new stitching mesh node is inserted at the midpoint of the two mesh nodes. The original two mesh nodes are deleted and the mesh elements of the adjacent blocks are reconnected to the newly inserted stitching mesh node to generate the stitched deformed mold parting surface geometric model.

[0170] In this embodiment, all deformed parting surface local blocks Bk_def are arranged according to the connection order of the original parting surface starting boundary curve. For two adjacent blocks Bk_def and B(k+1)_def, the corresponding mesh node pair (Na, Nb) on their common boundary is obtained, and the spatial distance Dab between Na and Nb is calculated. When Dab is greater than zero, the midpoint position Pmid = ((Xa+Xb) / 2, (Ya+Yb) / 2, (Za+Zb) / 2) is calculated, and a new stitching mesh node Nmid is inserted at Pmid. The original two mesh nodes Na and Nb are deleted, and the node index of the mesh element containing Na or Nb is updated to Nmid. All mesh node pairs on the common boundary are traversed to complete the stitching process, generating the stitched deformed mold parting surface geometric model Mdef.

[0171] Step S760: Input the stitched deformed mold parting surface geometry model into the injection filling fidelity evaluation generative adversarial network. The injection filling fidelity evaluation generative adversarial network includes a filling mode discriminator network and a fidelity generator network. The filling mode discriminator network takes the stitched deformed mold parting surface geometry model as input and outputs a filling rationality score. When the filling rationality score is lower than a preset reasonable filling threshold, the filling rationality score is backpropagated to the fidelity generator network as a loss signal. The fidelity generator network generates a corrected displacement field and superimposes it onto the stitched deformed mold parting surface geometry model. This process is repeated until the filling rationality score reaches or exceeds the reasonable filling threshold. The parting surface geometry model at this time is used as the target mold parting surface geometry model for injection flow optimization.

[0172] In this embodiment, the geometric model Mdef of the deformed mold parting surface after stitching is input into the injection filling fidelity evaluation generative adversarial network Net_fill, which includes a filling pattern discriminator network D_fill and a fidelity generator network G_fill. The filling pattern discriminator network D_fill takes Mdef as input, extracts features through multiple 3D convolutional layers, and finally outputs a filling rationality score Sfill through a fully connected layer, ranging from 0 to 1. The preset reasonable filling threshold St_fill is 0.85. When Sfill is lower than St_fill, the loss signal Lfill=1-Sfill is backpropagated to the fidelity generator network G_fill. G_fill takes Mdef as input, generates a corrected displacement field Drorr through multiple deconvolutional layers, and superimposes Drorr onto the grid node coordinates of Mdef to obtain the corrected model Mcorr. Mcorr is input again into D_fill to recalculate Sfill, and this process is repeated until Sfill reaches or exceeds St_fill. The parting surface geometric model Mopt_fill at this time is used as the target mold parting surface geometric model for injection flow optimization.

[0173] For example, before step S120, the following step may be included: Step S810: Obtain parting line design case data of multiple historical injection molds. Each parting line design case data includes the geometric attribute field of the appearance surface unit of the historical injection mold, the set of width distribution of the historical appearance continuous transition area, and the sequence of spatial positions of the historical parting line. The geometric attribute field of the appearance surface unit in each parting line design case data is encoded into a first feature matrix, the set of width distribution of the appearance continuous transition area is encoded into a second feature matrix, and the sequence of spatial positions of the historical parting line is encoded into a third feature matrix to construct a parting line design case database.

[0174] In this embodiment, a large amount of historical injection mold parting line design case data is collected. For each case, a six-dimensional feature vector of all sampling points is extracted from the geometric attribute field of the surface unit of the historical injection mold, and arranged into matrix M1 as the first feature matrix; all width values ​​are extracted from the set of width distribution of the historical appearance continuity transition region, and arranged into vector Vw as the second feature matrix; the coordinates of all trajectory points in the historical parting line spatial position sequence are arranged into matrix M3 as the third feature matrix. The combination of (M1, Vw, M3) of all cases is stored in the parting line design case database DBcase.

[0175] Step S820: Input all the first feature matrices, second feature matrices, and third feature matrices from the fractal design case database into a variational autoencoder for unsupervised pre-training. The variational autoencoder includes an encoder neural network and a decoder neural network. The encoder neural network maps each fractal design case data into a mean vector and a log-variance vector in a latent space. The decoder neural network reconstructs the third feature matrix from the latent vectors sampled from the latent space. After pre-training, the network weight parameters of the encoder neural network are fixed, and the mean vector output by the encoder neural network is used as the latent encoding vector for each fractal design case data.

[0176] In this embodiment, all the first feature matrices M1, second feature matrices Vw, and third feature matrices M3 from the fractal design case database DBcase are input into the variational autoencoder (VAE). This VAE comprises an encoder network Enc_v and a decoder network Dec_v. The encoder network Enc_v consists of multiple fully connected layers, taking the concatenated feature vectors of M1, Vw, and M3 as input, and outputting the mean vector μ and the log-variance vector logσ^2 in the latent space. A latent vector z is sampled from a normal distribution N(μ, σ^2). The decoder network Dec_v, also composed of multiple fully connected layers, reconstructs the third feature matrix M3' from z. The training loss function includes reconstruction error and KL divergence. After pre-training, the weight parameters of the encoder network Enc_v are fixed, and the mean vector μ obtained by inputting each case data into Enc_v is used as the latent encoding vector Zcase for that case.

[0177] Step S830: Construct a latent space graph structure for fractal design cases. The latent space graph structure for fractal design cases uses the latent encoding vector of each fractal design case data as a graph node, and the reciprocal of the Euclidean distance between any two latent encoding vectors as the weight value of the graph edge. Perform graph convolutional autoencoder training on the latent space graph structure for fractal design cases. The graph convolutional autoencoder consists of a graph convolutional encoding layer and a graph convolutional decoding layer. The graph convolutional encoding layer convolves and aggregates the latent encoding vector of each graph node with the latent encoding vectors of its neighboring graph nodes to obtain the high-order feature vector of each graph node. The graph convolutional decoding layer reconstructs the original graph structure from the high-order feature vector. After training, the high-order feature vector of each graph node is extracted as the deep design feature vector of the fractal design case data.

[0178] In this embodiment, a latent space graph structure Gcase for fractal line design cases is constructed. Each node in this graph is a latent encoding vector Zcase_i for each case, and the edge weight Wij between node i and node j is 1 / (||Zcase_i-Zcase_j||2+ε), where ε is a small constant to prevent division by zero. A graph convolutional autoencoder is trained on this graph, which includes a graph convolutional encoding layer Enc_g and a graph convolutional decoding layer Dec_g. The graph convolutional encoding layer Enc_g contains two graph convolutional layers: the first layer aggregates the features of neighboring nodes for each node, outputting the first layer features; the second layer outputs a higher-order feature vector Hcase_i. The graph convolutional decoding layer Dec_g reconstructs the graph structure from Hcase_i, i.e., predicts the edge weights between nodes. After training, the higher-order feature vector Hcase_i of each node is extracted as the deep design feature vector Fdeep_i for that case.

[0179] Step S840: Input the surface geometric attribute field of each surface unit in the current injection molded part appearance data set and the set of transition region width distribution of each appearance continuity transition region into the variational autoencoder. Output the current latent encoding vector of the current injection molded part through the encoder neural network. Calculate the Euclidean distance between the current latent encoding vector and each graph node in the latent space graph structure of the parting line design case. Select the parting line design case data corresponding to the multiple graph nodes with the smallest Euclidean distance as the reference case set.

[0180] In this embodiment, the surface geometric attribute field of each surface unit and the set of transition region width distributions of each continuous transition region are extracted from the current injection molded part appearance data set. This data is encoded into feature matrices M1_cur and Vw_cur. M1_cur and Vw_cur are input into the encoder network Enc_v trained in step S820, outputting the current latent encoding vector Zcur for the current injection molded part. The Euclidean distance Di = ||Zcur - Zcase_i||2 between Zcur and the latent encoding vector Zcase_i of each graph node in the latent space graph structure Gcase for parting line design cases is calculated. The top M graph nodes with the smallest Di are selected, and the parting line design case data corresponding to these nodes are used as the reference case set Ref_set.

[0181] Step S850: Input the parting line spatial position sequence of each reference case data in the reference case set and the current injection molded part appearance data set into a cross-case parting line transfer neural network. The cross-case parting line transfer neural network includes a spatial attention mechanism module and a curve deformation module. The spatial attention mechanism module calculates the spatial similarity between each trajectory point in the parting line spatial position sequence of the reference case data and the width center trajectory of each appearance continuity transition region in the current injection molded part appearance data set, and generates the attention weight of each trajectory point. The curve deformation module performs a spatial affine transformation on the parting line spatial position sequence of the reference case data according to the attention weight, and projects the transformed parting line spatial position sequence onto the appearance surface unit of the current injection molded part appearance data set to generate the transferred candidate parting line spatial position sequence.

[0182] In this embodiment, the spatial position sequence of the parting line of each reference case data in the reference case set Ref_set, Pseq_ref, and the current injection molded part appearance data set Ds are input into the cross-case parting line transfer neural network Net_trans, which includes a spatial attention mechanism module Att and a curve deformation module Def. The spatial attention mechanism module Att calculates the spatial similarity Sjk between each trajectory point Pj in Pseq_ref and the width center trajectory Wcenter_k of each appearance continuity transition region of the current injection molded part, Sjk=exp(-||Pj-Qjk||2 / σ), where Qjk is the nearest point of Pj on Wcenter_k. For each Pj, all Sjk are normalized to obtain the attention weight αj. The curve deformation module Def performs a spatial affine transformation on Pseq_ref according to the attention weight αj and calculates the weighted average transformation matrix T=Σαj. Tj, where Tj is the transformation from Pj to the corresponding nearest point Qjk. The transformed parting line spatial position sequence is projected onto the appearance surface unit of the current injection molded part appearance data set to generate the migrated candidate parting line spatial position sequence Pseq_cand.

[0183] Step S860: Perform local fine-tuning on the migrated candidate fractal spatial position sequence, extract the set of curvature changes in the transition region of the appearance continuity transition area at the location of each trajectory point in the migrated candidate fractal spatial position sequence, calculate the curvature change gradient at each trajectory point according to the normal vector angle change rate sequence in the transition region curvature change set, move the spatial coordinates of the trajectory point along the direction of the curvature change gradient, the moving step size is proportional to the magnitude of the curvature change gradient, obtain the fine-tuned spatial coordinates of the trajectory point, and arrange all the fine-tuned spatial coordinates of the trajectory point in the original order to generate the initial fractal spatial position sequence.

[0184] In this embodiment, for each trajectory point Pj_cand in the spatial position sequence Pseq_cand of the migrated candidate fractal lines, the appearance continuity transition region at the location of that point is determined, and the normal vector angle change rate sequence Rseq of that region is extracted from the curvature change set of the transition region generated in step S115. The curvature change gradient Grad_j at Pj_cand is calculated, specifically the forward difference value of the corresponding position in Rseq. Pj_cand is moved along the direction of the curvature change gradient Grad_j, with a movement step size Step_j = β. |Grad_j|, where β is a preset step size coefficient, yields the fine-tuned trajectory point coordinates Pj_opt. All fine-tuned trajectory point coordinates are arranged in their original order to generate the initial parting line spatial position sequence Pseq_init, which replaces the initial parting line spatial position sequence generated in step S125.

[0185] Figure 2 This application illustrates an intelligent drawing system 100 for mold parting surfaces integrated with injection molded part appearance design, comprising a processor 1001 and a memory 1003. The processor 1001 and memory 1003 are connected, for example, via a bus 1002. Optionally, the intelligent drawing system 100 for mold parting surfaces integrated with injection molded part appearance design may further include a transceiver 1004. The transceiver 1004 can be used for data interaction between this intelligent drawing system and other intelligent drawing systems for mold parting surfaces integrated with injection molded part appearance design, such as sending and / or receiving data. It should be noted that in actual scheduling, the transceiver 1004 is not limited to one, and the structure of this intelligent drawing system 100 for mold parting surfaces integrated with injection molded part appearance design does not constitute a limitation on the embodiments of this application.

[0186] The memory 1003 is used to store program code for executing the embodiments of this application, and its execution is controlled by the processor 1001. The processor 1001 is used to execute the program code stored in the memory 1003 to implement the steps shown in the foregoing method embodiments.

[0187] The above description is only an optional implementation method for some implementation scenarios of this application. It should be noted that for those skilled in the art, other similar implementation methods based on the technical concept of this application, without departing from the technical concept of this application, also fall within the protection scope of the embodiments of this application.

Claims

1. A method for intelligently drawing the parting surface of a mold combined with the appearance design of injection molded parts, characterized in that, The method includes: Obtain a set of appearance data for injection molded parts, wherein the set of appearance data includes appearance surface units and appearance continuity transition areas between appearance surface units; The parting line direction deduction process is performed on the injection molded part appearance data set, and an initial parting line spatial position sequence is generated based on the extreme value trajectory of the curvature change of the appearance surface unit and the width center trajectory of the appearance continuous transition region. Based on the initial parting line spatial position sequence, the appearance surface unit is subjected to parting surface starting boundary anchoring processing, and the parting surface starting boundary curve is determined on the appearance surface unit to obtain a set of parting surface starting boundary curves carrying the starting boundary identifier. The parting surface spatial extension generation process is performed by calling the set of parting surface initial boundary curves carrying the initial boundary identifier and the initial parting line spatial position sequence. Starting from the parting surface initial boundary curve, the process extends along the initial parting line spatial position sequence into the internal solid region of the injection molded part to generate parting surface spatial sheet units, thereby obtaining a preliminary geometric model of the parting surface composed of parting surface spatial sheet units. The parting surface preliminary geometric model is subjected to intersection avoidance trimming processing between the parting surface and the appearance continuity transition area. The part of the parting surface spatial sheet unit that intrudes into the appearance continuity transition area is cut off to obtain the target mold parting surface geometric model.

2. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 1, characterized in that, The set of appearance data for the injection molded part includes: Obtain the original three-dimensional appearance surface model of the injection molded product, perform appearance surface unit segmentation processing on the original three-dimensional appearance surface model, and divide the original three-dimensional appearance surface model into multiple non-overlapping appearance surface units according to the abrupt boundary of the surface normal vector and the zero intersection point of the surface Gaussian curvature in the original three-dimensional appearance surface model. Each appearance surface unit corresponds to a continuous appearance area of ​​the injection molded product. For two adjacent surface units, a continuous transition region identification process is performed. The common boundary line between the two adjacent surface units is extracted. A preset transition zone width is extended into the surface units on both sides using the common boundary line as a baseline. Within the transition zone width, the normal vector angle sequence and the principal curvature ratio sequence of the surface points are collected. The transition zone region where the normal vector angle sequence changes continuously and the principal curvature ratio sequence changes monotonically is marked as the continuous transition region. Thus, the corresponding continuous transition region for each adjacent surface unit is obtained. For each surface unit, surface geometric property quantization is performed. The normal vector coordinate components, principal curvature maxima and principal curvature minima at each sampling point are calculated to generate the surface geometric property field of the surface unit. The surface geometric property field includes the normal vector coordinate component values, principal curvature maxima and principal curvature minima values ​​at each sampling point. For each appearance continuity transition region, perform transition region width distribution acquisition processing. In the appearance continuity transition region, set multiple width measurement sampling lines along the direction perpendicular to the common boundary line. Calculate the surface arc length from the common boundary line to the boundary of the transition region on each width measurement sampling line as the transition region width value at the corresponding sampling position, and obtain a transition region width distribution set composed of multiple transition region width values. For each continuous transition region, a transition region curvature change acquisition process is performed. Multiple curvature sampling trajectories are set in the continuous transition region along the direction parallel to the common boundary line. On each curvature sampling trajectory, the normal vector angle change rate sequence and the principal curvature ratio change rate sequence are recorded to generate a transition region curvature change set containing the normal vector angle change rate sequence and the principal curvature ratio change rate sequence. The surface geometric attribute field of the appearance surface unit, the width distribution set of the transition region, and the curvature change set of the transition region are associated and stored to obtain a set of injection molded part appearance data containing appearance surface unit identifier, adjacent relationship identifier, and transition region identifier.

3. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 1, characterized in that, The step of performing parting line trajectory deduction processing on the injection molded part appearance data set, generating an initial parting line spatial position sequence based on the curvature change extreme value trajectory of the appearance surface unit and the width center trajectory of the appearance continuous transition region, includes: Extract the surface geometric attribute field of each surface unit in the injection molded part appearance data set. Search for the spatial coordinates of surface points where the principal curvature maxima and principal curvature minima are local maxima and local minima in the surface geometric attribute field. Connect the spatial coordinates of all the searched surface points to form a curve trajectory to generate the curvature change extreme value trajectory corresponding to each surface unit. The curvature change extreme value trajectory is composed of multiple extreme point spatial coordinates arranged in order of geodesic distance on the surface. Extract the width distribution set of each appearance continuity transition region in the injection molded part appearance data set. Calculate the spatial coordinates of the midpoint of the width value of the transition region on each width measurement sampling line in the appearance continuity transition region. Connect all the spatial coordinates of the midpoints sequentially along the extension direction of the common boundary line to generate the width center trajectory corresponding to each appearance continuity transition region. The width center trajectory is composed of multiple midpoint spatial coordinates arranged in the order of the sampling lines. The width center trajectory of the continuous transition region shared between adjacent surface units and the curvature change extreme value trajectory of each of the two adjacent surface units are processed to solve the trajectory intersection point. The spatial straight-line distance between the endpoint spatial coordinates of the width center trajectory and the endpoint spatial coordinates of each curvature change extreme value trajectory is calculated. When the spatial straight-line distance is less than a preset distance threshold, the endpoints of the width center trajectory and the endpoints of the curvature change extreme value trajectory are connected to obtain the parting line backbone trajectory network containing the trajectory intersection point. The main trajectory network of the fractal line is subjected to trajectory smoothing optimization processing. At each trajectory intersection point in the main trajectory network of the fractal line, the angle between the tangent vectors of adjacent trajectory segments is calculated. When the angle between the tangent vectors is greater than the preset angle threshold, a Bézier curve transition segment is inserted near the trajectory intersection point to replace the original broken line segment, thus obtaining the smoothed main trajectory network of the fractal line. The smoothed parting line backbone trajectory network is subjected to trajectory spatial sorting processing. The spatial projection length of each trajectory segment is calculated with the demolding direction of the injection molded part as the reference direction. The trajectory segments are sorted in descending order of spatial projection length, and the spatial coordinates of each trajectory sampling point are extracted sequentially along the sorted trajectory segments to generate an initial parting line spatial position sequence according to the priority of the demolding direction.

4. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 1, characterized in that, The process involves anchoring the initial boundary of the appearance surface unit based on the initial parting line spatial position sequence, determining the initial boundary curve of the parting surface on the appearance surface unit, and obtaining a set of initial boundary curves of the parting surface carrying the initial boundary identifier, including: The spatial coordinates of the first trajectory sampling point are extracted from the initial parting line spatial position sequence as the starting anchor point spatial coordinates. A starting boundary search circle is generated in the tangent plane of the appearance surface unit with the starting anchor point spatial coordinates as the center. The starting boundary search circle intersects with the surface of the appearance surface unit to form a starting boundary closed curve. The initial boundary curve extension direction determination process is performed at the spatial coordinates of each trajectory sampling point by sequentially traversing each trajectory sampling point along the spatial position sequence of the initial parting line. The spatial vector between the current trajectory sampling point and the next trajectory sampling point is calculated as the direction vector of the current trajectory segment, and the direction vector is projected onto the tangential plane of the appearance surface unit to which the current trajectory sampling point belongs to obtain the tangential extension direction of the initial boundary curve at the current trajectory sampling point. At each trajectory sampling point, dynamic adjustment of the width of the initial boundary curve is performed. The maximum and minimum principal curvature values ​​of the current trajectory sampling point are obtained in the surface geometric attribute field of the appearance surface unit to which the current trajectory sampling point belongs. The ratio of the maximum to the minimum principal curvature values ​​is calculated as a surface curvature index. Based on the surface curvature index, a preset width mapping relationship is queried to determine the local width value of the initial boundary curve at the current trajectory sampling point. The width mapping relationship is configured such that the larger the surface curvature index, the smaller the determined local width value. At each trajectory sampling point, a local segment of the initial boundary curve is generated based on the initial boundary search circle, the tangential extension direction, and the local width value. Taking the current trajectory sampling point as the center, the local width value is extended to both sides along the perpendicular tangential extension direction to obtain two boundary control points on the surface of the appearance surface unit. The current trajectory sampling point and the two boundary control points are connected by a spline curve to generate a local segment of the initial boundary curve at the current trajectory sampling point. All local segments of the initial boundary curve generated at all trajectory sampling points are sequentially spliced ​​together according to the order of the trajectory sampling points, and tangential continuity adjustment is performed at the splicing points of adjacent local segments to make the tangent vector directions of the common endpoints of adjacent local segments consistent, thereby generating a continuous parting surface initial boundary curve. The starting boundary curve of the parting surface is processed by adding a starting boundary identifier. The unit identifier of the appearance surface unit and the sequential number of the trajectory sampling point are added to each curve segment of the starting boundary curve of the parting surface to obtain a set of starting boundary curves of the parting surface carrying the starting boundary identifier.

5. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 1, characterized in that, The process involves calling the set of parting surface initial boundary curves carrying the initial boundary identifier and the initial parting line spatial position sequence to perform parting surface spatial extension generation. Starting from the parting surface initial boundary curve, the process extends along the initial parting line spatial position sequence into the internal solid region of the injection molded part to generate parting surface spatial sheet units, resulting in a preliminary geometric model of the parting surface composed of parting surface spatial sheet units, including: Extract the set of spatial coordinate points of each curve segment of the parting surface starting boundary curve from the set of parting surface starting boundary curves carrying the starting boundary identifier, and use the set of spatial coordinate points of each curve segment as the starting contour line of the parting surface spatial extension to obtain the set of starting contour lines. Each starting contour line corresponds to a parting surface starting boundary curve segment on an appearance surface unit. For each starting contour line in the set of starting contour lines, perform starting contour line discretization processing, extract contour sampling points at equal intervals along the arc length direction of each starting contour line, and obtain the contour sampling point sequence corresponding to each starting contour line. Each contour sampling point carries the spatial coordinates of the contour sampling point and the arc length parameter on the starting contour line. For each contour sampling point, perform parting surface extension direction vector calculation processing to obtain the surface unit normal vector of the appearance surface unit to which the contour sampling point belongs at the position of the contour sampling point, obtain the trajectory direction vector of the contour sampling point at the corresponding position in the initial parting line spatial position sequence, and calculate the vector product of the surface unit normal vector and the trajectory direction vector as the initial extension direction vector of the parting surface at the contour sampling point. For each contour sampling point, perform an extension path generation process. Starting from the contour sampling point, generate an extension ray along the initial extension direction vector towards the internal solid area of ​​the injection molded part. Calculate the first intersection point between the extension ray and the boundary of the internal solid area of ​​the injection molded part. Record the spatial line segment from the starting point to the first intersection point as the extension path line segment of the contour sampling point, and obtain the extension path line segment corresponding to each contour sampling point. Perform extended path surface skinning on all extended path segments corresponding to all contour sampling points. Align the corresponding positions of the extended path segments of adjacent contour sampling points according to the arc length parameter and connect them into quadrilateral mesh units. Splice all quadrilateral mesh units to generate a continuous spatial surface sheet, which serves as the parting surface spatial sheet unit obtained from the initial contour line. All parting surface spatial sheet elements corresponding to the initial contour lines are spliced ​​together according to the arrangement order of the initial contour lines on the initial boundary curve of the parting surface. The overlapping mesh nodes at the common boundary of adjacent parting surface spatial sheet elements are merged to generate a preliminary geometric model of the parting surface composed of multiple seamlessly connected parting surface spatial sheet elements.

6. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 1, characterized in that, The step of performing intersection avoidance trimming on the preliminary geometric model of the parting surface and the appearance continuity transition area, and removing the portion of the parting surface spatial sheet unit that intrudes into the appearance continuity transition area, to obtain the geometric model of the target mold parting surface, includes: Obtain the spatial boundary description of each appearance continuity transition region in the injection molded part appearance data set. The spatial boundary description includes the spatial coordinate point sequence of the boundary contour line of each appearance continuity transition region on the appearance surface of the injection molded part. Traverse each parting surface spatial sheet unit in the preliminary geometric model of the parting surface, extract the spatial coordinates of all grid nodes of the parting surface spatial sheet unit, compare the spatial coordinates of each grid node with the spatial boundary description of each appearance continuity transition area, and determine whether each grid node is located within the internal space of any appearance continuity transition area. When at least one mesh node in any parting surface space sheet element is located within the internal space of the appearance continuity transition region, the parting surface space sheet element is marked as a parting surface space sheet element to be trimmed. For each parting surface spatial sheet element to be trimmed, perform intrusion region boundary extraction processing, calculate the intersection coordinates of all mesh edges of the parting surface spatial sheet element to be trimmed with the spatial boundary contour line of the appearance continuity transition area, arrange all intersection coordinates in the order of their appearance on the spatial boundary contour line to form an intrusion region boundary curve, and obtain the set of intrusion region boundary curves. For each parting surface space sheet unit to be trimmed, perform the intrusion region internal mesh unit deletion process. Using the boundary curve of the intrusion region as the dividing line, divide the parting surface space sheet unit to be trimmed into a set of mesh units in the retained region and a set of mesh units in the region to be deleted. Delete all mesh units in the set of mesh units in the region to be deleted from the parting surface space sheet unit to be trimmed. The remaining set of mesh elements in the retained area after deletion is processed by trimming the boundary and reconstructing the boundary. The boundary mesh edges located on the boundary curve of the intrusion area are extracted from the set of mesh elements in the retained area. The endpoints of the boundary mesh edges are connected in spatial order to generate a new trimmed boundary curve. The internal mesh nodes in the set of mesh elements in the retained area are reconnected to generate a closed trimmed parting surface spatial sheet element. Merge all trimmed parting surface space sheet elements with the unmarked parting surface space sheet elements to generate a target mold parting surface geometry model that does not contain any intrusive parts of the appearance continuity transition region.

7. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 6, characterized in that, The step of performing the intrusion region deletion process on each parting surface spatial sheet unit to be trimmed, dividing the parting surface spatial sheet unit to be trimmed into a set of retained region mesh units and a set of deleted region mesh units using the boundary curve of the intrusion region as the dividing line, includes: Obtain each intrusion region boundary curve from the set of intrusion region boundary curves of the spatial sheet unit to be trimmed, and extract the spatial coordinate point sequence of each intrusion region boundary curve and the spatial coordinate point sequence of the spatial boundary contour of the appearance continuity transition area corresponding to the intrusion region boundary curve. The mesh cell classification and traversal process is performed on the spatial sheet unit of the parting surface to be trimmed. Each mesh cell in the spatial sheet unit of the parting surface to be trimmed is obtained in turn. The spatial coordinates of all mesh nodes of each mesh cell are extracted, and the spatial position relationship between the spatial coordinates of each mesh node and the boundary curve of the intrusion area is determined. When the spatial coordinates of all grid nodes of a grid cell are located on the same side of the boundary curve of the intrusion area and that side is in a direction away from the center of the appearance continuity transition area, the grid cell is marked as a reserved area grid cell and added to the reserved area grid cell set. When the spatial coordinates of at least one grid node in a grid cell are located on the side of the boundary curve of the intrusion area that is close to the center of the appearance continuity transition area, the grid cell is marked as a candidate grid cell to be deleted and added to the candidate set to be deleted. For each candidate mesh cell to be deleted in the candidate set to be deleted, the intersection judgment process between the mesh cell and the boundary curve of the intrusion area is performed. The number of intersection points between each mesh edge of the candidate mesh cell to be deleted and the boundary curve of the intrusion area is calculated. When the number of intersection points is greater than zero, the coordinates of the intersection points located on the boundary curve of the intrusion area and the coordinates of the mesh nodes located on the side of the boundary curve of the intrusion area near the center of the appearance continuity transition area are extracted from the candidate mesh cell to be deleted. Based on the extracted intersection coordinates and grid node coordinates, the candidate grid cells to be deleted are subdivided into multiple sub-grid cells. The candidate grid cells to be deleted are re-divided into multiple sub-grid cells using the original grid nodes and newly added intersection coordinates of the candidate grid cells to be deleted as vertices. Among them, the sub-grid cells located on the side of the boundary curve of the intrusion area that is closer to the center of the appearance continuity transition area are marked as the grid cells to be deleted and added to the set of grid cells to be deleted. The sub-grid cells located on the side of the boundary curve of the intrusion area that is farther away from the center of the appearance continuity transition area are marked as the grid cells to be retained and added to the set of grid cells to be retained. When all the grid nodes of the candidate grid cell to be deleted are located on the side of the boundary curve of the intrusion area that is close to the center of the appearance continuity transition area and no grid edge intersects with the boundary curve of the intrusion area, the candidate grid cell to be deleted is marked as a grid cell to be deleted and added to the set of grid cells to be deleted. Remove all mesh cells from the set of mesh cells in the region to be deleted from the parting surface space sheet element to be trimmed, and retain all mesh cells in the set of mesh cells in the region to be retained as the trimmed parting surface space sheet element.

8. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 1, characterized in that, After performing intersection avoidance trimming on the preliminary geometric model of the parting surface and the appearance continuity transition area, and cutting off the portion of the parting surface spatial sheet unit that intrudes into the appearance continuity transition area to obtain the geometric model of the target mold parting surface, the method further includes: The gap filling process between the parting surface boundary and the appearance continuity transition area is performed on the geometric model of the target mold parting surface. The spatial coordinate point sequence of each trimmed boundary curve and the spatial coordinate point sequence of the spatial boundary contour line of the appearance continuity transition area corresponding to the trimmed boundary curve are extracted from the geometric model of the target mold parting surface. Calculate the shortest spatial distance between each trimming boundary point on the trimming boundary curve and the spatial boundary contour of the appearance continuity transition area. When the shortest spatial distance is greater than zero, generate a filled connection line between the trimming boundary point and the nearest point on the spatial boundary contour. Construct a gap-filled surface sheet based on all fill connecting lines, with the trimmed boundary curve as the first boundary and the spatial boundary contour line of the appearance continuity transition area as the second boundary, and all fill connecting lines as guide lines. Construct a gap-filled surface sheet connecting the trimmed boundary curve and the spatial boundary contour line through the lofted surface generation method. The gap-filling curved surface sheet is merged with the target mold parting surface geometry model. The overlapping mesh nodes on the boundary of the gap-filling curved surface sheet and the trimmed boundary curve of the target mold parting surface geometry model are merged to generate the target mold parting surface geometry model after gap filling. After gap filling, the contact boundary smoothing process between the parting surface and the continuous transition area of ​​the appearance is performed on the geometric model of the target mold parting surface. All boundary lines in contact with the continuous transition area of ​​the appearance are extracted from the geometric model of the target mold parting surface after gap filling. The angle between the surface tangent vector of the target mold parting surface geometric model and the surface tangent vector of the continuous transition area of ​​the appearance is calculated at each boundary point on each boundary line. When the angle is greater than a preset smoothing angle threshold, the surface normal vector of the target mold parting surface geometric model is rotated and adjusted at the boundary point so that the angle between the surface tangent vector of the adjusted target mold parting surface geometric model and the surface tangent vector of the continuous transition area of ​​the appearance is reduced to below the smoothing angle threshold. The target mold parting surface geometric model after contact boundary smoothing is obtained and output as the final mold parting surface geometric model.

9. The intelligent drawing method for mold parting surface combined with injection molded part appearance design according to claim 8, characterized in that, After performing contact boundary smoothing processing on the parting surface and the continuous transition area between the parting surface and the appearance of the target mold parting surface geometric model after gap filling, the method further includes: Obtain all parting surface spatial sheet elements in the final mold parting surface geometry model, perform parting surface thickness direction expansion processing on each parting surface spatial sheet element, extend the preset thickness expansion distance along the surface normal vector direction of the mesh node to the internal solid area of ​​the injection molded part with each mesh node as the reference point, and generate the spatial coordinates of the thickness expansion endpoint corresponding to each mesh node. Connect each grid node with its corresponding thickness extension endpoint spatial coordinates to form a thickness direction line segment. Based on all thickness direction line segments and the grid topology relationship of the original parting surface spatial sheet unit, construct the thickness-extended solid grid unit. Each thickness-extended solid grid unit is formed by a grid unit of the original parting surface spatial sheet unit and its corresponding grid unit obtained by projection along the thickness direction. Perform solid mesh unit combination processing on all thickness-expanded solid mesh units, merge the common surfaces of adjacent thickness-expanded solid mesh units, delete duplicate mesh nodes on the common surfaces, and generate a continuous solid geometry model composed of solid mesh units. The interference check process between the parting surface entity and the internal structure of the injection molded part is performed on the continuous solid geometry model to obtain the boundary description of the internal solid region of the injection molded part, and the spatial coordinates of each mesh node of the continuous solid geometry model and the positional relationship between the boundary description of the internal solid region of the injection molded part are calculated. When the spatial coordinates of any mesh node in the continuous solid geometry model are outside the boundary description of the solid region inside the injection molded part, the mesh node is backed back by a preset back distance in the opposite direction of the line segment of its thickness direction until the spatial coordinates of the mesh node return to the boundary description of the solid region inside the injection molded part, thus obtaining the continuous solid geometry model after interference correction. The continuous solid geometry model after interference correction is output as a mold parting surface structure model with solid thickness.

10. A mold parting surface intelligent drawing system that integrates injection molded part appearance design, characterized in that, The invention includes a processor and a computer-readable storage medium storing machine-executable instructions, which, when executed by the processor, implement the intelligent drawing method for mold parting surfaces in conjunction with the appearance design of injection molded parts as described in any one of claims 1-9.

Images