A method and system for generating a casting process model of an automobile chassis part based on multi-modal feature recognition

Abstract

The application provides a kind of automobile chassis parts casting process model generation method and system based on multi-modal feature recognition, full-process automatic generation, effectively solve the traditional process modeling excessive dependence on manual experience, design cycle is long, the industry pain point of poor consistency.This application integrates multi-strategy feature recognition algorithm based on rules, section analysis and topological spread, can accurately extract key process features such as main shaft hole, machining hole and complex curved surface and establish a unified local coordinate system, greatly improve the recognition accuracy of complex structure features.Through standardized process rule base and standard part library, realize blank automatic processing, standard part automatic selection assembly and mold automatic generation, the traditional modeling cycle of several days is compressed to several hours.At the same time, the application is optimized after selecting the chassis parts, the process scheme is in line with the actual production, effectively ensures the consistency and reliability of the process model, significantly reduces the casting process design cost and human error risk.

Description

Technical Field

[0001] This invention relates to the field of automotive chassis component casting technology, and in particular to a method and system for generating automotive chassis component casting process models based on multimodal feature recognition. Background Technology

[0002] As the core system for vehicle load-bearing and transmission, the automotive chassis's key components (such as steering knuckles and wheel hubs) directly affect the vehicle's driving safety and handling stability. These components typically feature complex structures, significant variations in wall thickness, and harsh stress environments, thus placing extremely high demands on their casting quality. Currently, in the traditional casting production of key automotive chassis components, process design remains a crucial factor determining product quality and production efficiency.

[0003] Existing casting process modeling workflows primarily rely on designers' experience and typical trial-and-error methods. Typically, designers, based on a 3D model of the part and combining personal experience with process manuals, initially determine the layout of the parting line, gating system, risers, and chills. Subsequently, numerical simulation software (such as ProCAST and MagmaSoft) is used to simulate and analyze the filling and solidification processes. By observing predicted shrinkage cavities, porosity, and stress concentration defects, the process plan is repeatedly modified until the simulation results meet quality requirements. Existing technologies face the following significant challenges: over-reliance on human experience, resulting in long design cycles and poor consistency; difficulty in quantifying the nonlinear relationship between feature recognition and process mapping; and a pronounced data silo effect, lacking the fusion and utilization of multimodal information.

[0004] Therefore, it is necessary to propose a method and system for generating casting process models of automotive chassis components based on multimodal feature recognition in order to solve or at least alleviate the above-mentioned defects. Summary of the Invention

[0005] The main objective of this invention is to provide a method and system for generating casting process models of automotive chassis components based on multimodal feature recognition, in order to solve the problems of excessive reliance on manual experience in casting process modeling in the prior art, resulting in long design cycles and poor consistency.

[0006] To achieve the above objectives, this invention provides a method for generating a casting process model for automotive chassis components based on multimodal feature recognition, comprising the following steps: S1, Obtain the corresponding initial model based on the target requirements; S2, Based on multimodal feature recognition technology, perform local feature recognition on the initial model, extract local features, and establish a local coordinate system based on the local features; S3, perform blanking on the initial model to generate a blank 3D model; S4. Based on the local features and the local coordinate system, match and import the appropriate standard parts from the standard parts library, and construct the 3D model shape of the mold after the automatic assembly of the standard parts is completed. S5, Subtract the blank 3D model and the assembled standard parts from the shape of the mold 3D model to obtain the mold 3D model; S6, export the 3D model of the mold and related geometric information to complete the automatic generation of the casting process model of the automobile chassis parts.

[0007] Preferably, step S1 specifically includes the following steps: S11, Create model entries based on target requirements, select modeling objects according to the model entries, and determine the corresponding modeling category based on the modeling objects; S12, Based on the modeling object and the modeling category, obtain the corresponding initial model.

[0008] Preferably, the modeling objects include wheels and steering knuckles; wherein, when the modeling object is selected as a wheel, the modeling category includes cast wheels and cast-rotor wheels, and the structural existence parameters of buckle structures and undercut structures are synchronously associated. When the modeling object is selected as a steering knuckle, the modeling category includes double wishbone front steering knuckle, MacPherson strut front steering knuckle, MacPherson strut clamp front steering knuckle, and multi-link rear steering knuckle.

[0009] Preferably, step S2 specifically includes the following steps: S21, invoke a multi-strategy, multi-modal feature recognition algorithm that integrates rules, cross-section analysis, and topology spread proximity search to parse the initial model; S22, Identify and extract key local features of the casting process of the initial model, the local features including spindle hole size parameters, set of spatial distribution of machined holes and curvature distribution of feature surfaces; S23. Using the identified spindle hole center axis as the reference axis and the intersection of the machined hole positioning point or the curved surface with the spindle hole center axis as the reference origin, a local coordinate system is constructed.

[0010] Preferably, step S3 specifically includes the following steps: Based on the modeling object and modeling category, the blank is processed using the corresponding rules; wherein, when the modeling object is a wheel, the offset curve and surface required to generate the blank are calculated, the self-intersecting part is trimmed and rounded, the rim section is thickened, the undercut structure is processed, and the spindle hole, valve hole and bolt hole are filled with process to generate the 3D blank model of the wheel. When the modeling object is a steering knuckle, a riser of the corresponding size is added, and the spindle hole is filled to generate a blank 3D model of the steering knuckle.

[0011] Preferably, step S4 specifically includes the following steps: S41, construct a standard parts 3D model library; wherein, the standard parts 3D model library includes sprue bushings, sprue cups, and runner cones; S42, Based on the local features, select the corresponding standard parts from the standard parts 3D model library; S43, the selected standard parts are assembled onto the blank 3D model according to the local coordinate system to construct the shape of the mold 3D model.

[0012] Preferably, step S5 specifically includes the following steps: S51, call the three-dimensional Boolean operation engine, use the shape of the mold 3D model as the tool body, and use the blank 3D model and the assembled standard parts group as the cutting body to perform multi-body spatial subtraction operation. S52 performs topological integrity verification on the mold cavity surface after subtraction, and finally outputs a continuous 3D model of the mold.

[0013] Preferably, step S6 specifically includes the following steps: S61, export the 3D model of the mold into multiple format models; wherein, the format models include x_t format, step format, and prt format; S62, and export the geometric information of the mold 3D model to complete the automatic generation of the automotive chassis component casting process model; wherein, the geometric information includes the main axis direction and reference point coordinates.

[0014] Preferably, the step S2 of extracting local features specifically includes the following steps: S221, a method library built to execute within the NX internal environment; S222, Extract key features and contour lines from the initial model based on the method library; S223, using a combination of slicing and layering and region merging, local features of the initial model are extracted.

[0015] This application also provides a system for generating casting process models of automotive chassis components based on multimodal feature recognition, applied to the method for generating casting process models of automotive chassis components based on multimodal feature recognition as described above. The system includes a modeling category management module, a product model management module, a blank 3D model management module, a multimodal feature recognition management module, a standard parts management module, a model assembly module, a process model rule management module, and a mold 3D model management module. Specifically, the modeling category management module manages the modal modeling categories of different modeling objects; the product model management module manages product models of different modeling objects; the blank 3D model management module manages blank 3D models; the multimodal feature recognition management module performs multimodal feature recognition on the initial model; the standard parts management module manages compatible standard parts; the model assembly module assembles compatible standard parts into the blank 3D model; the process model rule management module manages process model rules and guides the model assembly module to complete automatic assembly according to the rules; and the mold 3D model management module manages, creates, and exports 3D models.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention provides a method and system for generating casting process models for automotive chassis components based on multimodal feature recognition. The entire process is automatically generated, effectively solving the industry pain points of traditional process modeling, such as excessive reliance on human experience, long design cycles, and poor consistency. This invention integrates multi-strategy feature recognition algorithms based on rules, cross-section analysis, and topology spread. It can accurately extract key process features such as spindle holes, machined holes, and complex curved surfaces and establish a unified local coordinate system, significantly improving the recognition accuracy of complex structural features. Through a standardized process rule library and standard parts library, it achieves automatic blank processing, automatic selection and assembly of standard parts, and automatic mold generation, compressing the traditional modeling cycle of several days to several hours. Furthermore, this invention performs specific optimizations after selecting chassis components, ensuring that the process scheme closely matches actual production, effectively guaranteeing the consistency and reliability of the process model, and significantly reducing casting process design costs and the risk of human error. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0018] Figure 1 This is a process flow diagram of the overall model generation method in one embodiment of the present invention.

[0019] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0023] Furthermore, the technical solutions of the various embodiments can be combined with each other, but only if they are feasible for those skilled in the art. If the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0024] Please see the appendix Figure 1 The present invention provides a method for generating a casting process model of automotive chassis components based on multimodal feature recognition, comprising the following steps: S1 is the initial model obtained based on the target requirements; it serves as a standardized input for the entire process.

[0025] S2, based on multimodal feature recognition technology, local feature recognition is performed on the initial model, local features are extracted, and a local coordinate system is established based on the local features; by integrating rule matching, cross-section analysis and topology propagation multi-strategy algorithms, local features are accurately extracted; and a unified local coordinate system is established around the local features to provide a geometric benchmark for all subsequent process operations, solving the problems of missed detection and false detection in traditional manual feature recognition.

[0026] S3, the initial model is processed to generate a blank 3D model; the conversion from product to blank is automatically completed according to the preset process rules, providing a foundation for the final establishment of the subsequent model.

[0027] S4. Based on the local features and the local coordinate system, match and import suitable standard parts from the standard parts library, and construct the mold 3D model shape after the standard parts are automatically assembled. Based on the identified feature parameters, automatically match process standard parts from the standardization library and complete error-free automatic assembly along the local coordinate system. Then, automatically determine the parting surface and mold size according to the blank outline and process category, and generate the mold shape including the mold closing guide structure to realize the integrated automatic design of the gating system and the mold base.

[0028] S5, subtract the blank 3D model and the assembled standard parts from the shape of the mold 3D model to obtain the mold 3D model; call the high-precision three-dimensional Boolean operation engine, take the mold shape as the target body and subtract the blank and all standard parts at once to automatically form the mold cavity and core; and simultaneously perform topology integrity verification to repair geometric defects such as broken surfaces and non-manifold edges to ensure that the generated mold model can be directly used for CNC machining.

[0029] S6, export the 3D model of the mold and related geometric information to automatically generate the casting process model of the automotive chassis parts; export the final mold as a whole and each independent entity into a general three-dimensional format, and extract core geometric information such as the main axis direction and reference point coordinates to achieve seamless data connection with subsequent casting simulation, CNC machining, cost accounting and other links, and complete the full-process automated closed loop.

[0030] In a preferred embodiment of the present invention, step S1 specifically includes the following steps: S11. Create model entries based on target requirements, select modeling objects according to the model entries, and determine the corresponding modeling category based on the modeling objects. First, receive the casting process target requirements input by the user, including core information such as part model, material grade, production batch, and quality grade, and generate standardized model entries containing a unique project identifier, modeling object type, and core process requirements. This entry serves as the main thread of the entire modeling process, uniformly linking and managing the parameters and results of all subsequent steps. Then, select modeling objects according to the preset selectable range of the model entries.

[0031] In a preferred embodiment, this application focuses on specific optimizations for the two most critical types of cast components in an automotive chassis—wheels and steering knuckles. When a wheel is selected as the modeling object, the system automatically loads the corresponding modeling category set, including cast wheels and cast-rotor wheels, and simultaneously associates the existence parameters of the riser structure and the undercut structure. This is because the wheel casting process is highly dependent on the presence or absence of these two structures: the riser structure directly affects the arrangement of the riser, while the undercut structure determines whether a side core-pulling mold structure is required.

[0032] When the modeling object is a steering knuckle, the system automatically loads four modeling categories: double wishbone front steering knuckle, MacPherson strut front steering knuckle, MacPherson strut clamping front steering knuckle, and multi-link rear steering knuckle. Different types of steering knuckles differ significantly in structural features, stress distribution, and casting process requirements. For example, the MacPherson strut front steering knuckle has a clamping arm structure, requiring a special gating and riser arrangement, while the multi-link rear steering knuckle has more and more complex arms, demanding stricter requirements on the filling sequence. By pre-defining these modeling categories, the system can automatically call the corresponding process rules, significantly improving modeling efficiency and accuracy.

[0033] S12, based on the modeling object and the modeling category, obtain the corresponding initial model; thus, according to the selected modeling object and modeling category, automatically route to the corresponding product model management module to retrieve and obtain the matching initial model. When the modeling object is a wheel, the initial model is the product model of the wheel. When the modeling object is a steering knuckle, the initial model is the product model of the steering knuckle.

[0034] In a preferred embodiment of the present invention, step S2 specifically includes the following steps: S21, a multi-strategy, multimodal feature recognition algorithm integrating rule-based, cross-section analysis, and topological propagation proximity search is invoked to analyze the initial model. It is worth noting that three complementary feature recognition strategies are integrated to achieve optimal recognition of different types of features: Rule-based method: suitable for recognizing standard features with well-defined geometric shapes, such as cylindrical holes and conical holes. By matching the geometric parameters of the features with preset rules, this method quickly and accurately identifies such features. Cross-section analysis method: suitable for recognizing features with uniform cross-sections, such as wheel rims and steering knuckles. By analyzing the shape and size variations of multiple cross-sections, the complete contour of the feature is extracted. Topological propagation proximity search method: suitable for recognizing complex, interconnected features, such as surface groups and hole systems. Starting from a known seed feature, it uses topological propagation search to identify all related features.

[0035] Specifically, a rule-based method is used to target standard features with clearly defined geometric attributes (cylindrical surfaces, planes, conical surfaces, spheres, etc.). Based on a preset set of geometric decision rules, the surfaces of the entity are screened, grouped, and their overall geometric attributes are calculated. After the rule-based method, a cross-sectional analysis method is used to perform multi-angle / multi-height cross-sectional scanning on the entity, reducing the 3D geometry to a 2D contour sequence. Then, signal processing techniques (frequency domain transformation, curvature analysis, threshold filtering) are used to extract the implicit geometric abrupt change points from the contour sequence. Finally, starting from the geometric abrupt change points output by the rule-based method, a breadth-first search propagation (i.e., topological propagation neighbor search method) is performed along the surface-edge topological adjacency graph. Adjacent surfaces that meet the conditions are absorbed layer by layer through surface type conditions and geometric constraints until convergence.

[0036] This invention is not a simple combination of three methods, but rather a hierarchical collaborative decision-making mechanism based on "rule-based method to prioritize locking standard features → cross-section analysis method to extract uniform cross-section features → topology propagation method to complete complex surface features". It also incorporates conflict resolution rules: when the results of the rule-based method and the topology propagation method are inconsistent, the result of the topology propagation method shall prevail, because it is based on real topological relationships rather than preset geometric assumptions.

[0037] S22, Identify and extract key local features of the casting process from the initial model. These local features include the spindle hole size parameters, the set of spatial distributions of machined holes, and the curvature distribution of characteristic surfaces. From all the initially identified features, select the key local features that have a decisive impact on the casting process design, including the spindle hole size parameters, the set of spatial distributions of machined holes, and the curvature distribution of characteristic surfaces. The spindle hole is the core reference of the entire component, and its size directly determines the selection of standard parts such as the sprue bushing and the flow divider cone. The position of the machined holes determines the arrangement of the core in the mold. The curvature distribution of the characteristic surfaces affects the filling flow and solidification sequence of the molten metal.

[0038] S23. Using the identified spindle hole center axis as the reference axis and the intersection of the machining hole positioning point or curved surface with the spindle hole center axis as the reference origin, a local coordinate system is constructed. This unified local coordinate system is established using the spindle hole as the most stable and critical feature. The spindle hole is the core reference for the machining and assembly of automotive chassis parts. Establishing a local coordinate system based on it ensures the consistency of the casting process model with the subsequent machining and assembly models. It is worth noting that the method for constructing the local coordinate system can be: traversing surfaces to extract cylindrical and planar surfaces; calculating the direction of the main coordinate system; and calculating the position of the origin along the main axis direction. Generally, when the direction of the local coordinate system is close to that of the absolute coordinate system, the direction of the absolute coordinate system should be used for calibration. When the modeling object is a wheel, the origin of the local coordinate system is aligned with the flange plane; when the modeling object is a steering knuckle, the origin of the local coordinate system is aligned with the bottom surface of the blank.

[0039] Specifically, the identified spindle hole center axis is taken as the Z-axis reference axis, the intersection of the machined hole positioning point or surface with the spindle hole center axis is taken as the coordinate origin O, the machined hole positioning feature direction in the plane perpendicular to the Z-axis and passing through the origin is taken as the X-axis, and the Y-axis is determined according to the right-hand rule, thereby constructing a local coordinate system {O-XYZ}.

[0040] When the spindle bore is irregular, the least squares method is used to fit the center axis; when there are multiple conflicting intersections, the point closest to the flange surface is selected. The specific steps include: Least-squares fitting of the central axis when the spindle hole is irregular. In actual parts, due to casting deviations, wear, thermal deformation, etc., the inner wall of the spindle hole is often not an ideal cylindrical surface, but has irregular shapes such as ellipticity deviation, taper deviation, or local depressions. If the line connecting the centers of any two cross-sectional circles is directly taken as the axis, it will lead to a large error in the direction and position of the axis, thus causing the entire local coordinate system to be skewed. The weighted least squares (WLS) method is used to fit the central axis of multiple cross-sectional point clouds of the inner wall of the spindle hole. The specific steps are as follows: Cross-sectional sampling: Along the axial direction of the spindle hole, extract cross-sections at equal intervals (recommended 2~5mm, adaptively adjusted according to the hole length; total number of cross-sections = [hole length / equal interval] + 1). For each cross-section, use an edge detection algorithm to extract the hole wall contour point set.

[0041] The center of each cross section is initially estimated; for each set of points on the cross section, the center and radius of the fitted circle are obtained by using the Taubin circle fitting algorithm (which has high algebraic accuracy and no bias).

[0042] Weighted least squares axis fitting: Treat the centers of all cross sections as sampling points on the axis, and use the weighted least squares method to fit a spatial straight line.

[0043] Outlier removal (RANSAC-assisted): Before fitting, RANSAC (Random Sample Consensus) is used to iteratively remove outliers, specifically including the following steps: Three cross-sectional centers are randomly selected, and a temporary axis is fitted. Calculate the distance from all circle centers to the temporary axis and calculate the proportion of interior points; Repeat K times (K=100), and take the axis with the largest proportion of interior points as the initial value; The distance threshold is twice the standard deviation of the distances from all circle centers to the initial axis.

[0044] Output the fitted axis; the output includes the axis direction vector, reference points on the axis, and axis quality indicators.

[0045] The nearest flange face selection strategy in case of multiple intersection point conflicts. When the intersection point of the curved surface and the central axis of the spindle hole is used as the reference origin, if multiple curved surfaces (such as the wall surfaces of multiple machined holes, multiple boss end faces, etc.) intersect with the axis, multiple candidate intersection points will be generated, resulting in a conflict. The flange face is the clamping reference surface of the part. The intersection point closest to the flange face means that the positioning feature participates in positioning first in the clamping state and has the highest positioning priority. The principle of prioritizing the point closest to the flange face is adopted, and the point closest to the flange face is selected as the coordinate origin.

[0046] As an optional embodiment of the present invention, step S2, which uses the identified spindle hole center axis as the reference axis and the intersection of the machined hole positioning point or curved surface with the spindle hole center axis as the reference origin, specifically includes the following steps: After the initial establishment of the coordinate system, an automatic calibration algorithm based on positioning features can be used to compare the known theoretical coordinates of the machined hole positioning points (from the CAD model) with the measured coordinates, and the coordinate system can be calibrated by improving the ICP (Iterative Closest Point) algorithm. The system provides self-checking and quality assessment functions for the established local coordinate system. Specific check items, judgment criteria, and handling of failures are listed in the table below: If all self-tests pass, the final coordinate system is output. If any fails, the local coordinate system needs to be rebuilt or automatic calibration needs to be performed again until the self-test passes.

[0047] In a preferred embodiment of the present invention, step S3 specifically includes the following steps: Based on the modeling object and modeling category, corresponding rules are used for blank processing. Specifically, when the modeling object is a wheel, the system calculates and generates the required offset curves and surfaces for the blank, trims and fillets the self-intersecting parts, adds rim section thickening, processes the undercut structure, and performs process filling processing on the spindle hole, valve hole, and bolt hole to generate a 3D blank model of the wheel. When the modeling object is a wheel, the system first calculates and generates the required offset curves and surfaces for the blank, uniformly thickening the product model's wall to the dimensions required by the casting process. Then, the self-intersecting parts generated during the offset process are automatically trimmed and filled to avoid casting defects caused by sharp edges. Next, rim section thickening is added to compensate for material thinning during the casting process. Finally, the undercut structure is processed, and based on the size and position of the undercut, a reserved space for the side core-pulling structure is automatically generated to form the 3D blank model of the wheel. When filling the spindle hole, the top surface is required to align with the flange surface, and the bottom surface is required to align with the lowest identifiable rotating surface, ignoring the drainage groove. The lowest rotating surface includes non-cylindrical surfaces. The method for filling valve holes can be as follows: obtain the inner cylindrical surface of the valve hole and the edges at both ends of the inner cylindrical surface of the valve hole, crawl the surface between the two cylindrical surfaces, extract the surface features of the valve hole, fill the surface based on the edges at both ends, and finally stitch the surface, where the outermost edge uses the edge of the cylindrical surface. Filling bolt holes can be done by referring to the method for filling valve holes, where the outermost edge includes the rotating surface of the cylindrical surface. The method for adding rim section thickness can be as follows: extract the protrusion on the back cavity side as the starting point, and then calculate according to the casting / spinning and undercut classification. For cast wheels, first find the outer side of the section [radial coordinate = starting point coordinate + casting rim root thickness] to determine the starting position of the rim. For cast wheels, select segmented offset. When there is an undercut, set the starting point to the lowest point of the back cavity reference profile. Then calculate and generate the rim curve, create a point set, create a spline curve based on the created point set, create a body of revolution based on the created spline curve to complete the rim creation, and finally set the rim section thickness as needed.

[0048] For wheel parts with the spindle bore center axis (Z-axis) as the rotation axis, an undercut refers to an area on the part surface where the normal vector makes an angle greater than 90° with the rotation axis direction. This area has a radial "backstripping" and cannot be directly demolded by axial mold opening; a side core-pulling mechanism is necessary. The processing of wheel undercut structures can be divided into two parts: identifying the undercut structure and parametrically calculating the space reserved for side core-pulling. The wheel undercut structure can be automatically identified using normal vector criteria, radial projection overlap criteria, or envelope interference criteria.

[0049] When the modeling object is a steering knuckle, a riser of the corresponding size is added, and the spindle hole is filled to generate a 3D model of the steering knuckle blank. When the modeling object is a steering knuckle, the system first adds a riser of the corresponding size at the hot spot location based on the steering knuckle's structure and material properties to achieve sequential solidification and prevent shrinkage cavities and porosity defects. Then, the spindle hole is filled to generate a 3D model of the steering knuckle blank. This is because the spindle hole is usually formed by subsequent machining and does not need to be cast during casting; filling the spindle hole simplifies the mold structure and improves the density of the casting. Before adding the riser, the draft angle of the riser needs to be determined based on the steering knuckle product model, the steering knuckle blank entity, and the local coordinate system of the steering knuckle. The calculation method can be as follows: based on the steering knuckle coordinate system, extract the bottom plane on the blank entity, then extract all edges of the bottom plane, calculate the closest distance to the spindle, and calculate the upper limit of the draft angle based on the bottom plane information and the size of the spindle hole (corresponding riser bottom size), taking the minimum value between 25° and 25°. Before generating the 3D model of the blank, it is necessary to calculate the mold envelope and mold groove based on the steering knuckle blank entity and the local coordinate system of the steering knuckle. The calculation method of the mold envelope can be as follows: calculate the rectangle based on the envelope in the local coordinate system of the steering knuckle; rotate along the principal axis to calculate the envelope in multiple coordinate systems, calculate the maximum cutting angle of each corner, and thus calculate the octagonal envelope.

[0050] The calculation method for the mold groove can differ for different types of steering knuckles: For the front steering knuckle of a double wishbone, the calculation method for the mold groove can be as follows: based on the middle section of the steering knuckle, without cutting through the steering tie rod side, extract all sections of the steering knuckle side view to form an overall contour, and based on the overall contour, calculate the trapezoid that can be excavated from the top mold above the main shaft and the trapezoid that can be excavated from the bottom mold below the arm, and then perform transverse cutting.

[0051] For MacPherson strut front steering knuckles and MacPherson clamping front steering knuckles, the calculation method for the mold groove can be as follows: taking the height of the steering tie rod as a reference, for steering tie rods that are too high, take the middle section; for steering tie rods that are too low, cut through the steering tie rod according to the actual section, extract all the sections of the steering knuckle side view, form the overall outline, and based on the overall outline, calculate the trapezoid that can be excavated from the top mold above the main shaft, and calculate the trapezoid that can be excavated from the bottom mold below the arm, and then perform transverse cutting.

[0052] For multi-link rear steering knuckles, the calculation method for mold grooves can be as follows: extract the surface of the steering knuckle visible from above, slice it along the main axis, calculate the largest rectangle on each slice that contains the main axis and does not interfere with the extracted surface, select the groove with the largest cuttable volume based on the slice position, and then cut it downwards.

[0053] Furthermore, step S4 specifically includes the following steps: S41, Construct a 3D model library of standard parts; wherein, the 3D model library of standard parts includes sprue bushings, sprue cups, and runner cones; parametrically model commonly used standard parts in the casting process to construct a hierarchical and categorized standardized process standard parts library. This library stores 3D models of sprue bushings, sprue cups, and runner cones of different specifications, and associates them with the applicable process parameters, matching features, and assembly tolerances of each standard part. For example, sprue bushings are divided into multiple series according to their inner diameter, and each series is associated with corresponding applicable casting weight, pouring temperature, and other parameters.

[0054] S42, based on the local features, the system selects corresponding standard parts from the standard parts 3D model library; it compares the identified key local features with the matching rules in the process model rule library, and automatically selects the most suitable standard parts. For example, the system selects a sprue bushing with a matching inner diameter based on the inner diameter of the spindle hole; and selects a sprue cup with a suitable capacity based on the weight of the casting and the pouring time requirements.

[0055] S43, the selected standard parts are assembled onto the blank 3D model according to the local coordinate system to construct the mold 3D model outline; based on the local coordinate system established in step S2, the selected standard parts are automatically assembled to the corresponding process positions of the blank 3D model according to the preset assembly positions and tolerance requirements in the process model rule library. For example, the central axis of the sprue bushing coincides with the central axis of the spindle hole, the bottom surface of the sprue cup coincides with the parting surface of the mold, and the bottom surface of the riser is aligned; after the standard parts are assembled, the system automatically determines the mold parting surface position and mold outline size parameters according to the selected modeling category and the outer contour of the blank 3D model, and generates the mold 3D model outline including the parting surface and mold closing guide structure.

[0056] Furthermore, step S5 specifically includes the following steps: S51, the 3D Boolean operation engine is invoked, using the shape of the mold 3D model as the tool body and the blank 3D model and the assembled standard parts group as the cutting bodies, to perform multi-body spatial subtraction operation; the high-precision 3D Boolean operation engine built into NX is invoked to perform a one-time multi-body Boolean difference operation. Compared with the traditional method of subtraction one by one, the one-time multi-body subtraction operation is more efficient and can avoid geometric topological errors caused by multiple operations. For example, based on the coordinate system and riser information, a riser entity is generated, the bottom surface and bottom edge of the riser are extracted, the riser and the initial blank are merged, the search starts from the bottom surface of the riser, the merged boundary edge is extracted, and fillets are created to complete the riser creation. As another example, based on the mold groove information, extrusion and cutting are performed, the mold edge extrusion is generated based on the mold envelope contour, and the extruded mold envelope is merged with the grooved mold entity to realize mold grooving.

[0057] S52 performs a topological integrity check on the mold cavity surface after subtraction, ultimately outputting a structurally continuous 3D mold model. A comprehensive topological check is performed on the mold model after Boolean operations, including checking for geometric defects such as broken surfaces, overlapping surfaces, and non-manifold edges. If defects are found, the system automatically repairs them to ensure the topological integrity of the mold model.

[0058] Furthermore, step S6 specifically includes the following steps: S61, export the mold 3D model into multiple format models; wherein, the format models include x_t format, step format, and prt format; export the final mold 3D model as a whole into multiple common 3D format files such as x_t, step, and prt. These common 3D formats are currently the most widely used neutral file formats in the manufacturing industry, and can be recognized by all CAD / CAM / CAE software, ensuring that the mold model can be seamlessly transferred between different software environments.

[0059] S62, and export the geometric information of the 3D model of the mold to complete the automatic generation of the casting process model for automotive chassis components; wherein, the geometric information includes the principal axis direction and reference point coordinates; export the x_t format, step format, and prt format files of each independent entity in the 3D model of the mold respectively, and simultaneously export the associated geometric information of each entity, including the principal axis direction, reference point coordinates, entity volume, and surface area parameters. This geometric information is an important basis for subsequent numerical simulation of the casting process, mold cost accounting, and production planning.

[0060] Furthermore, the extraction of local features in step S2 specifically includes the following steps: S221 is a method library built and executed within the NX internal environment. Based on the secondary development interface of the NX 3D design software, it constructs a natively executed feature extraction method library. This library pre-stores dedicated algorithm logic for extracting key features and contours from wheel, steering knuckle, and M1 mold models, and can directly call NX's core geometry engine without data format conversion. This library is developed based on the NXOpenAPI (C++ / Python) and NXJournal interfaces, and all algorithm logic directly calls NX's internal UF_ The UnigraphicsFunction and NXOpen libraries eliminate the need to export to intermediate formats such as STEP / IGES. Data flows entirely using NX's native Tag identifiers and Parasolid kernel data structures.

[0061] S222, based on the method library, extract key features and contour lines from the initial model; call the dedicated algorithm in the method library to perform overall analysis on the initial model, and extract all key features and contour lines that may be related to the casting process, including hole features, boss features, groove features, parting line candidate contours, etc.

[0062] S223 employs a combination of slicing and region merging to extract local features of the initial model. The 3D initial model is sliced ​​and layered along multiple directions, transforming the complex 3D geometry problem into a relatively simple 2D contour problem. Then, region merging analysis is performed on the contours of adjacent slices to reconstruct the complete boundaries of the 3D features. This method is particularly suitable for extracting features that are difficult to identify using a single rule, such as wheel back cavities and complex steering knuckle surfaces.

[0063] This application also provides a system for generating casting process models of automotive chassis components based on multimodal feature recognition, applied to the method for generating casting process models of automotive chassis components based on multimodal feature recognition as described above. The system includes a modeling category management module, a product model management module, a blank 3D model management module, a multimodal feature recognition management module, a standard parts management module, a model assembly module, a process model rule management module, and a mold 3D model management module. Specifically, the modeling category management module manages the modal modeling categories of different modeling objects; the product model management module manages product models of different modeling objects; the blank 3D model management module manages blank 3D models; the multimodal feature recognition management module performs multimodal feature recognition on the initial model; the standard parts management module manages compatible standard parts; the model assembly module assembles compatible standard parts into the blank 3D model; the process model rule management module manages process model rules and guides the model assembly module to complete automatic assembly according to the rules; and the mold 3D model management module manages, creates, and exports 3D models.

[0064] Understandably, the Modeling Category Management module manages the modal modeling categories of different modeling objects, stores various modeling categories for wheels and steering knuckles and their associated process parameters, provides users with a modeling category selection interface, and passes the selected modeling category information to other modules. The Product Model Management module manages product models for different modeling objects, providing upload, retrieval, and download functions for product models. Based on the information passed by the Modeling Category Management module, it passes the matching product model to the Multimodal Feature Recognition Management module and the Blank 3D Model Management module. The Blank 3D Model Management module manages blank 3D models, receives product models from the Product Model Management module, calls the product-to-blank conversion rules in the Process Model Rule Management module, and converts the product model into a blank 3D model. It also supports the upload and management of pre-built blank 3D models and passes the final blank 3D model to the Model Assembly module and the Mold 3D Model Management module. The multimodal feature recognition and management module performs multimodal feature recognition on the initial model, invoking a multi-strategy algorithm that integrates rule-based, cross-section analysis, and topology-spreading proximity search to extract key local features and establish a local coordinate system. It then transmits the feature information and local coordinate system information to the model assembly module. The model assembly module assembles standard parts to fit the blank 3D model. It receives the feature information and local coordinate system information from the multimodal feature recognition and management module, invokes standard part assembly rules from the process model rule management module, retrieves matching standard parts from the standard part management module, completes automatic assembly, and transmits the assembled model to the mold 3D model management module. The process model rule management module manages process model rules, including product-to-blank rules, standard part selection rules, assembly position rules, and mold shape generation rules. It provides rule support to other modules, guiding the entire modeling process to execute automatically according to process specifications. Mold 3D Model Management Module: Used to manage, create, and export 3D models. It receives blank 3D models and assembled standard part models, automatically creates the mold 3D model outline and performs Boolean difference operations to generate the final mold 3D model, and exports the model and related geometric information to a specified format.

[0065] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for generating a casting process model for automotive chassis components based on multimodal feature recognition, characterized in that, Includes the following steps: S1, Obtain the corresponding initial model based on the target requirements; S2, Based on multimodal feature recognition technology, perform local feature recognition on the initial model, extract local features, and establish a local coordinate system based on the local features; S3, perform blanking on the initial model to generate a blank 3D model; S4. Based on the local features and the local coordinate system, match and import the appropriate standard parts from the standard parts library, and construct the 3D model shape of the mold after the automatic assembly of the standard parts is completed. S5, Subtract the blank 3D model and the assembled standard parts from the shape of the mold 3D model to obtain the mold 3D model; S6, Export the 3D model of the mold and related geometric information to complete the automatic generation of the casting process model of the automobile chassis parts; Step S1 specifically includes the following steps: S11, Create model entries based on target requirements, select modeling objects according to the model entries, and determine the corresponding modeling category based on the modeling objects; S12, Based on the modeling object and the modeling category, obtain the corresponding initial model; Step S2 specifically includes the following steps: S21, invoke a multi-strategy, multi-modal feature recognition algorithm that integrates rules, cross-section analysis, and topology spread proximity search to parse the initial model; S22, Identify and extract key local features of the casting process of the initial model, the local features including spindle hole size parameters, set of spatial distribution of machined holes and curvature distribution of feature surfaces; S23. Using the identified spindle hole center axis as the reference axis and the intersection of the machined hole positioning point or the curved surface with the spindle hole center axis as the reference origin, a local coordinate system is constructed. Step S3 specifically includes the following steps: Based on the modeling object and modeling category, the blank is processed using the corresponding rules; wherein, when the modeling object is a wheel, the offset curve and surface required to generate the blank are calculated, the self-intersecting part is trimmed and rounded, the rim section is thickened, the undercut structure is processed, and the spindle hole, valve hole and bolt hole are filled with process to generate the 3D blank model of the wheel. When the modeling object is a steering knuckle, a riser of the corresponding size is added, and the spindle hole is filled to generate a blank 3D model of the steering knuckle.

2. The method for generating a casting process model for automotive chassis components based on multimodal feature recognition according to claim 1, characterized in that, The modeling objects include wheels and steering knuckles; wherein, when the modeling object is selected as a wheel, the modeling category includes cast wheels and cast-rotor wheels, and the structural existence parameters of the buckle structure and undercut structure are simultaneously associated. When the modeling object is selected as a steering knuckle, the modeling category includes double wishbone front steering knuckle, MacPherson strut front steering knuckle, MacPherson strut clamp front steering knuckle, and multi-link rear steering knuckle.

3. The method for generating a casting process model for automotive chassis components based on multimodal feature recognition according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41, construct a standard parts 3D model library; wherein, the standard parts 3D model library includes sprue bushings, sprue cups, and runner cones; S42, Based on the local features, select the corresponding standard parts from the standard parts 3D model library; S43, the selected standard parts are assembled onto the blank 3D model according to the local coordinate system to construct the shape of the mold 3D model.

4. The method for generating a casting process model for automotive chassis components based on multimodal feature recognition according to claim 1, characterized in that, Step S5 specifically includes the following steps: S51, call the three-dimensional Boolean operation engine, use the shape of the mold 3D model as the tool body, and use the blank 3D model and the assembled standard parts group as the cutting body to perform multi-body spatial subtraction operation. S52 performs topological integrity verification on the mold cavity surface after subtraction, and finally outputs a continuous 3D model of the mold.

5. The method for generating a casting process model for automotive chassis components based on multimodal feature recognition according to claim 1, characterized in that, Step S6 specifically includes the following steps: S61, export the 3D model of the mold into multiple format models; wherein, the format models include x_t format, step format, and prt format; S62, and export the geometric information of the mold 3D model to complete the automatic generation of the automotive chassis component casting process model; wherein, the geometric information includes the main axis direction and reference point coordinates.

6. The method for generating a casting process model for automotive chassis components based on multimodal feature recognition according to claim 4, characterized in that, The extraction of local features in step S2 specifically includes the following steps: S221, a method library built to execute within the NX internal environment; S222, Extract key features and contour lines from the initial model based on the method library; S223, using a combination of slicing and layering and region merging, local features of the initial model are extracted.

7. A system for generating casting process models of automotive chassis components based on multimodal feature recognition, applied to the method for generating casting process models of automotive chassis components based on multimodal feature recognition as described in any one of claims 1-6, characterized in that, It includes a modeling category management module, a product model management module, a blank 3D model management module, a multimodal feature recognition management module, a standard parts management module, a model assembly module, a process model rule management module, and a mold 3D model management module. Specifically, the modeling category management module manages the modal modeling categories of different modeling objects; the product model management module manages product models of different modeling objects; the blank 3D model management module manages blank 3D models; the multimodal feature recognition management module performs multimodal feature recognition on the initial model; the standard parts management module manages compatible standard parts; the model assembly module assembles compatible standard parts into the blank 3D model; the process model rule management module manages process model rules and guides the model assembly module to complete automatic assembly according to the rules; and the mold 3D model management module manages, creates, and exports 3D models.

Images