A method and system for rapid generation of injection mold structures based on parametric design

Abstract

This application provides a method and system for rapid generation of injection mold structures based on parametric design, relating to the field of injection mold design technology. First, it receives a parametric description group of an injection molded product model containing a set of product external shape parameters and a set of product internal shape parameters. Next, it calls a cavity generation rule library based on the product external shape parameter set to obtain a cavity geometric parameter group, and calls a core generation rule library based on the product internal shape parameter set to obtain a core geometric parameter group. Then, it performs assembly clearance matching processing on the cavity and core geometric parameter groups to generate a mold assembly parameter group. Finally, it calls a mold frame parameter component library based on the mold assembly parameter group to generate a complete mold parameter file containing a fixed mold component parameter group and a moving mold component parameter group. This invention can improve the efficiency and quality of injection mold structure design, and reduce costs and cycle time.

Description

Technical Field

[0001] This application relates to the field of injection mold design technology, and more specifically, to a method and system for rapid generation of injection mold structures based on parametric design. Background Technology

[0002] In the field of injection mold design and manufacturing, traditional methods mainly rely on engineers' manual design and experience accumulation. Engineers need to first conduct a detailed analysis of the shape, size and other characteristics of the injection molded product, and then, based on their own professional knowledge and experience, gradually determine the various structural parameters of the mold, such as the size and shape of the cavity and core, the assembly clearance of the mold, the position of the parting surface and the demolding direction, etc.

[0003] The aforementioned traditional design methods have several drawbacks. Firstly, the design process is cumbersome and time-consuming, requiring engineers to spend considerable time and effort on repeated calculations and modifications. This is especially true when dealing with injection-molded products with complex shapes, significantly extending the design cycle and leading to low mold development efficiency, failing to meet rapidly changing market demands. Secondly, because the design process heavily relies on the individual experience of engineers, the results from different engineers may vary considerably, making it difficult to guarantee the accuracy and consistency of mold structure design. This can easily lead to mold manufacturing defects or injection-molded product quality problems due to unreasonable design, increasing the cost of mold debugging and modification. Furthermore, traditional methods are insufficient in handling parametric design, making it difficult to quickly adjust the mold structure according to minor changes in product parameters, lacking flexibility and versatility. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method and system for rapid generation of injection mold structures based on parametric design.

[0005] According to a first aspect of this application, a method for rapid generation of injection mold structures based on parametric design is provided, the method comprising: Receive a parametric description set of the injection molded product model, the parametric description set including a product external shape parameter set and a product internal shape parameter set; The cavity generation rule library is called according to the product shape parameter set to perform matching and mapping processing to obtain the cavity geometric parameter set, which includes the cavity size parameter set and the cavity curvature parameter set; The core generation rule library is called according to the product internal shape parameter set to perform matching and mapping processing to obtain the core geometric parameter set, which includes the core size parameter set and the core protrusion parameter set; The cavity geometry parameter set and the core geometry parameter set are subjected to assembly clearance matching processing to generate a mold assembly parameter set, which includes a parting surface curve parameter set and a demolding direction parameter set. The mold assembly parameter group is used to call the mold frame parameter component library for component matching and generate a complete mold parameter file. The complete mold parameter file includes a fixed mold component parameter group and a moving mold component parameter group.

[0006] According to a second aspect of this application, a rapid generation system for injection mold structures based on parametric design is provided. The rapid generation system for injection mold structures based on parametric design 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 rapid generation system for injection mold structures based on parametric design implements the aforementioned rapid generation method for injection mold structures based on parametric design.

[0007] Based on any of the above aspects, the technical effect of this application is as follows: By receiving the parametric description set of the injection molded product model, and performing matching and mapping processing based on the product's external shape parameter set and the cavity generation rule library, a cavity geometric parameter set is obtained. This makes the cavity design more consistent with the product's external shape requirements, improving the efficiency and accuracy of cavity design. Similarly, based on the product's internal shape parameter set and the core generation rule library, a core geometric parameter set is obtained, ensuring precise matching between the core and the product's internal shape. Assembly clearance matching processing is performed on the cavity and core geometric parameter sets to generate a mold assembly parameter set containing parting surface curve parameter sets and demolding direction parameter sets. This effectively solves key problems in the mold assembly process, ensuring reasonable assembly and smooth demolding between various mold components. Finally, based on the mold assembly parameter set and the mold frame parameter component library, a complete mold parameter file containing fixed mold component parameter sets and moving mold component parameter sets is generated. This achieves rapid and complete generation of the mold structure, greatly improving the design efficiency and quality of injection mold structures, and reducing mold development costs and cycles. Attached Figure Description

[0008] Figure 1 A flowchart illustrating the rapid generation method for injection mold structure based on parametric design provided in this application embodiment is shown. Figure 2 A schematic diagram of the component structure of the rapid generation system for injection mold structure based on parametric design provided in this application embodiment is shown. Detailed Implementation

[0009] Figure 1 This paper illustrates a flowchart of a method and system for rapid generation of injection mold structures based on parametric design, as provided in an embodiment of this application. The detailed steps include: Step S110: Receive the parametric description set of the injection molded product model, the parametric description set including the product external shape parameter set and the product internal shape parameter set.

[0010] The product's external shape parameter set includes a set of external boundary curves and a set of surface curvature gradients. The external boundary curve set consists of multiple spatial curve segments, each containing a start-point coordinate triplet, an end-point coordinate triplet, and a curve type identifier. The surface curvature gradient set is a multi-dimensional numerical array used to characterize the rate of curvature change of the product's external surface at different parameter domain locations. The product's internal shape parameter set includes a set of internal cavity boundary surfaces, a set of internal protrusion position coordinates, and a set of internal protrusion height dimensions. The internal cavity boundary surface set consists of a set of closed parametric surface patches. The internal protrusion position coordinates record the coordinates of the reference point for each protrusion feature in three-dimensional space. The internal protrusion height dimensions record the vertical distance from the product's internal reference surface to the top surface of each protrusion feature.

[0011] Step S120: Based on the product shape parameter set, call the cavity generation rule library for matching and mapping processing to obtain the cavity geometric parameter set, which includes the cavity size parameter set and the cavity curvature parameter set.

[0012] The process involves calling the cavity generation rule library. Parsing the product's outer boundary curves from the product shape parameter set, inputting them into the projection rule module, and projecting them onto the parting surface along the normal direction to generate projected boundaries. Extracting the product surface curvature gradient set, inputting it into the shrinkage compensation rule layer, and multiplying it by the material shrinkage compensation coefficient to obtain the shrinkage compensation curvature parameter set. Calling the skeleton construction rule, the process merges the projected boundaries and the shrinkage compensation curvature parameter set to generate the initial cavity skeleton wireframe. Calling the parameter extraction rule, the process parses the cavity dimension parameter set and cavity curvature parameter set from the initial cavity skeleton wireframe.

[0013] Step S121: Analyze the product outer boundary curve group in the product shape parameter set, and determine the projection contour polygon group of the cavity on the parting surface based on the product outer boundary curve group.

[0014] Extract the product's outer boundary curve set from the product's shape parameter set. Determine the parting surface spatial plane. For each spatial curve segment in the product's outer boundary curve set, project its start point, end point, and internal points sampled at equal parameter intervals onto the parting surface along the normal direction of the parting surface to obtain projection points. Connect the projection points in their original order to form projection curve segments. Traverse all projection curve segments and calculate the Euclidean distance between the endpoints of adjacent projection curve segments. When the distance is less than a tolerance threshold, force the endpoint coordinates to align, connecting the beginning and end of the projection curve segments into a closed loop. Select the closed loop with the largest area and sort its vertices counterclockwise to generate a projection contour polygon set.

[0015] Step S122: Extract the product surface curvature gradient group from the product shape parameter set, and input the product surface curvature gradient group into the cavity surface shrinkage compensation rule layer for processing to obtain the shrinkage compensation curvature parameter group.

[0016] Extract the surface curvature gradient set from the product shape parameter set, denoted as Guv. Input Guv into the cavity surface shrinkage compensation rule layer. Read the linear shrinkage rate parameter S of the injection molding material and calculate the compensation coefficient A, A=1 / (1-S). Perform the operation on each component in Guv: Gcuv=Guv×A^2. Perform the above operation on all parameter coordinates to generate the shrinkage compensation curvature parameter set Gcuv.

[0017] Step S123: Generate an initial skeleton wireframe group for the cavity based on the projected contour polygon group and the shrinkage compensation curvature parameter group. The initial skeleton wireframe group for the cavity includes a cavity depth dimension chain group and a cavity bottom fillet radius group.

[0018] A two-dimensional boundary mesh framework is constructed on the parting surface based on the vertex sequence of the projected contour polygon group. For each mesh node, the compensated curvature gradient value is obtained by bilinear interpolation sampling from Gcuv based on its two-dimensional coordinates. This compensated curvature gradient value is integrated along the normal direction of the parting surface from the boundary of the projected contour polygon to the mesh node position to obtain the depth offset Dij of that node. The Dij of all nodes constitutes a depth distribution field, from which the depth value sequence at different positions along the boundary of the projected contour polygon is extracted as the cavity depth dimension chain group. Based on the local curvature changes of the bottom surface in the depth distribution field, the curvature radius value sequence of each corner region is calculated as the cavity bottom fillet radius group. The mesh node coordinates in the depth distribution field are combined with Dij to generate the initial skeleton wireframe group of the cavity.

[0019] Step S124: Call the cavity draft angle generation rule to add draft angle to the initial skeleton wireframe group of the cavity, and generate the cavity sidewall draft angle parameter group, which contains draft angle value groups for different depth sections.

[0020] The cavity draft angle generation rules are invoked. The sidewall mesh nodes in the initial cavity skeleton wireframe group are analyzed, and the angle between the normal vector at each sidewall node and the parting surface normal direction is calculated. The cavity sidewall is divided into multiple segments along the depth direction, each segment corresponding to a depth interval. For each depth segment, the normal angles of all sidewall nodes within that segment are calculated, and the median value of these calculations is taken as the reference draft angle θd for that depth segment. A preset draft angle safety margin β is added to θd to obtain the final draft angle αd = θd + β for that depth segment. The αd values ​​for all depth segments are arranged in ascending order of depth to generate the cavity sidewall draft angle parameter set.

[0021] Step S125: Perform boundary fusion processing on the surface patch boundary curve group in the shrinkage compensation curvature parameter group and the cavity sidewall angle parameter group to obtain the cavity surface patch topology group, which includes the continuity order identifier group of adjacent surface patches.

[0022] The boundary curve group Bs of each surface patch is extracted from Gcuv. The αd in the cavity sidewall angle parameter group is converted into the boundary tilt constraint of the sidewall surface. For each boundary curve, the desired continuity order Cl is determined according to the type of its adjacent surface patches. The value of Cl includes C0, C1, or C2. Using the boundary curve and its Cl as constraints, a surface skinning algorithm is used to generate transition surface patches connecting the bottom surface and the sidewall surface. All generated transition surface patches, together with the original bottom surface patch and sidewall surface patch, constitute the cavity surface patch topology group, and the continuity order identifier group between each pair of adjacent surface patches is recorded.

[0023] Step S126: Perform surface smoothing filtering on the cavity surface topology group to generate a cavity surface control point offset group, which is used to adjust the local unevenness of the cavity surface.

[0024] Each surface patch in the cavity surface patch topology group is converted into a Bezier representation, and the control point mesh Pij for each surface patch is obtained. For each control point Pij, its Laplacian coordinates Lij are calculated, Lij = Pij - (Pi-1j + Pi+1j + Pij-1 + Pij+1) / 4. A smoothing energy function E = Σ(Lnew_ij - Lij)^2 + λ × Σ(Pnew_ij - Pij)^2 is constructed, where Lnew_ij is the Laplacian coordinate of the updated control point, Pnew_ij is the updated control point coordinate, and λ is the weight coefficient. By solving the system of linear equations that minimizes E, the optimal offset ΔPij = Pnew_ij - Pij for each control point is obtained. All ΔPij are organized into a cavity surface control point offset group.

[0025] Step S127: Generate a cavity solid boundary structure group based on the cavity depth dimension chain group and the cavity surface control point offset group. The cavity solid boundary structure group includes the surface normal group and edge curvature group of the cavity space.

[0026] The depth values ​​from the cavity depth dimension chain are assigned to the corresponding mesh nodes in the initial cavity skeleton wireframe group to generate the three-dimensional spatial coordinates of the cavity bottom surface. ΔPij is applied to the control points of each surface patch in the cavity to update the geometry of the surface patches. Triangular patch representations are extracted from the updated bottom and sidewall surfaces to obtain the cavity solid boundary structure group. For each triangular patch, its normal vector Nf = (V1 × V2) / |V1 × V2| is calculated, where V1 and V2 are the two edge vectors of the triangular patch. For each edge, its curvature value Ke = Δθ / Δs is calculated, where Δθ is the angle between the tangential vectors at the two endpoints of the edge, and Δs is the arc length of the edge.

[0027] Step S128: Extract the cavity volume expression group and cavity area expression group from the cavity entity boundary structure group, and use the cavity volume expression group and cavity area expression group as the volume parameter subset and area parameter subset in the cavity size parameter group, respectively. Generate a cavity geometric parameter group based on the cavity size parameter group containing the volume parameter subset and area parameter subset and the cavity surface control point offset group. The cavity geometric parameter group includes the cavity size parameter group and the cavity curvature parameter group.

[0028] Extract the vertex coordinates of all triangular facets from the cavity solid boundary structure group. For each triangular facet, calculate its directed volume contribution value Vtet=(Pa·(Pb×Pc)) / 6, where Pa, Pb, and Pc are the vector formed by the coordinates of the three vertices of the triangular facet and the origin. Sum the directed volume contributions of all triangular facets and take the absolute value to obtain the volume value Vcv=|ΣVtet| in the cavity volume expression group. For each triangular facet, calculate its area contribution value Atri=|(Pb-Pa)×(Pc-Pa)| / 2. Sum the area contributions of all triangular facets to obtain the area value Acv=ΣAtri in the cavity area expression group. Use Vcv as a subset of volume parameters in the cavity dimension parameter group, and use Acv as a subset of area parameters in the cavity dimension parameter group. Encapsulate the cavity dimension parameter group containing the volume parameter subset and the area parameter subset together with the cavity surface control point offset group ΔPij to generate the cavity geometric parameter group.

[0029] Step S130: Based on the product internal shape parameter set, call the core generation rule library for matching and mapping processing to obtain the core geometric parameter set, which includes the core size parameter set and the core protrusion parameter set.

[0030] The core generation rule library is invoked. The internal cavity boundary surface group of the product's internal shape parameter set is parsed and used as the inverse constraint of the core's outer contour, generating the core's outer contour surface group. The coordinate groups of the internal protrusion positions and the height dimensions of the internal protrusions are extracted and mapped to the distribution group of the core surface recess positions and the depth dimensions of the recesses. The core draft angle generation rule is invoked to add draft angles to the core's basic shape boundary group, generating the distribution group of draft angle angles on the core's side surfaces. Boolean difference operations are performed on the core's basic shape boundary group and the core surface recess depth dimensions to generate a complete core geometric description including recess features. From this complete core geometric description, the span dimensions of the core in each direction are extracted as the core dimension parameter group, and the height, width, and shape parameters of each protrusion feature are extracted as the core protrusion parameter group.

[0031] Step S131: Analyze the product internal cavity boundary surface group in the product internal shape parameter set, and determine the core outer contour surface group based on the product internal cavity boundary surface group. The core outer contour surface group and the product internal cavity boundary surface group have a spatial complementary filling relationship.

[0032] Extract the internal cavity boundary surface group Hin from the product's internal shape parameter set. Offset each surface patch in Hin outwards along its normal direction by a preset gap value δ, δ = δ0 + δf, where δ0 is the basic gap constant and δf is a function term related to the surface curvature. The offset operation is as follows: for each point Ps on the surface patch, its offset point Po = Ps + δ × Ns, where Ns is the unit normal vector at that point. All offset points constitute new surface patches, and all new surface patches together form the outer contour surface group Cc of the core. The spatial relationship between Cc and Hin is: Cc is located inside Hin, and the normal distance between them is everywhere equal to δ.

[0033] Step S132: Extract the coordinate set of the product's internal protrusion position and the dimension set of the product's internal protrusion height from the product's internal shape parameter set, and generate the distribution set of the core surface pit position and the dimension set of the core surface pit depth based on the coordinate set of the product's internal protrusion position and the dimension set of the product's internal protrusion height.

[0034] Extract the coordinate set Qp of the internal protrusion positions and the height dimension set Hp of the internal protrusions from the product's internal shape parameter set. For each protrusion feature, map the coordinate values ​​in Qp onto the surface of the core's outer contour surface set Cc to obtain the coordinates Qk of the recess center point. Add a preset recess depth margin Δd to the height value in Hp to obtain the recess depth value Dk = Hp + Δd. Collect the Qk corresponding to all protrusion features as the recess position distribution set on the core surface, and collect the Dk corresponding to all protrusion features as the recess depth dimension set on the core surface.

[0035] Step S133: Perform feature spatial position fusion processing based on the outer contour surface group of the core and the distribution group of pit positions on the core surface to generate the basic shape boundary group of the core body. The basic shape boundary group of the core body includes the outer surface sheet group of the core body and the coordinate group of the volume center point of the core body.

[0036] The core's outer contour surface group Cc is used as the basic geometry. For each pit center point in the pit location distribution group Qk on the core surface, the pit region boundary is marked at the corresponding position in Cc. Using Boolean subtraction, the volume corresponding to the pit region is removed from Cc. The removal process is as follows: for each pit, a prism or pyramid geometry with depth Dk and base shape is constructed. This geometry is then subtracted from Cc using Boolean subtraction to generate a new surface group with pit features. After all pits are processed, the core's basic shape boundary group Bc is obtained. All surface patches constituting the core's outer surface are extracted from Bc as the outer surface patch group Ss. The coordinates of the volume center point Gc of Bc are calculated as Gc = (∫xdV / ∫dV, ∫ydV / ∫dV, ∫zdV / ∫dV), where the integration domain is the entire volume space enclosed by Bc.

[0037] Step S134: Call the core demolding draft angle generation rule to add demolding draft angle to the basic shape boundary group of the core body, and generate a core side surface demolding draft angle distribution group, which includes demolding draft angle value groups corresponding to different height segments.

[0038] The core draft angle generation rule is invoked. The side surface region in the core body's basic shape boundary group Bc is analyzed. The core side surface is divided into K segments along the height direction, each segment corresponding to a height interval [hk-1, hk]. For the k-th segment, the angle between the normal vector of all side surface mesh nodes within the segment and the draft direction vector is calculated, and the median value of this angle is taken as the reference draft angle γk for that segment. A preset draft angle safety increment γs is added to γk to obtain the final draft angle ηk = γk + γs for that segment. The ηk values ​​of all segments are arranged in ascending order of height to generate a core side surface draft angle distribution group.

[0039] Step S135: Generate a core body root fillet radius distribution group based on the outer surface sheet group in the core body basic shape boundary group. The core body root fillet radius distribution group includes a fillet radius value group for the connection part between the core body and the core mounting plate.

[0040] Extract the boundary curves of the connection points between the core body base shape boundary group Bc and the outer surface sheet group Ss. For each sampling point on the boundary curve, calculate the dihedral angle φj between the core side surface and the mounting plate surface at that point. Set the functional relationship between the fillet radius Rf and the dihedral angle φj as: Rf = R0 + R1 × (π / 2 - φj), where R0 is the reference fillet radius and R1 is the scaling factor. Calculate Rf for all sampling points and take the median value of all Rf values ​​as the fillet radius value of the connection point. If there are multiple independent connection points, calculate the fillet radius value for each point separately, and collect the fillet radius values ​​of all points into a core body root fillet radius distribution group.

[0041] Step S136: Perform fillet processing on the bottom sidewall transition of the core body basic shape boundary group and the core surface pit depth dimension group to generate a pit feature fillet radius group. The pit feature fillet radius group includes the fillet radius value group of the bottom corner position and the sidewall edge position of each pit feature.

[0042] For each pit in the core surface pit location distribution group Qk, extract the bottom boundary curve and sidewall edge curve of the pit from the core body basic shape boundary group Bc. For the pit bottom corner position, calculate the angle ψb between the bottom surface and the sidewall, and set the bottom corner fillet radius Rb=Rb0+Rb1×(π / 2-ψb). For the pit sidewall edge position, calculate the angle ψe between adjacent two sidewalls, and set the edge fillet radius Re=Re0+Re1×(π / 2-ψe). Collect the Rb and Re of all pits into a pit feature transition fillet radius group.

[0043] Step S137: Based on the distribution group of the demolding draft angle on the side surface of the core and the distribution group of the radius of the fillet at the root of the core body, perform shape correction processing on the basic shape boundary group of the core body to generate the corrected boundary group of the core body.

[0044] The ηk from the core side surface draft angle distribution group is applied to the side surface of the core body basic shape boundary group Bc. For side surface mesh nodes whose height is within the k-th segment, they are expanded outward in a direction perpendicular to the draft direction by a distance Lexp = h × tan(ηk), where h is the height of the node from the bottom of the core. Simultaneously, the Rf from the core body root fillet radius distribution group is applied to the root connection region of Bc, replacing the original sharp corners with rounded transition surfaces of radius Rf. After all expansion and fillet replacement operations are completed, the corrected core body boundary group Bcm is generated.

[0045] Step S138: Based on the pit feature transition fillet radius group, perform local rounding processing on the pit geometry corresponding to the pit position distribution group on the core surface to generate a complete geometric description group of the core pit containing the transition fillet surface.

[0046] For each pit in the core surface pit location distribution group Qk, obtain its bottom boundary and sidewall edges. Apply the corresponding Rb from the pit feature transition fillet radius group to the pit bottom corner position, replacing the original sharp bottom corner with a rounded transition surface of radius Rb. Apply the corresponding Re to the pit sidewall edge position, replacing the original sharp edge with a rounded transition surface of radius Re. After all fillet replacement operations are completed, a core pit geometry description containing complete transition fillet surfaces is generated. Collect all pit geometry descriptions into a complete core pit geometry description group Pit.

[0047] Step S139: Perform Boolean difference operation on the modified boundary group of the core body and the complete geometric description group of the core recess to generate a core geometric parameter group, which includes a core size parameter group and a core protrusion parameter group.

[0048] The modified boundary group Bcm of the core body is used as the minuend geometry, and each pit geometry in the complete geometric description group Pit of the core pits is used as the subtrahend geometry. A Boolean difference operation is performed on each pit geometry sequentially: Btmp = Bcm - Pitk, where Pitk is the k-th pit geometry. After processing all pits, the final core geometry Cf is obtained. The maximum span dimensions of the core in the X, Y, and Z directions are extracted from Cf to form the core dimension parameter group Dc. The depth, opening width, bottom width, and sidewall angle of each pit feature are extracted from Cf to form the core protrusion parameter group Pc. Dc and Pc are then encapsulated together as the core geometry parameter group.

[0049] Step S140: Perform assembly gap matching processing on the cavity geometry parameter group and the core geometry parameter group to generate a mold assembly parameter group, which includes a parting surface curve parameter group and a demolding direction parameter group.

[0050] Extract the cavity opening boundary curve set from the cavity geometry parameter set, and extract the core maximum outer boundary curve set from the core geometry parameter set. Compare the overlap of the two sets of curves in three-dimensional space, calculate the average spatial distance and maximum spatial distance, and generate the cavity-core mating clearance distribution set. Determine the spatial plane position equation of the parting surface based on the clearance width of the uniformly distributed area in the mating clearance distribution set, and generate the parting surface curve parameter set. Extract the central axis direction vector of the cavity sidewall pull-out angle parameter set as the first candidate demolding direction, and extract the central axis direction vector of the core side surface demolding slope angle distribution set as the second candidate demolding direction. Compare the direction consistency of the two direction vectors, and generate a unified demolding direction vector when the included angle is less than a preset direction deviation threshold. Verify the demolding feasibility based on the unified demolding direction vector, identify the undercut area, and generate the side core pulling mechanism configuration requirement set. Generate the mold assembly parameter set based on the parting surface curve parameter set, the unified demolding direction vector, and the side core pulling direction vector set.

[0051] Step S141: Extract the cavity opening boundary curve group from the cavity geometry parameter group, and simultaneously extract the core maximum outer boundary curve group from the core geometry parameter group.

[0052] Extract the cavity opening boundary curve group Bco from the cavity geometry parameter group generated in step S128. This cavity opening boundary curve group describes the opening profile of the cavity at the parting surface. Extract the core maximum outer boundary curve group Bcmax from the core geometry parameter group generated in step S139. Both Bco and Bcmax are spatial closed curves located near the same parting surface plane.

[0053] Step S142: Compare the overlap of the cavity opening boundary curve group and the core maximum outer boundary curve group in three-dimensional space, and calculate the average spatial distance and maximum spatial distance between the two curve groups.

[0054] Project the cavity opening boundary curve group Bco and the core maximum outer boundary curve group Bcmax onto the parting surface plane to obtain two-dimensional closed curves Bc2d and Bm2d. For each point Pci on Bc2d, search for the nearest point Pmni on Bm2d and calculate the Euclidean distance between the two points, di = |Pci - Pmni|. Calculate the arithmetic mean davg = (Σdi) / N of all di, where N is the total number of sampling points on Bc2d. Calculate the maximum value dmax = max(di) of all di. Use davg and dmax as statistical characteristic quantities of the cavity-core mating clearance.

[0055] Step S143: Generate a cavity-core mating clearance distribution group based on the average spatial distance value group and the maximum spatial distance value group, wherein the cavity-core mating clearance distribution group includes clearance width value groups for different mating areas.

[0056] The parting plane is divided into multiple sector-shaped regions or grid regions. For each region Rz, the distance di between corresponding points Bc2d and Bm2d within the region is calculated, and the median value of all di values ​​within the region is taken as the fit clearance width gz for that region. The gz values ​​of all regions are collected into a cavity-core fit clearance distribution group Gg={g1, g2, ..., gM}, where M is the total number of regions.

[0057] Step S144: Determine the spatial plane position equation of the parting surface based on the gap width values ​​of the uniformly distributed area in the cavity-core mating gap distribution group, and generate the parting surface curve parameter group, which includes the intersection parameter group of the parting surface and the cavity opening boundary curve group.

[0058] From the cavity-core mating clearance distribution group Gg, uniformly distributed regions with clearance width variation rates less than a preset threshold are selected, and clearance width value groups for these regions are extracted. Based on the median value gmid of this clearance width value group, the offset Δzp = gmid / 2 of the parting surface in the mold height direction is determined. The spatial planar position equation of the parting surface is: z = z0 + Δzp, where z0 is the average height of the maximum outer boundary curve group of the core. The intersection line of this parting surface plane and the cavity opening boundary curve group Bco is calculated, resulting in a series of intersection curve segments. These intersection curve segments are connected sequentially to generate the parting surface curve parameter group Cp.

[0059] Step S145: Extract the central axis direction vector of the cavity sidewall draft angle parameter group in the cavity geometry parameter group as the first candidate demolding direction vector, and simultaneously extract the central axis direction vector of the core side surface draft angle distribution group in the core geometry parameter group as the second candidate demolding direction vector. Perform direction consistency comparison processing on the first candidate demolding direction vector and the second candidate demolding direction vector, calculate the included angle value between the two direction vectors, and generate a unified demolding direction vector when the included angle value is less than the preset direction deviation threshold.

[0060] Extract the central axis direction vector Vcav of the draft angle of the cavity sidewall from the cavity sidewall draft angle parameter set. Extract the central axis direction vector Vcor ​​of the draft angle of the core side surface from the core side surface draft angle distribution set. Calculate the angle θali between Vcav and Vcor ​​= arccos((Vcav·Vcor) / (|Vcav|×|Vcor|)). When θali is less than the preset direction deviation threshold θth, use the weighted average of Vcav and Vcor ​​as the unified draft direction vector Vdr = (w1×Vcav+w2×Vcor) / (w1+w2), where w1 and w2 are weighting coefficients.

[0061] Step S146: Perform demolding feasibility verification processing on the cavity geometric parameter group and the core geometric parameter group according to the unified demolding direction vector, generate the cavity internal undercut area position group and the core external undercut area position group, and generate the lateral core pulling mechanism configuration requirement group according to the cavity internal undercut area position group and the core external undercut area position group. The lateral core pulling mechanism configuration requirement group includes the lateral core pulling direction vector group and the lateral core pulling distance value group.

[0062] The unified demolding direction vector Vdr is used as the main demolding direction. For each surface patch in the cavity geometry parameter set, the dot product between the normal vector Ncv at any point on the surface patch and the projection vector along the Vdr direction at that point is calculated. When the dot product is less than zero, the region where that point is located is an undercut region. The position coordinates of all undercut regions are collected into the cavity internal undercut region position group Uud. For each surface patch in the core geometry parameter set, the dot product between the normal vector and the demolding direction is calculated using the same method to generate the core external undercut region position group Lud. For each undercut region, its lateral core-pulling direction vector Vsl is determined. This direction vector is perpendicular to Vdr and points out of the undercut region. The maximum indentation depth of the undercut region in the Vsl direction is calculated as the lateral core-pulling distance Dsl. The Vsl and Dsl corresponding to all undercut regions are collected into the lateral core-pulling mechanism configuration requirement group.

[0063] Step S147: Generate a demolding direction parameter group based on the unified demolding direction vector and the lateral core-pulling direction vector group, wherein the demolding direction parameter group includes a main demolding direction vector and an auxiliary lateral core-pulling direction vector group.

[0064] The unified demolding direction vector Vdr is used as the primary demolding direction vector. All Vsl vectors extracted from the lateral core-pulling mechanism configuration requirement group are collected into an auxiliary lateral core-pulling direction vector group Vsls. Vdr and Vsls are then encapsulated together into a demolding direction parameter group Ddir.

[0065] Step S148: Generate a mold assembly parameter set based on the parting surface curve parameter set and the demolding direction parameter set, wherein the mold assembly parameter set includes the parting surface curve parameter set and the demolding direction parameter set.

[0066] The parting surface curve parameter group Cp generated in step S144 and the demolding direction parameter group Ddir generated in step S147 are associated and encapsulated to generate the mold assembly parameter group Masm=(Cp, Ddir). This mold assembly parameter group will be used in subsequent steps of calling the mold base parameter component library.

[0067] Step S150: Call the mold frame parameter component library according to the mold assembly parameter group to perform component matching processing and generate a complete mold parameter file. The complete mold parameter file includes a fixed mold component parameter group and a moving mold component parameter group.

[0068] The spatial height position of the parting surface in the mold thickness direction is determined based on the parting surface curve parameter group in the mold assembly parameter group, serving as the reference value for the boundary height between the fixed mold assembly and the moving mold assembly. The first dimensional constraint in the thickness direction of the fixed mold platen is calculated based on the maximum depth value of the cavity depth dimension chain group and the boundary height reference value. Standard specification parameter groups for fixed mold plates with thickness dimensions greater than this constraint are then retrieved from the mold frame parameter component library. Similarly, the first dimensional constraint in the thickness direction of the moving mold platen is calculated based on the maximum height value of the core dimension parameter group and the boundary height reference value. Standard specification parameter groups for moving mold plates with thickness dimensions greater than this constraint are then retrieved from the mold frame parameter component library. The minimum bounding rectangle dimension value group of the projected contour polygon group of the cavity space on the parting surface is extracted as the width and length constraints of the cavity layout area. Standard specification groups for fixed mold platen plane dimensions and moving mold platen plane dimensions with plane dimensions greater than this constraint are then retrieved from the mold frame parameter component library. Finally, initial parameter groups for the fixed mold assembly and moving mold assembly are generated based on the thickness, length, and width parameters of the fixed and moving mold plates. The system calls the guide pillar and guide bushing configuration rules to generate guide pillar length value groups, guide bushing depth value groups, and guide pillar distribution position coordinate groups. It then calls the ejector mechanism configuration rules to generate ejector pin distribution position coordinate groups and ejector pin length value groups. These component parameters are added to the initial parameter groups for both the fixed mold component and the moving mold component, generating a complete mold parameter file containing both parameter groups.

[0069] Step S151: Determine the spatial height position value of the parting surface in the mold thickness direction according to the parting surface curve parameter group in the mold assembly parameter group, and use the spatial height position value as the boundary height reference value between the fixed mold assembly and the moving mold assembly.

[0070] Extract the parting surface curve parameter set Cp from the mold assembly parameter set Masm. Calculate the spatial height coordinate value Zp of the plane where Cp lies. Use this Zp as the reference value for the boundary height between the fixed mold assembly and the moving mold assembly in the mold thickness direction.

[0071] Step S152: Calculate the first dimensional constraint in the thickness direction of the fixed template based on the maximum depth value of the cavity depth dimension chain group in the cavity geometry parameter group and the boundary height reference value; call the fixed template standard specification parameter group in the mold frame parameter component library whose thickness dimension is greater than the first dimensional constraint in the thickness direction of the fixed template; calculate the first dimensional constraint in the thickness direction of the moving template based on the maximum height value of the core dimension parameter group in the core geometry parameter group and the boundary height reference value; call the moving template standard specification parameter group in the mold frame parameter component library whose thickness dimension is greater than the first dimensional constraint in the thickness direction of the moving template.

[0072] Extract the maximum depth value Dcmax from the cavity depth dimension chain. The first dimensional constraint in the thickness direction of the fixed template is Tfmin = Zp + Dcmax + Tsc, where Tsc is the safe thickness at the bottom of the cavity. Search the mold frame parameter component library for fixed template standard specifications with a thickness greater than Tfmin, and obtain the corresponding thickness value Tf, length value Lf, and width value Wf. Extract the maximum height value Hcmax from the core dimension parameter group. The first dimensional constraint in the thickness direction of the moving template is Tmmin = (total height - Zp) + Hcmax + Tsd, where Tsd is the safe thickness at the bottom of the core. Search the mold frame parameter component library for moving template standard specifications with a thickness greater than Tmmin, and obtain the corresponding thickness value Tm, length value Lm, and width value Wm.

[0073] Step S153: Extract the minimum bounding rectangle size value group of the projection contour polygon group of the cavity space on the parting surface in the cavity geometry parameter group as the width and length constraints of the cavity layout area, and call the fixed template plane size standard specification group and the moving template plane size standard specification group in the mold frame parameter component library whose template plane size is greater than the width and length constraints of the cavity layout area.

[0074] Extract the coordinates of all vertices from the projected contour polygon group, and calculate the minimum and maximum X-coordinates and Y-coordinates of these vertices on the parting plane. The length Ll of the minimum bounding rectangle is calculated as Xmax - Xmin, and the width Wl is calculated as Ymax - Ymin. Search the mold frame parameter component library for standard specifications where the length Lf and width Wf of the fixed template are both greater than (Ll + Lmarg), where Lmarg is the layout edge margin. Simultaneously search the standard specifications where the length Lm and width Wm of the moving template are both greater than (Ll + Lmarg).

[0075] Step S154: Generate an initial set of parameters for the fixed mold component based on the first dimensional constraint in the thickness direction of the fixed mold and the standard specification group of the planar dimensions of the fixed mold. The initial set of parameters for the fixed mold component includes the values ​​of the fixed mold thickness, the fixed mold length, and the fixed mold width. Generate an initial set of parameters for the moving mold component based on the first dimensional constraint in the thickness direction of the moving mold and the standard specification group of the planar dimensions of the moving mold. The initial set of parameters for the moving mold component includes the values ​​of the moving mold thickness, the moving mold length, and the moving mold width.

[0076] The Tf, Lf, and Wf obtained in step S152 are used as the three core parameters of the initial parameter group Initf for the fixed mold component. The Tm, Lm, and Wm obtained in step S152 are used as the three core parameters of the initial parameter group Initm for the moving mold component.

[0077] Step S155: Call the guide post and guide sleeve configuration rules in the mold frame parameter component library, generate guide post length value group and guide sleeve depth value group according to the fixed template thickness value and the moving template thickness value, and generate guide post distribution position coordinate group according to the fixed template length value and width value; call the ejector mechanism configuration rules in the mold frame parameter component library, generate ejector pin distribution position coordinate group and ejector pin length value group according to the moving template plane size standard specification group and the core protrusion parameter group.

[0078] This step includes two independent sub-steps, namely step S1551 and step S1552.

[0079] Step S1551: Call the guide post and guide sleeve configuration rules in the mold frame parameter component library, generate guide post length value group and guide sleeve depth value group according to the fixed template thickness value and the moving template thickness value, and generate guide post distribution position coordinate group according to the fixed template length value and width value.

[0080] The configuration rules for guide pillars and guide sleeves in the template parameter component library include guide pillar diameter specification groups and guide sleeve inner diameter specification groups. Based on the smaller of the fixed template length Lf and width Wf, the guide pillar diameter specification group is called, and the corresponding guide pillar diameter value Dg is selected. Based on Dg, the guide sleeve inner diameter specification group is called, and the guide sleeve inner diameter value Db that matches Dg is selected. The total template thickness Ttot = Tf + Tm that the guide pillar needs to pass through is calculated, and the total guide pillar length Lg = Ttot + Lgt, where Lgt is the constant reserved length at the end of the guide pillar. The guide sleeve mounting hole depth Hbm = Tm - Hbb, where Hbb is the constant reserved thickness at the bottom of the guide sleeve. The guide sleeve guide section length Hbg = Tf. Based on Lf and Wf, the coordinate range of the four corner areas of the fixed template is determined, and the coordinates of the first, second, third, and fourth guide pillars are generated at positions at a first safety distance Ds1 from the edge of the fixed template. Based on Lf and Wf, determine the coordinate range of the midpoints of the long and wide sides of the fixed template. At a distance of Ds2 from the edge of the fixed template, generate the coordinates of the fifth, sixth, seventh, and eighth guide pillars. Combine these eight guide pillar coordinates into a guide pillar distribution position coordinate group Pg. Based on the corresponding position of each guide pillar coordinate in Pg on the moving template, generate a guide sleeve distribution position coordinate group Pb. Pb and Pg correspond one-to-one in the parting surface normal direction and have the same coordinates. Encapsulate Dg, Lg, and Pg into a complete guide pillar parameter configuration group, and encapsulate Db, Hbm, Hbg, and Pb into a complete guide sleeve parameter configuration group.

[0081] Step S1552: Call the ejector mechanism configuration rules in the mold frame parameter component library, and generate the ejector pin distribution position coordinate group and ejector pin length value group according to the moving template plane size standard specification group and the core protrusion parameter group.

[0082] The ejector pin diameter specification grading group included in the ejector mechanism configuration rules in the mold frame parameter component library is analyzed. The ejector pin diameter specification grading group is called based on the smaller of the moving template length Lm and width Wm, and the corresponding ejector pin diameter value De is selected. The position coordinate group Qk of each protrusion feature on the core surface and the projected area group Ap of each protrusion feature on the core surface are extracted from the core protrusion parameter group Pc. Based on the Qk corresponding to the protrusion features in Ap whose projected area is greater than a preset area threshold Ath, a first set of candidate ejector pin position coordinates is generated, including the center point coordinates of each large projected area protrusion feature. Based on the Qk corresponding to the protrusion features in Ap whose projected area is less than or equal to Ath, a second set of candidate ejector pin position coordinates is generated, including the edge area coordinates of each small projected area protrusion feature. The coordinate range of the moving template's central area is determined based on Lm and Wm, and a third set of candidate ejector pin position coordinates is generated within this central area coordinate range according to a uniform grid spacing value Gs. The first, second, and third sets of candidate ejector pin position coordinates are merged and deduplicated to generate the initial coordinate group Pei for the ejector pin distribution positions. The total ejector pin length, Le = Hcmax + Let, is generated by adding the maximum height value Hcmax of the core dimension parameter group Dc to the constant ejection distance Let from the ejector pin end protruding from the core surface. The installation length of the ejector pin within the moving template, Lem = Tm - Hep, is generated by subtracting the constant ejector pin base plate thickness Hep from the moving template thickness Tm. The protrusion length of the ejector pin protruding from the moving template surface is generated by subtracting Le from Le. A preliminary ejector pin parameter configuration group is generated based on De, Le, and Pei. The position of Pei is fine-tuned and optimized based on the spatial distance values ​​between each ejector pin position and the core protrusion parameter group Pc. The optimized ejector pin distribution position coordinate group Pe, along with De and Le, is encapsulated into an ejector pin distribution position coordinate group and an ejector pin length value group.

[0083] Step S156: Based on the guide post length value group, guide sleeve depth value group, guide post distribution position coordinate group, ejector pin distribution position coordinate group, and ejector pin length value group, perform component addition and expansion processing on the fixed mold component parameter initial group and the moving mold component parameter initial group to generate a complete mold parameter file containing the fixed mold component parameter group and the moving mold component parameter group.

[0084] Add Dg, Lg, and Pg from the complete guide post parameter configuration group generated in step S1551 to the initial group Initf of the fixed mold component parameters, generating the first extended group Fix1 of the fixed mold component parameters containing the guide post parameter configuration subgroup. Add Db, Hbm, Hbg, and Pb from the complete guide sleeve parameter configuration group generated in step S1551 to the initial group Initm of the moving mold component parameters, generating the first extended group Mov1 of the moving mold component parameters containing the guide sleeve parameter configuration subgroup. Add De, Le, and Pe from the complete ejector pin parameter configuration group generated in step S1552 to Mov1, generating the second extended group Mov2 of the moving mold component parameters containing the ejector pin parameter configuration subgroup. Calculate the range dimension value group of the ejector pin distribution area based on the maximum and minimum distribution boundary coordinate values ​​of the ejector pin position in Pe. Call the ejector pin base plate plane dimension specification classification group in the mold frame parameter component library, and select the ejector pin base plate length value Lep and ejector pin base plate width value Wep that are greater than this range dimension value group. Based on the overall structural layout of the mold, determine the ejector base plate mounting space height value Hsp. Call the ejector base plate thickness specification grading group from the mold base parameter component library and select an ejector base plate thickness value Tep that is less than Hsp. Add Lep, Wep, and Tep as ejector base plate parameter configuration subgroups to Mov2, generating the third extended group of moving mold component parameters, Mov3, which includes the ejector base plate parameter configuration subgroups. Perform interference checks based on the spatial distance between the guide post distribution position coordinate group Pg and the ejector pin distribution position coordinate group Pe. When the spatial distance between the guide post coordinates in Pg and the ejector pin coordinates in Pe is less than the safety distance threshold Ds, generate a position adjustment command. Based on this position adjustment command, perform offset correction processing on the guide post distribution position coordinates or ejector pin distribution position coordinates where interference occurs, generating a corrected guide post distribution position coordinate correction group Pgm or an ejector pin distribution position coordinate correction group Pem. Replace the original guide post distribution position coordinate group Pg in Fix1 with Pgm, generating the second extended group of fixed mold component parameters, Fix2. Replace the original ejector pin distribution coordinate group Pe in Mov3 with Pem to generate the fourth extended group of moving mold component parameters, Mov4. Package Fix2 and Mov4 into a complete mold parameter file containing both the fixed mold component parameter group and the moving mold component parameter group.

[0085] Step S210: Input the product external shape parameter set and the product internal shape parameter set into the injection molding flow simulation parametric encoder for feature compression processing to generate a potential feature vector set of injection molding filling behavior. The potential feature vector set of injection molding filling behavior includes the parameterized distribution features of melt front advance velocity and the parameterized distribution features of melt pressure attenuation gradient.

[0086] The product's external shape parameter set and internal shape parameter set are concatenated into a unified input feature tensor, which is then input into the injection molding flow simulation parametric encoder. This injection molding flow simulation parametric encoder adopts a convolutional neural network architecture, containing multiple convolutional layers and pooling layers. The input feature tensor first passes through the first convolutional layer, which contains 32 convolutional kernels, each with a size of 3×3, a stride of 1, and uniform padding, outputting the first convolutional feature map. The first convolutional feature map passes through the first pooling layer, using max pooling with a pooling window size of 2×2 and a stride of 2, outputting the first pooled feature map. The first pooled feature map then passes through the second convolutional layer (64 convolutional kernels), the second pooling layer, the third convolutional layer (128 convolutional kernels), the third pooling layer, the fourth convolutional layer (256 convolutional kernels), and the fourth pooling layer, resulting in a high-dimensional abstract feature map. This high-dimensional abstract feature map is input into a global average pooling layer, which compresses the spatial dimension of each feature channel into a single value, resulting in a global feature vector. The global feature vector is input to two independent fully connected branches: the first fully connected branch outputs a parameterized distribution feature vector Vflow of the melt front advance velocity, where each dimension corresponds to the predicted melt front advance velocity value for different regions of the product; the second fully connected branch outputs a parameterized distribution feature vector Pgrad of the melt pressure decay gradient, where each dimension corresponds to the predicted pressure decay gradient value along the flow path for different regions of the product. Vflow and Pgrad are concatenated to form a latent feature vector group Zflow for injection molding filling behavior.

[0087] Step S220: Input the cavity geometry parameter set and the core geometry parameter set into the mold structure topology parameterization encoder for structural abstraction processing to generate a mold topology constraint potential feature vector set. The mold topology constraint potential feature vector set includes the cavity and core mating clearance parameterized topology features and the demolding direction parameterized constraint features.

[0088] The geometric parameters of the cavity and core are converted into a graph structure representation. Each surface patch of the cavity and core is treated as a graph node, and the adjacency relationships between surface patches are treated as graph edges. The initial feature vector of each graph node contains geometric attributes such as the area, mean curvature, variance of curvature, and normal vector components of that surface patch. This graph structure is input into a topological parameterized encoder for the mold structure, which employs a graph convolutional neural network architecture. The encoder contains three graph convolutional layers. The first graph convolutional layer performs neighborhood aggregation on the feature vector of each node. The aggregation method is as follows: for node i, the feature vectors of all its neighboring nodes are summed, multiplied by a weight matrix W1, and then added to the result of multiplying node i's own feature vector by its own weight matrix W2. After processing by a nonlinear activation function, the first layer of node features is output. The second and third graph convolutional layers perform the same neighborhood aggregation operation sequentially, with each layer outputting feature dimensions of 64, 128, and 256, respectively. The node features output from the third convolutional layer are input into a global graph pooling layer. The feature vectors of all nodes are summed and averaged to obtain the global topological feature vector Ftopo for the mold. Ftopo inputs two independent fully connected branches: the first branch outputs a parameterized topological feature vector Tgap for the cavity-core mating clearance, which encodes the distribution pattern of the mating clearance between the cavity and core at different spatial locations; the second branch outputs a parameterized constraint feature vector Ddir for the demolding direction, which encodes the geometric constraint relationship between the main demolding direction and the auxiliary lateral core-pulling direction. Tgap and Ddir are concatenated to form the latent feature vector group Ztopo for mold topological constraints.

[0089] Step S230: Call the pre-trained injection mold parametric generative adversarial network to perform cross-domain feature joint discrimination processing on the latent feature vector group of injection filling behavior and the latent feature vector group of mold topology constraints. The injection mold parametric generative adversarial network includes a parametric generator module and a parametric discriminator module. The parametric generator module is used to convert the latent feature vector group of injection filling behavior into a mold structure correction parametric offset group. The parametric discriminator module is used to determine the compatibility discrimination probability value between the mold structure correction parametric offset group and the mold assembly parameter group.

[0090] The pre-trained parameterized generative adversarial network for injection molds is invoked. The Zflow generated in step S210 and the Ztopo generated in step S220 are concatenated along the feature dimension to obtain a joint feature vector Zjnt. This Zjnt is then input into the parameterization generator module. The parameterization generator module consists of four deconvolutional layers. Zjnt is first input into the first deconvolutional layer, which expands the dimension of the input vector to a 4×4×512 feature map. This feature map then passes through the second deconvolutional layer (output 4×4×256), the third deconvolutional layer (output 8×8×128), and the fourth deconvolutional layer (output 16×16×64). The output feature map of the fourth deconvolutional layer is input into a fully connected output layer, which outputs a one-dimensional vector, where each element corresponds to the correction offset of a parameter in the mold assembly parameter set. This one-dimensional vector is then reshaped into a mold structure correction parameterization offset set ΔM with the same structure as the mold assembly parameter set. Add ΔM to the original mold assembly parameter set Masm to obtain the corrected mold assembly parameter set Mc = Masm + ΔM. Input Mc into the parameterized discriminator module. The parameterized discriminator module consists of four convolutional layers and a fully connected output layer. Mc is first converted into a fixed-size feature map, which is then processed through four convolutional layers for feature extraction. Each convolutional layer is followed by a batch normalization layer and a LeakyReLU activation function. The output feature map of the fourth convolutional layer is then processed by global average pooling and input into the fully connected output layer. The fully connected output layer outputs a scalar value, which is mapped to the range of 0 to 1 using the Sigmoid function and serves as the compatibility discrimination probability value Pcpt.

[0091] Step S240: When the compatibility discrimination probability value is lower than the preset compatibility threshold, the updated mold structure correction parameter offset group is regenerated through the parameterization generator module, and the parting surface curve parameter group and demolding direction parameter group in the mold assembly parameter group are parameterized offset correction processed according to the updated mold structure correction parameter offset group to generate the mold assembly parameter optimization group.

[0092] When the compatibility discrimination probability value Pcpt output in step S230 is lower than the preset compatibility threshold Pth, it is determined that the currently generated mold assembly parameter set Mc does not meet the compatibility requirements between injection flow and mold structure. Zjnt is input into the parameterization generator module again, but during backpropagation, the internal weights of the generator are adjusted according to the feedback signal of the discriminator, and the updated mold structure correction parameterization offset set ΔM2 is regenerated. ΔM2 is added to the original mold assembly parameter set Masm to obtain the new Mc2=Masm+ΔM2. The updated parting surface curve parameter set Cp2 and the updated demolding direction parameter set Ddir2 are extracted from Mc2. Cp2 and Ddir2 are used as the mold assembly parameter optimization set Mopt.

[0093] Step S250: Input the optimized mold assembly parameters into the parameterized discriminator module for secondary compatibility discrimination processing. When the secondary compatibility discrimination probability value reaches the preset compatibility threshold, output the optimized mold assembly parameters as the target mold assembly parameters to the step of calling the mold frame parameter component library for component matching processing based on the optimized mold assembly parameters, replacing the original optimized mold assembly parameters.

[0094] The optimized mold assembly parameter group Mopt generated in step S240 is input again into the parameterized discriminator module, and the same discrimination calculation process as in step S230 is performed, outputting the secondary compatibility discrimination probability value Pcpt2. When Pcpt2≥Pth, it is determined that Mopt meets the compatibility requirements. Mopt is then used as the target mold assembly parameter group and output to step S150, replacing the original mold assembly parameter group Masm.

[0095] Step S310: Extract the product wall thickness parameterized distribution set and the product rib parameterized distribution set from the parameterized description group of the injection molded product model. Input the product wall thickness parameterized distribution set and the product rib parameterized distribution set into a pre-trained injection molded defect prediction graph convolutional network for node feature aggregation processing. The injection molded defect prediction graph convolutional network includes a parameterized graph convolutional layer and a parameterized attention aggregation layer. The parameterized graph convolutional layer is used to perform neighborhood information convolution processing on the wall thickness features and rib features of each node in the product geometric topology graph to generate node hidden feature groups. The parameterized attention aggregation layer is used to adaptively allocate the influence weights between different node hidden feature groups to generate a global defect-sensitive feature vector group.

[0096] The parametric distribution sets Tdis for product wall thickness and Rdis for product rib positions are extracted from the parametric description set of the injection-molded product model. The product geometry model is discretized into a graph structure, where each node corresponds to a sampling region on the product surface, and each edge connects adjacent sampling regions. The initial feature vector of each node contains the wall thickness value Tnod and the rib presence identifier Rflg. This graph structure is input into a pre-trained injection molding defect prediction graph convolutional network, which first contains three parametric graph convolutional layers. For each node i, the first parametric graph convolutional layer multiplies the feature vectors of its neighboring nodes j by the weight matrix Wc1, sums the results, and then adds the sum of the sum of the feature vectors of node i itself multiplied by the weight matrix Wc2. After passing through the ReLU activation function, the first layer node hidden feature Hi1 is output. The second and third parametric graph convolutional layers perform the same operation sequentially, outputting the node hidden feature set Hnod. Then, the hidden features of all nodes are input into a parameterized attention aggregation layer. This layer first calculates the attention weight coefficients between each pair of nodes i and j: eij = LeakyReLU(At[WattHi||WattHj]), where At is the learnable attention parameter vector, Watt is the learnable weight matrix, and || denotes vector concatenation. Softmax normalization is applied to eij to obtain the attention weights αij = exp(eij) / Σkexp(eik). The weighted feature vector Hi_att for each node i is Hi_att = Σjαij × WattHj. The summation and averaging of Hi_att values ​​from all nodes yields the global defect-sensitive feature vector set Gdfs.

[0097] Step S320: Based on the global defect sensitive feature vector group, call the injection molding defect type parameterized classifier to perform defect category probability distribution prediction processing, and generate an injection molding defect probability distribution group containing the probability values ​​of shrinkage defects, warpage defects, and cavitation defects.

[0098] The global defect-sensitive feature vector group Gdfs generated in step S310 is input into the injection molding defect type parameterized classifier. This injection molding defect type parameterized classifier is a three-layer fully connected network. Gdfs is first input into the first fully connected layer, which contains 128 neurons and outputs the first layer defect feature vector Fd1. Fd1 is input into the second fully connected layer, which contains 64 neurons and outputs the second layer defect feature vector Fd2. Fd2 is input into the output fully connected layer, which contains 3 neurons, corresponding to shrinkage defects, warpage defects, and cavitation defects, respectively. The three output values ​​of the output fully connected layer are processed by the Softmax function to obtain the probability values ​​of the three defect categories: shrinkage defect probability value Psink, warpage defect probability value Pwarp, and cavitation defect probability value Pvoid. Psink, Pwarp, and Pvoid ​​are combined into the injection molding defect probability distribution group Pdef.

[0099] Step S330: When the probability value of the shrinkage defect exceeds the preset shrinkage alarm threshold, the set of parameterized expressions of the curvature of the cavity surface in the cavity geometric parameter group is input into the shrinkage compensation parameterized inverse network for local curvature correction inversion processing. The shrinkage compensation parameterized inverse network includes a parameterized inversion convolutional layer and a parameterized curvature regression layer. The parameterized inversion convolutional layer is used to calculate the curvature correction requirement intensity parameter group of each region of the cavity surface in reverse according to the shrinkage defect probability value. The parameterized curvature regression layer is used to map the curvature correction requirement intensity parameter group into a set of displacement vectors of local control points of the cavity surface.

[0100] When the shrinkage defect probability value Psink output in step S320 exceeds the preset shrinkage alarm threshold Psth, the cavity surface curvature parameterized representation set Ccrv is extracted from the cavity geometric parameter set. This Ccrv is then input into the shrinkage compensation parameterized inverse network. This shrinkage compensation parameterized inverse network first contains three parameterized inversion convolutional layers. The first inversion convolutional layer takes Ccrv as input, performs a transposed convolution operation to double the feature map size while halving the number of channels, and outputs the first inversion feature map. The second and third inversion convolutional layers sequentially perform the same transposed convolution operation, outputting a curvature correction requirement intensity parameter set Rint. Rint is a feature map of the same size as Ccrv, where each element represents the required curvature correction intensity for the corresponding region. Rint is then input into a parameterized curvature regression layer, which contains two fully connected sublayers. The first fully connected sublayer flattens Rint into a one-dimensional vector and maps it to a 256-dimensional intermediate feature vector Fmid. The second fully connected sublayer maps Fmid to an output vector, where each element of the output vector corresponds to the displacement vector component of a control point on the cavity surface. This output vector is then reshaped into a set of displacement vectors Dp for the local control points of the cavity surface.

[0101] Step S340: Based on the displacement vector group of the local control points of the cavity surface, perform local curvature lifting correction processing on the set of cavity surface curvature parameterization expressions in the cavity geometric parameter group to generate an anti-shrinkage optimized cavity geometric parameter group, and output the anti-shrinkage optimized cavity geometric parameter group as the updated cavity geometric parameter group to the step of performing assembly gap matching processing on the cavity geometric parameter group and the core geometric parameter group.

[0102] The displacement vector set Dp of the local control points of the cavity surface generated in step S330 is applied to the cavity surface curv parameterization expression set Ccrv in the cavity geometric parameter set. For each control point Pc in Ccrv, the updated control point coordinates Pcn = Pc + Dpc, where Dpc is the displacement vector in Dp corresponding to that control point. After all control points are updated, the corrected cavity surface curvature parameterization expression set Ccrv2 is obtained. The entire cavity geometric parameter set is regenerated based on Ccrv2 to obtain the cavity geometric parameter optimization set Cav_opt after anti-shrinkage optimization. This Cav_opt is used as the updated cavity geometric parameter set and output to step S140 to replace the original cavity geometric parameter set for assembly gap matching processing.

[0103] Step S350: When the warpage defect probability value exceeds the preset warpage alarm threshold, the parameterized constraint set of the core body size in the core geometric parameter set is input into the warpage compensation parameterized inverse network for cooling channel layout parameterized offset inversion processing. The warpage compensation parameterized inverse network includes a parameterized warpage inversion convolutional layer and a parameterized channel layout regression layer. The parameterized warpage inversion convolutional layer is used to calculate the position offset requirement parameter set of each segment of the cooling channel inside the core in reverse according to the warpage defect probability value. The parameterized channel layout regression layer is used to map the position offset requirement parameter set into a cooling channel centerline control point coordinate offset vector set.

[0104] When the warpage defect probability value Pwarp output in step S320 exceeds the preset warpage alarm threshold Pwth, the parameterized constraint set Dcon for the core body size is extracted from the core geometric parameter set. This Dcon is then input into the warpage compensation parameterized inverse network. This warpage compensation parameterized inverse network contains three parameterized warpage inversion convolutional layers, whose structure is similar to the parameterized inversion convolutional layer in step S330. These layers progressively upsample the input features and extract the position offset requirement features, outputting the position offset requirement parameter set Oreq. Oreq is then input into the parameterized channel layout regression layer, which contains two fully connected sublayers. These sublayers map Oreq to a set of coordinate offset vectors Dch for the control points of the cooling channel centerline. Each element of Dch corresponds to the three-dimensional coordinate offset of a control point on the cooling channel centerline.

[0105] Step S360: Based on the coordinate offset vector group of the control point of the cooling channel centerline, perform channel centerline position offset correction processing on the preset initial cooling channel layout parameterized description group to generate an anti-warping optimized cooling channel layout parameterized description group, and add the anti-warping optimized cooling channel layout parameterized description group as an additional parameter group to the mold assembly parameter group.

[0106] Obtain the preset initial cooling channel layout parameterized description group Cchl, which records the initial three-dimensional coordinates of each control point on the centerline of the cooling channel. Apply the cooling channel centerline control point coordinate offset vector group Dch generated in step S350 to Cchl: for each control point, the updated coordinate Pcn = Pci + Dc, where Dc is the coordinate offset in Dch corresponding to that control point. After all control points are updated, the updated cooling channel centerline description is obtained. Based on this centerline, the complete cooling channel geometric parameters are regenerated to obtain the anti-warping optimized cooling channel layout parameterized description group Cch2. Add this Cch2 as an additional parameter group to the mold assembly parameter group Masm.

[0107] Step S370: When the probability value of the cavitation defect exceeds the preset cavitation alarm threshold, the parting surface curve parameter group in the mold assembly parameter group is input into the cavitation escape channel parameterized generation network for parameterized layout processing of the venting groove position. The cavitation escape channel parameterized generation network includes a parameterized parting surface curvature analysis layer and a parameterized venting groove position regression layer. The parameterized parting surface curvature analysis layer is used to calculate the cavitation accumulation risk index parameter group of each curve segment in the parting surface curve parameter group. The parameterized venting groove position regression layer is used to generate the venting groove's slotting position parameterized coordinate group and slotting depth parameterized value group on the parting surface according to the cavitation accumulation risk index parameter group.

[0108] When the cavitation defect probability value Pvoid ​​output in step S320 exceeds the preset cavitation alarm threshold Pvth, the parting surface curve parameter set Cp is extracted from the mold assembly parameter set Masm. This Cp is then input into the cavitation escape channel parameterization generation network. The parameterized parting surface curvature analysis layer of this cavitation escape channel parameterization generation network first calculates the curvature value κp at each sampling point on Cp. Then, it calculates the cavitation accumulation risk index Rv = κp × Vfr at each sampling point, where Vfr is the melt front velocity at that location obtained from injection flow simulation. The Rv values ​​of all sampling points constitute the cavitation accumulation risk index parameter set Rvs. Rvs is input into the parameterized venting groove location regression layer, which contains two fully connected sublayers. The first fully connected sublayer maps Rvs to an intermediate feature vector Fmid2, and the second fully connected sublayer maps Fmid2 to an output vector. The output vector consists of two parts: the first part is the parameterized coordinate group Pvt of the exhaust groove position, where each element is a two-dimensional coordinate on the parting surface; the second part is the parameterized numerical group Dvt of the exhaust groove depth, where each element is the groove depth value at the corresponding position.

[0109] Step S380: Based on the parameterized coordinate group of the slotting position and the parameterized numerical group of the slotting depth, the parting surface curve parameter group in the mold assembly parameter group is processed to add venting groove features, generating an optimized parting surface curve parameter group with venting groove structure, and outputting the optimized parting surface curve parameter group with venting groove structure as the updated parting surface curve parameter group to the step of calling the mold frame parameter component library for component matching based on the mold assembly parameter group.

[0110] Based on the parameterized coordinate group Pvt for the slot position and the parameterized numerical group Dvt for the slot depth generated in step S370, the parting surface curve parameter group Cp is processed to add venting groove features. For each slot position coordinate in Pvt, an venting groove geometric feature is generated at the corresponding position in Cp, and the depth of the venting groove geometric feature is determined by the corresponding depth value in Dvt. After all venting groove features are added, an optimized parting surface curve parameter group Cp3 containing the venting groove structure is generated. This Cp3 is used as the updated parting surface curve parameter group and output to step S150, replacing the original parting surface curve parameter group.

[0111] Step S410: Input the cavity geometry parameter set and the core geometry parameter set into the mold structure topology parameterization encoder for structural abstraction processing to generate a mold topology constraint potential feature vector set. The mold topology constraint potential feature vector set includes the cavity and core mating clearance parameterized topology features and the demolding direction parameterized constraint features.

[0112] The cavity depth dimension chain and cavity sidewall draft angle parameter set are extracted from the cavity geometry parameter set. The core dimension parameter set and core side surface draft angle distribution set are extracted from the core geometry parameter set. These parameter sets are combined into a unified input feature tensor Fin. This Fin is input into the mold structure topology parameterized encoder, which adopts a multilayer perceptron architecture. Fin first passes through the first fully connected layer, which contains 512 neurons and outputs the first layer abstract feature vector F1. F1 is input into the second fully connected layer, which contains 256 neurons and outputs the second layer abstract feature vector F2. F2 is input into the third fully connected layer, which contains 128 neurons and outputs the third layer abstract feature vector F3. F3 inputs two independent fully connected layers: the first fully connected layer outputs a parameterized topological feature vector Tgap for the cavity-core mating clearance, where each dimension corresponds to the theoretical clearance prediction value of the cavity and core in different mating regions; the second fully connected layer outputs a parameterized constraint feature vector Ddir for the demolding direction, where each dimension corresponds to the component of the demolding direction vector in three-dimensional space and the direction cosine value of the lateral core-pulling direction vector. Tgap and Ddir are concatenated to form the mold topological constraint latent feature vector group Ztopo.

[0113] Step S420: Call the pre-trained injection mold parametric generative adversarial network to perform cross-domain joint feature discrimination processing on the latent feature vector group of injection filling behavior and the latent feature vector group of mold topology constraints. The injection mold parametric generative adversarial network includes a parametric generator module and a parametric discriminator module. The parametric generator module is used to convert the latent feature vector group of injection filling behavior into a mold structure correction parametric offset group. The parametric discriminator module is used to determine the compatibility discrimination probability value between the mold structure correction parametric offset group and the mold assembly parameter group.

[0114] The pre-trained parameterized generative adversarial network for injection molds is invoked. The Zflow generated in step S210 and the Ztopo generated in step S410 are concatenated along the feature dimension to obtain a joint feature vector Zjoint. This Zjoint is then input into the parameterized generator module. The parameterized generator module consists of four fully connected layers. Zjoint is first input into the first fully connected layer of the generator, which contains 256 neurons and outputs the first layer generated feature vector G1. G1 is input into the second fully connected layer of the generator, which contains 128 neurons and outputs the second layer generated feature vector G2. G2 is input into the third fully connected layer of the generator, which contains 64 neurons and outputs the third layer generated feature vector G3. G3 is input into the generator output fully connected layer, which outputs a one-dimensional vector whose dimension is the same as the unfolded dimension of the mold assembly parameter group Masm, and each element corresponds to the correction offset of a certain parameter in Masm. This one-dimensional vector is then reshaped into a mold structure correction parameterized offset group ΔM consistent with the Masm structure. ΔM is element-wise added to the original mold assembly parameter set Masm to obtain the corrected mold assembly parameter set Mc = Masm + ΔM. Mc is then input into the parameterized discriminator module. The discriminator module also consists of four fully connected layers. Mc is first flattened into a one-dimensional vector and input into the first fully connected layer of the discriminator, which contains 64 neurons and outputs the first layer discriminant feature vector D1. D1 is input into the second fully connected layer of the discriminator, which contains 32 neurons and outputs the second layer discriminant feature vector D2. D2 is input into the third fully connected layer of the discriminator, which contains 16 neurons and outputs the third layer discriminant feature vector D3. D3 is input into the discriminator output fully connected layer, which contains 1 neuron and outputs a scalar value. This scalar value is activated by the Sigmoid function and mapped to the interval [0,1] to obtain the compatibility discrimination probability value Pcpt.

[0115] Step S430: When the compatibility discrimination probability value is lower than the preset compatibility threshold, the updated mold structure correction parameter offset group is regenerated through the parameter generator module, and the parting surface curve parameter group and demolding direction parameter group in the mold assembly parameter group are parameterized offset correction processed according to the updated mold structure correction parameter offset group to generate the mold assembly parameter optimization group.

[0116] When the compatibility discrimination probability value Pcpt output in step S420 is lower than the preset compatibility threshold Pth, parameter optimization iteration is performed. The joint feature vector Zjoint is input into the parameterization generator module again. At the same time, the gradient signal output by the discriminator module is passed to the generator module through the backpropagation algorithm to update the weight parameters of the generator module. The updated generator module performs forward calculation on Zjoint to regenerate the updated mold structure correction parameterized offset set ΔM_new. The offset quantum vector ΔCp corresponding to the parting surface curve parameter set Cp and the offset quantum vector ΔDdir corresponding to the demolding direction parameter set Ddir are extracted from ΔM_new. The original parameters are corrected: Cp_new=Cp+ΔCp, Ddir_new=Ddir+ΔDdir. The corrected Cp_new and Ddir_new are repackaged to generate the mold assembly parameter optimization set Mopt_new=(Cp_new,Ddir_new).

[0117] Step S440: Input the optimized mold assembly parameters into the parameterized discriminator module for secondary compatibility discrimination processing. When the secondary compatibility discrimination probability value reaches the preset compatibility threshold, output the optimized mold assembly parameters as the target mold assembly parameters to the step of calling the mold frame parameter component library for component matching processing based on the optimized mold assembly parameters, replacing the original optimized mold assembly parameters.

[0118] The mold assembly parameter optimization group Mopt_new generated in step S430 is input into the parameterized discriminator module, and the same discrimination calculation process as in step S420 is performed to obtain the secondary compatibility discrimination probability value Pcpt2. It is then determined whether Pcpt2 is greater than or equal to the preset compatibility threshold Pth. If the condition is met, Mopt_new is determined as the target mold assembly parameter group Masm_final. This Masm_final is output and passed to step S150 to replace the original mold assembly parameter group Masm in the original process, serving as the input for calling the mold base parameter component library for component matching processing.

[0119] For example, the method may further include: step S510: inputting the product shape parameter set into the trained cavity generation adversarial network for cavity structure generation processing, wherein the cavity generation adversarial network includes a cavity shape generator module and a cavity shape discriminator module.

[0120] Key features are extracted from the product shape parameter set received in step S110, including the bounding box dimensions of the product's outer boundary curve group and the statistical features (mean, variance) of the product's surface curvature gradient group, and combined into a conditional feature vector C_cond. Simultaneously, a random noise vector Z_noise is sampled from a standard normal distribution. The conditional feature vector C_cond and the random noise vector Z_noise are concatenated to obtain the generator's input vector Z_input. Z_input is then input into the cavity shape generator module of the trained cavity generative adversarial network.

[0121] Step S520: Receive the random noise vector and the conditional constraint parameters of the product shape parameter set through the cavity shape generator module, and perform multi-layer deconvolution operation on the random noise vector and the conditional constraint parameters to generate an initial cavity geometric parameter set.

[0122] The cavity shape generator module receives the input vector Z_input generated in step S510. This cavity shape generator module consists of four fully connected layers. Z_input is first input to the first fully connected layer (containing 256 neurons), and outputs a feature vector G_fc1. G_fc1 is input to the second fully connected layer (containing 512 neurons), and outputs a feature vector G_fc2. G_fc2 is input to the third fully connected layer (containing 1024 neurons), and outputs a feature vector G_fc3. G_fc3 is input to the generator output fully connected layer, whose number of neurons is the same as the parameterized representation dimension of the cavity geometry parameter set. The activation values ​​of the output layer are reshaped to generate the initial cavity geometry parameter set G_init. The parameter structure of G_init includes preliminary estimates of cavity size parameters (such as depth chain, volume, and area) and cavity curvature parameters (such as coordinates of surface control points).

[0123] Step S530: Input the initial cavity geometric parameter set and the cavity geometric parameter set output by the cavity generation rule library into the cavity shape discriminator module for authenticity discrimination processing. The cavity shape discriminator module outputs authenticity scoring parameter set and structural rationality scoring parameter set.

[0124] The rule-driven cavity geometric parameter set, denoted as G_rule, is obtained from the output of step S120 (cavity generation rule base processing flow) and used as the "real" sample. The initial cavity geometric parameter set G_init generated in step S520 is used as the "generated" sample. G_rule and G_init are respectively input into the cavity shape discriminator module. The discriminator module first extracts features through several fully connected layers. For the input sample X (which may be G_rule or G_init), it first passes through the first discriminative fully connected layer (128 neurons), outputting D_fc1. D_fc1 passes through the second discriminative fully connected layer (64 neurons), outputting D_fc2. D_fc2 is input to two parallel output heads: the first output head (realism discriminator head) is a single-neuron fully connected layer, followed by a sigmoid function, which outputs a scalar value P_real, representing the probability that the sample is judged to be from the real data distribution, i.e., the realism score. The second output head (structural plausibility discriminator) is a fully connected layer containing multiple neurons. Its output vector is processed by a custom plausibility evaluation function (e.g., checking if the depth is positive, if the curvature is continuous, etc.) and mapped to a scalar value S_valid, representing the structural plausibility score of the sample. Therefore, for each input sample, the discriminator outputs a tuple (P_real, S_valid).

[0125] Step S540: Calculate the generator adversarial loss value set and the discriminator adversarial loss value set according to the authenticity scoring parameter set and the structural rationality scoring parameter set. Perform backpropagation update processing on the network weight parameters of the cavity shape generator module according to the generator adversarial loss value set to obtain the updated cavity shape generator module. Perform backpropagation update processing on the network weight parameters of the cavity shape discriminator module according to the discriminator adversarial loss value set to obtain the updated cavity shape discriminator module.

[0126] In one training iteration, the generator loss L_G and the discriminator loss L_D are calculated respectively.

[0127] The generator loss L_G consists of two parts: adversarial loss and rationality loss. The adversarial loss L_adv_G = -log(D(G(Z_input))), where D(G(Z_input)) is the discriminator's realism score P_real for the generated sample G_init. The rationality loss L_valid_G = max(0, S_threshold - S_valid_G), where S_valid_G is the structural rationality score of the generated sample, and S_threshold is a preset rationality threshold. The total generator loss is L_G = L_adv_G + λ. L_valid_G, where λ is the balance coefficient.

[0128] The discriminator loss L_D also consists of two parts: the real sample loss and the generated sample loss. The real sample loss L_real = -log(D(G_rule)), and the generated sample loss L_fake = -log(1-D(G_init)). The total discriminator loss is L_D = L_real + L_fake.

[0129] Using the backpropagation algorithm and an optimizer (such as Adam), the gradient is calculated based on L_G, and all trainable weight parameters of the cavity shape generator module are updated to obtain the weight-updated generator module. Similarly, the gradient is calculated based on L_D, and all trainable weight parameters of the cavity shape discriminator module are updated to obtain the weight-updated discriminator module.

[0130] Step S550: Iteratively execute the generative adversarial training process until the authenticity score parameter set exceeds the preset authenticity threshold and the structural rationality score parameter set exceeds the preset rationality threshold, thereby obtaining the trained cavity generative adversarial network. Call the cavity shape generator module in the trained cavity generative adversarial network to perform forward propagation calculation on the product shape parameter set to generate an optimized cavity geometric parameter set. Use the optimized cavity geometric parameter set as the replacement data of the cavity geometric parameter set in the subsequent assembly gap matching process.

[0131] Repeat steps S510 to S540 for multiple rounds of iterative training. After each round of training, evaluate the performance of the current generator module using an independent validation set. Calculate the average realism score P_real_avg and the average structural plausibility score S_valid_avg of the generated samples on the validation set. Stop training when P_real_avg exceeds a preset realism threshold P_th_true (e.g., 0.7) and S_valid_avg exceeds a preset plausibility threshold S_th_valid (e.g., 0.8). At this point, the saved cavity shape generator module becomes the core generation component of the trained cavity generative adversarial network. During the inference phase, this trained generator module is invoked. The product shape parameter set of the current injection molded product model (after the same feature extraction as during training) is used as conditional constraint parameters and input into the generator along with a fixed or randomly sampled noise vector. The generator performs forward propagation calculations and outputs the optimized cavity geometry parameter set G_optimized. The cavity geometry parameter set directly generated by the cavity generation rule base in step S120 is completely replaced by G_optimized, and it is passed as input to step S140 (assembly gap matching processing) and its subsequent steps.

[0132] For example, the method may further include: step S610: inputting the mold assembly parameter set into the trained injection flow graph neural network for filling balance analysis processing, wherein the injection flow graph neural network includes an assembly structure diagram encoder module and a flow behavior decoder module.

[0133] The parting surface curve parameter set Cp and the demolding direction parameter set Ddir are extracted from the mold assembly parameter set Masm. A discrete mesh of the region on the parting surface plane is constructed based on Cp, with each mesh cell considered a graph node. The main demolding direction vector in the demolding direction parameter set Ddir is used to determine the mainstream direction of melt flow. The mold assembly parameter set Masm is then input into the trained injection flow graph neural network. The assembly structure graph encoder module of this injection flow graph neural network receives the graph structure data composed of the aforementioned nodes and edges, where the initial features of each node include its spatial coordinates and the type of region it belongs to (e.g., gate area, end area, sidewall area, etc.).

[0134] Step S620: The assembly structure diagram encoder module performs graph structure construction processing on the parting surface curve parameter group and the demolding direction parameter group to generate a mold assembly topology diagram, which includes cavity nodes, core nodes, parting surface edge nodes and demolding direction edge nodes.

[0135] The assembly structure diagram encoder module executes the diagram construction. First, based on the cavity geometry parameter set, a series of nodes representing different regions of the cavity are generated within the parting surface projection area, labeled as the cavity node set V_cav. Based on the core geometry parameter set, nodes representing different regions of the core are generated, labeled as the core node set V_cor. The parting surface itself is abstracted as a set of nodes located on the parting surface, labeled as the parting surface edge node set V_par. These nodes describe the contour and key internal positions of the parting surface. The demolding direction parameters are transformed into connection relationships: starting from each cavity node and core node, along the demolding direction vector, connection edges are established with the parting surface edge nodes or adjacent region nodes. These edges imply the directional constraints of the flow path and can be regarded as the demolding direction edge node (or edge with directional weight) set E_dir. Finally, all nodes and edges are combined to generate the mold assembly topology diagram G_ass={V_cav∪V_cor∪V_par,E_conn∪E_dir}, where E_conn is the edge representing the spatial adjacency relationship of the regions.

[0136] Step S630: Extract local geometric feature vector groups for each node in the mold assembly topology graph, input the local geometric feature vector groups into the multi-layer graph convolutional layer of the assembly structure graph encoder module for neighborhood information aggregation processing, and generate a hidden layer feature representation group for each node.

[0137] For each node v_i in the mold assembly topology graph G_ass, its local geometric feature vector f_i is extracted. Features f_i may include: the node's 3D coordinates, its part type (one-hot encoded), the average curvature of adjacent surfaces, the distance to the nearest boundary, and the theoretical wall thickness at the node. The graph containing the initial features {f_i} of all nodes is input into the assembly structure graph encoder module, which contains K layers of graph convolutional layers. The update formula for node v_i in the k-th layer graph convolution operation is: h_i^(k)=σ(W^(k)AGGREGATE({h_j^(k-1):j∈N(i)})+B^(k)h_i^(k-1)), where h_i^(k) is the feature of node v_i in the k-th layer, h_i^(0)=f_i, N(i) is the set of neighboring nodes of node v_i, AGGREGATE is an aggregation function (such as mean, summation or maximum), W^(k) and B^(k) are learnable weight matrices, and σ is a non-linear activation function. After K layers of graph convolution, the final hidden feature representation h_i^(K) of each node is obtained.

[0138] Step S640: Input the hidden layer feature representation group of all nodes into the global pooling layer of the assembly structure diagram encoder module for feature compression processing to generate a global feature vector of mold assembly. Input the global feature vector of mold assembly into the flow behavior decoder module for layer-by-layer inverse graph convolution processing to generate a set of predicted melt flow front position parameters and a set of predicted melt flow velocity vectors for each region inside the cavity. Calculate the time difference parameter group for each flow path to reach the end of the cavity based on the set of predicted melt flow front position parameters and the set of predicted melt flow velocity vectors. Identify the position coordinate group of the unbalanced filling region and the quantification parameter group of the degree of imbalance based on the time difference parameter group.

[0139] The hidden feature representations {h_i^(K)} of all nodes obtained in step S630 are input into a global pooling layer (such as global average pooling or global max pooling) to obtain a fixed-dimensional global feature vector g_global for mold assembly. This global feature vector g_global is then input into the flow behavior decoder module. The decoder module typically uses a multilayer perceptron or inverse graph convolutional structure, and its goal is to decode the global features into the predicted flow behavior value for each graph node (or each spatial region). The decoder outputs two vectors that correspond one-to-one with each node: 1) the predicted arrival time parameter T_i of the melt flow front; 2) the predicted average flow velocity vector V_i of the melt at the node. Based on the predicted arrival time T_i, the flow path is traced from the set gate node (departure time T=0) to each end node (cavity boundary). The maximum time difference Δt_max = max(T_end) - min(T_end) for different flow paths to their respective end nodes is calculated. Simultaneously, identify the terminal nodes and their upstream regions that arrive significantly later than the average time, and record the node coordinates of these regions as the location coordinate set L_unbal to fill the imbalanced region. The imbalance quantification parameter set Q_unbal can include: the maximum time difference Δt_max, the time delay of each imbalanced region, and the area percentage of the affected region.

[0140] Step S650: Call the filling optimization suggestion generation module of the injection flow diagram neural network to process the coordinate group of the filling imbalance area and the quantification parameter group of the imbalance degree, generate a filling optimization parameter group including the gate position adjustment vector group and the runner size correction parameter group, and feed the filling optimization parameter group back to the mold assembly parameter group for parameter correction processing to generate an optimized mold assembly parameter group.

[0141] The injection flow graph neural network incorporates a fill optimization suggestion generation module, which receives the coordinate set L_unbal of the unbalanced filling region and the unbalance degree quantification parameter set Q_unbal as input. Internally, the module generates optimization suggestions based on predefined optimization rules or a small strategy network. These suggestions typically include: 1) A gate location adjustment vector set ΔG: Calculates a two-dimensional or three-dimensional position offset vector for each existing or candidate gate location to shorten the flow path in areas far from the gate. 2) A runner size correction parameter set ΔR: Includes adjustments to the diameters of the main runner and branch runners to balance the flow resistance of each path. The generated fill optimization parameter set Opt_fill={ΔG,ΔR} is fed back to the original mold assembly parameter set Masm. Based on ΔG, the geometric descriptions or positional parameters related to the gate in the mold assembly parameters are adjusted. Based on ΔR, the dimensional parameters related to the runner system in the mold assembly parameters are adjusted. After these corrections, the optimized mold assembly parameter set Masm_optimized_fill is generated. This optimized parameter set can be used for a more balanced injection filling process.

[0142] Figure 2 This application illustrates a rapid injection mold structure generation system 100 based on parametric 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 rapid injection mold structure generation system 100 may further include a transceiver 1004, which can be used for data interaction between this rapid injection mold structure generation system and other rapid injection mold structure generation systems based on parametric 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 rapid injection mold structure generation system 100 based on parametric design does not constitute a limitation on the embodiments of this application.

[0143] 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.

[0144] 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 rapid generation of injection mold structures based on parametric design, characterized in that, The method includes: Receive a parametric description set of the injection molded product model, the parametric description set including a product external shape parameter set and a product internal shape parameter set; The cavity generation rule library is called according to the product shape parameter set to perform matching and mapping processing to obtain the cavity geometric parameter set, which includes the cavity size parameter set and the cavity curvature parameter set; The core generation rule library is called according to the product internal shape parameter set to perform matching and mapping processing to obtain the core geometric parameter set, which includes the core size parameter set and the core protrusion parameter set; The cavity geometry parameter set and the core geometry parameter set are subjected to assembly clearance matching processing to generate a mold assembly parameter set, which includes a parting surface curve parameter set and a demolding direction parameter set. The mold assembly parameter group is used to call the mold frame parameter component library for component matching and generate a complete mold parameter file. The complete mold parameter file includes a fixed mold component parameter group and a moving mold component parameter group.

2. The rapid generation method for injection mold structure based on parametric design according to claim 1, characterized in that, The step of calling the cavity generation rule base according to the product shape parameter set to perform matching and mapping processing to obtain the cavity geometric parameter set includes: The product's outer boundary curve set in the product shape parameter set is analyzed, and the projected contour polygon set of the cavity on the parting surface is determined based on the product's outer boundary curve set; Extract the product surface curvature gradient group from the product shape parameter set, and input the product surface curvature gradient group into the cavity surface shrinkage compensation rule layer for processing to obtain the shrinkage compensation curvature parameter group. A cavity initial skeleton wireframe group is generated based on the projected contour polygon group and the shrinkage compensation curvature parameter group. The cavity initial skeleton wireframe group includes a cavity depth dimension chain group and a cavity bottom fillet radius group. The draft angle generation rule for the cavity is called to add draft angle to the initial skeleton wireframe group of the cavity, generating a cavity sidewall draft angle parameter group, which contains draft angle value groups for different depth segments. The boundary curve group of the surface patch in the shrinkage compensation curvature parameter group and the cavity sidewall angle parameter group are subjected to boundary fusion processing to obtain the cavity surface patch topology group, which includes the continuity order identifier group of adjacent surface patches. The cavity surface topology group is subjected to surface smoothing filtering to generate a cavity surface control point offset group, which is used to adjust the local unevenness of the cavity surface. A cavity solid boundary structure group is generated based on the cavity depth dimension chain group and the cavity surface control point offset group. The cavity solid boundary structure group includes the surface normal group and the edge curvature group of the cavity space. Extract the cavity volume expression group and cavity area expression group from the cavity entity boundary structure group. Use the cavity volume expression group and cavity area expression group as the volume parameter subset and area parameter subset in the cavity size parameter group, respectively. Generate a cavity geometric parameter group based on the cavity size parameter group containing the volume parameter subset and area parameter subset and the cavity surface control point offset group. The cavity geometric parameter group includes the cavity size parameter group and the cavity curvature parameter group.

3. The method for rapid generation of injection mold structures based on parametric design according to claim 2, characterized in that, The step of analyzing the product's outer boundary curve set in the product shape parameter set, and determining the projected contour polygon set of the cavity on the parting surface based on the product's outer boundary curve set, includes: Extract the spatial boundary curve segment group of the product shape from the product shape parameter set. The spatial boundary curve segment group includes the starting coordinate group, the ending coordinate group, and the curve type identifier group of each curve segment. The spatial boundary curve segment group is subjected to closed loop search processing. According to the endpoint coordinate matching principle, the curve segments in the spatial boundary curve segment group are connected sequentially into multiple closed boundary loop groups. Each closed boundary loop corresponds to a closed boundary loop of the product shape. From the multiple closed boundary ring groups, the boundary ring containing the projection point of the coordinate origin and having the largest enclosed area is selected as the product outer boundary curve group. The product outer boundary curve group contains all the curve segment parameter groups that constitute the maximum outer boundary. The starting coordinate group and ending coordinate group of each curve segment in the outer boundary curve group of the product are projected along the normal direction of the parting surface onto the plane where the parting surface is located, generating the projected starting coordinate projection group and ending coordinate projection group. Based on the projected starting coordinate projection group and ending coordinate projection group, the projected curve segment group on the projection plane is reconstructed, maintaining the consistency of the curve type identification of each curve segment before and after projection. The projection curve segment group is corrected by aligning the coordinates of the first and last connection points of the curve segments after projection. This process eliminates the slight deviation of the endpoint coordinates of the curve segments caused by differences in numerical precision during the projection process, resulting in a closed and continuous projection boundary curve correction group. Based on the parameter expression of each curve segment in the projection boundary curve correction group, the inclusion of points inside the projection area is judged to determine the range of the internal area enclosed by the projection boundary and generate a projection contour polygon vertex sequence group. The vertex coordinates in the vertex sequence group of the projected contour polygon are sorted in a counterclockwise direction to generate a projected contour polygon group with a unified direction identifier. The projected contour polygon group is used for parameterized reference of the boundary of the cavity projection area in the subsequent process.

4. The method for rapid generation of injection mold structures based on parametric design according to claim 2, characterized in that, The step of generating the initial skeleton wireframe group of the cavity based on the projected contour polygon group and the shrinkage compensation curvature parameter group includes: The projection region boundary mesh framework on the fractal surface is constructed based on the vertex sequence group in the projection contour polygon group. The projection region boundary mesh framework includes the direction vector group of the boundary edge and the coordinate position group of the boundary vertex. Extract the coordinates of the highest and lowest points of the product surface from the shrinkage compensation curvature parameter set. Calculate the vertical distance from each coordinate point in the highest point coordinate set to the parting surface, obtaining the highest point vertical distance value set as the first dimensional constraint in the cavity depth direction. Calculate the vertical distance from each coordinate point in the lowest point coordinate set to the parting surface, obtaining the lowest point vertical distance value set as the second dimensional constraint in the cavity depth direction. Generate a cavity depth dimensional chain based on the highest and lowest point vertical distance value sets. A two-dimensional parametric coordinate system is established on the parting surface based on the boundary vertex coordinate position set in the boundary grid frame of the projection area, and the boundary vertex coordinate position set is converted into a two-dimensional parametric coordinate set. Using the coordinates of each vertex in the two-dimensional parameter coordinate group as the two-dimensional coordinate reference of the cavity bottom surface grid node, and combining the maximum depth value of the corresponding position in the cavity depth dimension chain group, a three-dimensional spatial coordinate group of each grid node on the cavity bottom surface is generated. The initial mesh surface of the cavity bottom surface is constructed based on the three-dimensional spatial coordinates of each mesh node on the cavity bottom surface. The initial mesh surface of the cavity bottom surface includes the topological connection relationship group of mesh patches and the initial normal group of mesh nodes. Starting from the boundary grid nodes of the initial grid surface of the cavity bottom, the initial grid surface of the cavity sidewall is generated by extending in the opposite direction along the parting surface normal. The initial grid surface of the cavity sidewall includes a set of vertical extension direction vectors of the sidewall grid nodes. The initial mesh surface of the cavity bottom surface and the initial mesh surface of the cavity sidewall are subjected to boundary stitching to generate an initial skeleton wireframe group of the cavity. The initial skeleton wireframe group of the cavity includes a bottom surface node coordinate group, a sidewall node coordinate group, and a bottom surface and sidewall boundary curve group.

5. The method for rapid generation of injection mold structures based on parametric design according to claim 1, characterized in that, The process of calling the core generation rule library based on the product's internal shape parameter set to perform matching and mapping processing, resulting in a core geometric parameter set, includes: The product internal cavity boundary surface group is analyzed in the product internal shape parameter set. The outer contour surface group of the core is determined based on the product internal cavity boundary surface group. The outer contour surface group of the core and the product internal cavity boundary surface group have a spatial complementary filling relationship. Extract the coordinate set of the product's internal protrusion position and the set of the product's internal protrusion height dimensions from the product's internal shape parameter set. Generate the core surface pit position distribution set and the core surface pit depth dimension set based on the product's internal protrusion position coordinate set and the product's internal protrusion height dimension set. Based on the outer contour surface group of the core and the distribution group of the pit positions on the core surface, feature spatial position fusion processing is performed to generate the basic shape boundary group of the core body. The basic shape boundary group of the core body includes the outer surface sheet group of the core body and the coordinate group of the volume center point of the core body. The core body basic shape boundary group is processed by calling the core demolding angle generation rule to add demolding angles, generating a core side surface demolding angle distribution group, which includes demolding angle value groups corresponding to different height segments. The core body root fillet radius distribution group is generated based on the outer surface sheet group in the core body basic shape boundary group. The core body root fillet radius distribution group includes the fillet radius value group of the connection part between the core body and the core mounting plate. The core body basic shape boundary group and the core surface pit depth dimension group are subjected to pit bottom sidewall transition fillet processing to generate pit feature transition fillet radius group. The pit feature transition fillet radius group includes the fillet radius value group of each pit feature bottom corner position and sidewall edge position. Based on the distribution group of the demolding draft angle on the side surface of the core and the distribution group of the fillet radius at the root of the core body, the basic shape boundary group of the core body is modified to generate the modified boundary group of the core body. Based on the set of transition fillet radii for the pit features, the geometry of the pits corresponding to the pit location distribution set on the core surface is locally rounded to generate a complete geometric description set of the core pits containing the transition fillet surface. After the core body is corrected, the boundary group is processed by Boolean difference set operation with the complete geometric description group of the core recess to generate the core geometric parameter group. The core geometric parameter group includes the core size parameter group and the core protrusion parameter group.

6. The method for rapid generation of injection mold structures based on parametric design according to claim 1, characterized in that, The step of performing assembly clearance matching processing on the cavity geometry parameter set and the core geometry parameter set to generate a mold assembly parameter set includes: Extract the cavity opening boundary curve group from the cavity geometry parameter group, and simultaneously extract the core maximum outer boundary curve group from the core geometry parameter group; The overlap ratio of the cavity opening boundary curve group and the core maximum outer boundary curve group is compared in three-dimensional space, and the average spatial distance and maximum spatial distance between the two curve groups are calculated. A cavity-core mating clearance distribution group is generated based on the average spatial distance value group and the maximum spatial distance value group, wherein the cavity-core mating clearance distribution group includes clearance width value groups for different mating areas; The spatial plane position equation of the parting surface is determined based on the gap width values ​​of the uniformly distributed area in the cavity and core fit gap distribution group, and the parting surface curve parameter group is generated. The parting surface curve parameter group includes the intersection parameter group of the parting surface and the cavity opening boundary curve group. The central axis direction vector of the cavity sidewall draft angle parameter group in the cavity geometry parameter group is extracted as the first candidate demolding direction vector. At the same time, the central axis direction vector of the core side surface draft angle distribution group in the core geometry parameter group is extracted as the second candidate demolding direction vector. The first candidate demolding direction vector and the second candidate demolding direction vector are compared for directional consistency. The included angle value between the two direction vectors is calculated. When the included angle value is less than the preset directional deviation threshold, a unified demolding direction vector is generated. Based on the unified demolding direction vector, the demolding feasibility verification process is performed on the cavity geometric parameter group and the core geometric parameter group to generate the cavity internal undercut area position group and the core external undercut area position group. Based on the cavity internal undercut area position group and the core external undercut area position group, the lateral core pulling mechanism configuration requirement group is generated. The lateral core pulling mechanism configuration requirement group includes the lateral core pulling direction vector group and the lateral core pulling distance value group. A set of demolding direction parameters is generated based on the unified demolding direction vector and the set of lateral core-pulling direction vectors. The set of demolding direction parameters includes a main demolding direction vector and an auxiliary lateral core-pulling direction vector set. A mold assembly parameter set is generated based on the parting surface curve parameter set and the demolding direction parameter set, wherein the mold assembly parameter set includes the parting surface curve parameter set and the demolding direction parameter set.

7. The method for rapid generation of injection mold structures based on parametric design according to claim 1, characterized in that, The step of calling the mold frame parameter component library according to the mold assembly parameter group to perform component matching processing and generate a complete mold parameter file includes: The spatial height position value of the parting surface in the mold thickness direction is determined according to the parting surface curve parameter group in the mold assembly parameter group, and the spatial height position value is used as the boundary height reference value between the fixed mold assembly and the moving mold assembly. The first dimensional constraint in the thickness direction of the fixed template is calculated based on the maximum depth value of the cavity depth dimension chain group in the cavity geometry parameter group and the boundary height reference value. The standard specification parameter group of the fixed template with a thickness dimension greater than the first dimensional constraint in the thickness direction of the fixed template is then called from the mold frame parameter component library. The first dimensional constraint in the thickness direction of the moving template is calculated based on the maximum height value of the core dimension parameter group in the core geometry parameter group and the boundary height reference value. The standard specification parameter group of the moving template with a thickness dimension greater than the first dimensional constraint in the thickness direction of the moving template is then called from the mold frame parameter component library. Extract the minimum bounding rectangle size value group of the projection contour polygon group of the cavity space on the parting surface from the cavity geometry parameter group as the width and length constraints of the cavity layout area. Call the fixed template plane size standard specification group and the moving template plane size standard specification group from the mold frame parameter component library whose template plane size is greater than the width and length constraints of the cavity layout area. An initial set of parameters for the fixed mold component is generated based on the first dimensional constraint in the thickness direction of the fixed mold and the standard specification group of the planar dimensions of the fixed mold. The initial set of parameters for the fixed mold component includes the values ​​of the fixed mold thickness, the fixed mold length, and the fixed mold width. An initial set of parameters for the moving mold component is generated based on the first dimensional constraint in the thickness direction of the moving mold and the standard specification group of the planar dimensions of the moving mold. The initial set of parameters for the moving mold component includes the values ​​of the moving mold thickness, the moving mold length, and the moving mold width. The system calls the guide post and guide sleeve configuration rules in the mold frame parameter component library to generate guide post length value groups and guide sleeve depth value groups based on the fixed template thickness value and the moving template thickness value, and generates guide post distribution position coordinate groups based on the fixed template length value and width value; the system also calls the ejector mechanism configuration rules in the mold frame parameter component library to generate ejector pin distribution position coordinate groups and ejector pin length value groups based on the moving template plane dimension standard specification group and the core protrusion parameter group. Based on the guide post length value group, guide sleeve depth value group, guide post distribution position coordinate group, ejector pin distribution position coordinate group, and ejector pin length value group, the initial parameter group of the fixed mold component and the initial parameter group of the moving mold component are expanded by adding components to generate a complete mold parameter file containing the parameter groups of the fixed mold component and the moving mold component.

8. The method for rapid generation of injection mold structures based on parametric design according to claim 1, characterized in that, The method further includes: The product external shape parameter set and the product internal shape parameter set are input into the injection flow simulation parametric encoder for feature compression processing to generate a potential feature vector set of injection filling behavior. The potential feature vector set of injection filling behavior includes the parameterized distribution features of melt front advance velocity and the parameterized distribution features of melt pressure attenuation gradient. The cavity geometric parameter set and the core geometric parameter set are input into the mold structure topology parameterization encoder for structural abstraction processing to generate a mold topology constraint potential feature vector set. The mold topology constraint potential feature vector set includes the cavity and core mating clearance parameterized topology features and the demolding direction parameterized constraint features. A pre-trained parametric generative adversarial network for injection molds is invoked to perform cross-domain joint feature discrimination processing on the latent feature vector group of injection filling behavior and the latent feature vector group of mold topology constraints. The parametric generative adversarial network for injection molds includes a parametric generator module and a parametric discriminator module. The parametric generator module is used to convert the latent feature vector group of injection filling behavior into a set of parameterized offsets for mold structure correction. The parametric discriminator module is used to determine the compatibility discrimination probability value between the set of parameterized offsets for mold structure correction and the set of mold assembly parameters. When the compatibility discrimination probability value is lower than the preset compatibility threshold, the updated mold structure correction parameter offset group is regenerated by the parameter generator module, and the parting surface curve parameter group and demolding direction parameter group in the mold assembly parameter group are parameterized offset correction processed according to the updated mold structure correction parameter offset group to generate the mold assembly parameter optimization group. The optimized mold assembly parameters are input into the parameterized discriminator module for secondary compatibility discrimination. When the secondary compatibility discrimination probability value reaches the preset compatibility threshold, the optimized mold assembly parameters are output as the target mold assembly parameters to the step of calling the mold frame parameter component library for component matching based on the optimized mold assembly parameters, replacing the original optimized mold assembly parameters.

9. The method for rapid generation of injection mold structures based on parametric design according to claim 8, characterized in that, The method further includes: Extract the product wall thickness parameterized distribution set and the product rib parameterized distribution set from the parameterized description group of the injection molded product model. Input the product wall thickness parameterized distribution set and the product rib parameterized distribution set into a pre-trained injection molded defect prediction graph convolutional network for node feature aggregation processing. The injection molded defect prediction graph convolutional network includes a parameterized graph convolutional layer and a parameterized attention aggregation layer. The parameterized graph convolutional layer is used to perform neighborhood information convolution processing on the wall thickness features and rib features of each node in the product geometric topology graph to generate node hidden feature groups. The parameterized attention aggregation layer is used to adaptively allocate the influence weights between different node hidden feature groups to generate a global defect-sensitive feature vector group. Based on the global defect sensitive feature vector group, the injection molding defect type parameterized classifier is called to perform defect category probability distribution prediction processing, generating an injection molding defect probability distribution group containing the probability values ​​of shrinkage defects, warpage defects, and cavitation defects. When the probability value of the shrinkage defect exceeds the preset shrinkage alarm threshold, the parameterized expression set of the cavity surface curvature in the cavity geometric parameter set is input into the shrinkage compensation parameterized inverse network for local curvature correction inversion processing. The shrinkage compensation parameterized inverse network includes a parameterized inversion convolutional layer and a parameterized curvature regression layer. The parameterized inversion convolutional layer is used to calculate the curvature correction requirement intensity parameter set of each region of the cavity surface in reverse according to the shrinkage defect probability value. The parameterized curvature regression layer is used to map the curvature correction requirement intensity parameter set into a local control point displacement vector set of the cavity surface. Based on the displacement vector group of local control points on the cavity surface, the set of parameterized expressions of the cavity surface curvature in the cavity geometric parameter group is subjected to local curvature lifting correction processing to generate an anti-shrink mark optimized cavity geometric parameter group. The anti-shrink mark optimized cavity geometric parameter group is then output as the updated cavity geometric parameter group to the step of performing assembly gap matching processing on the cavity geometric parameter group and the core geometric parameter group. When the warpage defect probability value exceeds the preset warpage alarm threshold, the parameterized constraint set of the core body size in the core geometric parameter set is input into the warpage compensation parameterized inverse network for cooling channel layout parameterized offset inversion processing. The warpage compensation parameterized inverse network includes a parameterized warpage inversion convolutional layer and a parameterized channel layout regression layer. The parameterized warpage inversion convolutional layer is used to calculate the position offset requirement parameter set of each segment of the cooling channel inside the core in reverse according to the warpage defect probability value. The parameterized channel layout regression layer is used to map the position offset requirement parameter set into a set of coordinate offset vectors of the control point of the cooling channel centerline. Based on the coordinate offset vector group of the control point of the cooling channel centerline, the preset initial cooling channel layout parameterized description group is subjected to channel centerline position offset correction processing to generate a cooling channel layout parameterized description group optimized for anti-warping, and the cooling channel layout parameterized description group optimized for anti-warping is added to the mold assembly parameter group as an additional parameter group. When the probability value of the cavitation defect exceeds the preset cavitation alarm threshold, the parting surface curve parameter group in the mold assembly parameter group is input into the cavitation escape channel parameterization generation network for parameterized layout processing of the venting groove position. The cavitation escape channel parameterization generation network includes a parameterized parting surface curvature analysis layer and a parameterized venting groove position regression layer. The parameterized parting surface curvature analysis layer is used to calculate the cavitation accumulation risk index parameter group of each curve segment in the parting surface curve parameter group. The parameterized venting groove position regression layer is used to generate the venting groove's slotting position parameterized coordinate group and slotting depth parameterized value group on the parting surface based on the cavitation accumulation risk index parameter group. Based on the parameterized coordinate group of the slotting position and the parameterized numerical group of the slotting depth, the parting surface curve parameter group in the mold assembly parameter group is processed to add venting groove features, generating an optimized parting surface curve parameter group with venting groove structure, and the optimized parting surface curve parameter group with venting groove structure is output as the updated parting surface curve parameter group to the step of calling the mold frame parameter component library for component matching processing based on the mold assembly parameter group.

10. A rapid generation system for injection mold structures based on parametric design, characterized in that, The method includes a processor and a computer-readable storage medium storing machine-executable instructions that, when executed by the processor, implement the rapid generation method for injection mold structures based on parametric design as described in any one of claims 1-9.

Images