Bronze ware cast parting design method and system based on Boolean modeling

Abstract

The invention discloses a bronze ware model casting parting design method and system based on Boolean modeling, particularly relates to the field of digital model casting modeling and simulation verification, is used for solving the problems of parting cutting error accumulation and model block seepage stability loss under the complex morphology of a bronze ware, and provides a model casting parting design method and system based on Boolean modeling by constructing a morphology noise mapping and semantic slicing linkage mechanism. Early recognition of pitting corrosion and back-off areas is achieved, Boolean segmentation is driven through curvature constraint, and a flexible buffer surface is introduced into a danger area to achieve transition; a cross-domain parameter evaluation and adjustment mechanism is formed in combination with multi-scale displacement optimization and physical simulation feedback, high-response parameter write-back is realized through local disturbance search, a full-link linkage path from modeling to casting is constructed, and parting integrity and model block stability are improved.

Description

Technical Field

[0001] This invention relates to the field of digital mold casting modeling and simulation verification, and more specifically, to a method and system for designing bronze mold casting patterns based on Boolean modeling. Background Technology

[0002] When digitally restoring the bronze artifacts in the museum's collection, staff first use laser scanning to capture the external surface and internal cavities of the objects. Then, they use Boolean modeling to automatically segment the parting surfaces and mold blocks for printing templates and subsequent casting replication. Years of oxidation have caused multiple layers of pitting and indentation on the surface of the artifacts, and the point cloud is filled with noise and gaps, making the geometry far more complex than that of conventional industrial parts.

[0003] However, the scanning error was directly entered into the Boolean operation without semantic annotation. The algorithm misjudged the deep hole etch as a penetrable area and generated a cut surface that penetrated the inner and outer walls during cutting. After the mold blocks were connected, extremely fine gaps were left. During the copper pouring stage, the seepage amplified the cracks and caused them to burst, resulting in the failure of the replication. This highlights that the existing process lacks the means to suppress the risk of coupling between morphological noise and Boolean logic.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a bronze casting parting design method and system based on Boolean modeling. By constructing a linkage mechanism between morphological noise mapping and semantic slicing, early identification of pitting and undercut areas is achieved. Boolean segmentation is driven by curvature constraints, and a flexible buffer surface is introduced in the danger zone to achieve transition. By combining multi-scale displacement optimization and physical simulation feedback, a cross-domain parameter evaluation and adjustment mechanism is formed. High-response parameter rewriting is achieved through local perturbation search, and a full-link linkage path from modeling to casting is constructed to improve parting integrity and mold stability, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The Boolean modeling-based parting line design method for bronze casting includes the following steps: S1: After collecting point clouds of the object's exterior and interior, density adaptive resampling is performed, and temporal identifiers are implanted for abnormal scattered points to generate a shape noise mapping. S2: Based on the topography noise mapping, the semantic slicing network is used to label the pitting and undercut regions, and simultaneously generate the curvature constraint matrix and the danger domain mask while maintaining the corresponding relationship. S3: Use the curvature constraint matrix to trigger the Boolean segmentation core, generate a flexible buffer surface in real time at the corresponding position of the danger domain mask, output a continuous fractal draft and mark the buffer surface parameters; S4: Input the buffer surface parameters and topology noise mapping into the multi-scale displacement optimization process, iteratively correct the micro-displacement of the cut surface and continuously check the connection risk with the curvature constraint matrix to obtain the assembleable block mesh. S5: Import the assemblable template mesh into the virtual casting simulation platform, extract cross-domain indicators through physical simulation to generate risk coefficients to evaluate the performance of the buffer surface, and automatically adjust the buffer surface parameters to eliminate seepage risk when the evaluation is insufficient, and output the final template file with controlled seepage risk.

[0007] In a preferred embodiment, step S1 includes the following: The original point cloud data set of the bronze artifact's exterior and interior was collected using a laser scanner, and the point cloud density distribution of the original point cloud data set was adjusted using density adaptive resampling technology to generate a resampled point cloud data set.

[0008] In a preferred embodiment, step S1 further includes the following: Anomaly point sets are identified based on the local curvature and median distance in the resampled point cloud dataset, and each anomaly point in the dataset is assigned a temporal identifier to record the time information during the scanning process. A three-dimensional mesh is constructed by combining the resampled point cloud dataset and the anomaly point set, and the noise intensity within the mesh cells in the three-dimensional mesh is calculated to generate a shape noise mapping that reflects the noise distribution characteristics of the bronze surface.

[0009] In a preferred embodiment, step S2 includes the following: Using the topography noise map as input, a pre-trained semantic slicing network is driven to annotate the geometric features in the resampled point cloud dataset, generating annotation results. Based on the annotation results, the curvature change rate in the neighborhood of points labeled as pitting or indentation is calculated, generating a curvature constraint matrix. Based on the annotation results and the topography noise map, a hazard region mask is generated to mark the parts of the pitting and indentation regions where the noise intensity is higher than a preset threshold. It is ensured that each item of the annotation results, curvature constraint matrix, and hazard region mask maintains a one-to-one correspondence with the points in the resampled point cloud dataset.

[0010] In a preferred embodiment, step S3 includes the following: Using the resampled point cloud dataset, annotation results, curvature constraint matrix, and hazard domain mask as input, the Boolean segmentation core is initiated to perform block segmentation. The curvature constraint matrix is ​​used to adjust the normal vector of the segmentation surface to ensure the smoothness of the segmentation surface in the pitting and undercut regions. The generation of a flexible buffer surface is triggered by the hazard domain mask to deal with the complex morphology of the high-noise region. The segmentation surface and the flexible buffer surface are integrated to generate a continuous fractal draft. The shape parameters of each surface patch in the flexible buffer surface and the corresponding hazard domain point information are recorded to form the buffer surface parameters.

[0011] In a preferred embodiment, step S4 includes the following: The continuous parting and drafting process is decomposed into coarse and fine scales through progressive mesh subdivision. At the coarse scale, the displacement field is calculated based on the buffer surface parameters and morphological noise mapping to adjust the large-scale displacement of the cutting surface. Then, it is interpolated to the original mesh scale, and at the fine scale, the local displacement field is applied to optimize the local details based on the curvature constraint matrix. After each round of optimization, the penetration risk of the cutting surface is evaluated using the curvature constraint matrix to ensure continuity and safety. Iterative optimization is performed until the neighborhood curvature consistency index is lower than the preset threshold or the maximum number of iterations is reached. Finally, the mesh of the assembled mold block is output.

[0012] In a preferred embodiment, step S5 includes the following: The assemblable grid is imported into the virtual casting simulation platform. Casting material, temperature, and pressure parameters are set to generate pressure gradient data and deformation data of the buffer surface. The seepage boundary delay index is calculated using the pressure gradient data. By sampling multiple points along the buffer surface, the pressure gradient at each sampling point is calculated, and the delay index component at each point is obtained by integrating along the flow direction. The average value is taken as the seepage boundary delay index. The buffer elastic residual rate is calculated using the deformation data. By tracking multiple nodes on the buffer surface, the displacement and original thickness of each node are recorded, and the residual rate of each node is calculated as the ratio of displacement to original thickness. The average value is taken as the buffer elastic residual rate.

[0013] In a preferred embodiment, step S5 further includes the following: The seepage boundary delay index and buffer elastic residual rate are input into the seepage instability early warning model. Based on historical simulation and experimental data, the probability distribution under safe and unsafe conditions is estimated. Bayes' theorem is applied to calculate the posterior probability of the unsafe condition. The buffer safety factor is obtained by subtracting the posterior probability from 1.

[0014] In a preferred embodiment, step S5 further includes the following: When the buffer safety factor is lower than a predetermined threshold, multiple disturbance schemes are generated around the current buffer surface parameters. For each disturbance scheme, the simulation is rerun and the buffer safety factor is calculated. The scheme that maximizes the buffer safety factor is selected to update the buffer surface parameters. This process is iterated until the buffer safety factor reaches or exceeds the threshold. Finally, the final version block file with seepage risk control is output.

[0015] A Boolean modeling-based parting line design system for bronze casting includes: Mapping configuration module: After collecting point clouds of the object's exterior and interior cavities, it performs density adaptive resampling and implants temporal identifiers for abnormal scattered points to generate a shape noise mapping; The parsing and annotation module: Based on the topographic noise mapping, the semantic slicing network is used to annotate the pitting and undercut regions, and simultaneously generate the curvature constraint matrix and the danger zone mask and maintain the correspondence. Segmentation Auxiliary Module: Uses curvature constraint matrix to trigger Boolean segmentation core, generates flexible buffer surfaces in real time at the corresponding positions of the danger domain mask, outputs continuous fractal drafts and marks buffer surface parameters; Displacement control module: Input the buffer surface parameters and topography noise mapping into the multi-scale displacement optimization process, iteratively correct the micro-displacement of the cut surface and continuously check the penetration risk with the curvature constraint matrix to obtain the assemblable block mesh; Stability control module: Import the assemblable template mesh into the virtual casting simulation platform, extract cross-domain indicators through physical simulation to generate risk coefficients to evaluate the performance of the buffer surface, and automatically adjust the buffer surface parameters to eliminate seepage risk when the evaluation is insufficient, and output the final template file with controlled seepage risk.

[0016] The technical effects and advantages of the bronze casting parting design method and system based on Boolean modeling of this invention are as follows: This invention ensures accurate annotation of complex features such as pitting and undercutting in the early stages of modeling by constructing a linkage mechanism between morphological noise mapping and semantic slicing. Curvature constraint logic is embedded in the Boolean segmentation core to control the generation of cut surfaces, and a flexible buffer surface is introduced to accommodate irregular shapes in dangerous areas, thus achieving a continuous transition in geometric division. The fractal draft is simultaneously checked for penetration risks during multi-scale displacement optimization to ensure the assembly accuracy of the assembled blocks, and the optimization results are fed back to update the cut surface morphology. Assembleable blocks are fed back cross-domain indicators through physical simulation, triggering penetration risk judgment and buffer structure adjustment strategies. The strongest parameter response interval is obtained through local perturbation search, forming an efficient and controllable feedback adjustment path. Each processing stage constructs a full-link linkage system from scanning noise annotation to casting safety control through cross-domain mapping and parameter write-back, significantly improving fractal integrity and block stability. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the bronze casting parting design method based on Boolean modeling of the present invention. Figure 2 This is a schematic diagram of the structure of the bronze casting parting design system based on Boolean modeling of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 This invention presents a method for designing the parting line for bronze casting based on Boolean modeling, including: S1: After collecting point clouds of the object's exterior and interior, density adaptive resampling is performed, and temporal identifiers are implanted for abnormal scattered points to generate a shape noise mapping.

[0020] S2: Based on the topographic noise mapping, the semantic slicing network is used to label the pitting and undercut regions, and simultaneously generate the curvature constraint matrix and the danger domain mask and maintain the correspondence.

[0021] S3: Use the curvature constraint matrix to trigger the Boolean segmentation core, generate a flexible buffer surface in real time at the corresponding position of the danger domain mask, output a continuous fractal draft and mark the buffer surface parameters.

[0022] S4: Input the buffer surface parameters and topography noise mapping into the multi-scale displacement optimization process, iteratively correct the micro-displacement of the cut surface and continuously check the connection risk with the curvature constraint matrix to obtain the assembleable block mesh.

[0023] S5: Import the assemblable template mesh into the virtual casting simulation platform, extract cross-domain indicators through physical simulation to generate risk coefficients to evaluate the performance of the buffer surface, and automatically adjust the buffer surface parameters to eliminate seepage risk when the evaluation is insufficient, and output the final template file with controlled seepage risk.

[0024] As an important legacy of ancient Chinese civilization, the digital restoration and preservation of bronze artifacts are of profound significance for cultural inheritance. Traditional methods for bronze artifact restoration in museums and other institutions rely on manual measurement and mold making, which are inefficient and struggle to handle the complex morphologies of the artifacts' surfaces. In recent years, the introduction of laser scanning technology has made it possible to acquire high-precision point cloud data of the bronze artifacts' exterior and interior cavities. Combined with Boolean modeling technology, parting lines and mold blocks can be automatically segmented for 3D printing templates and casting replication. However, the pitting and indentation features formed by long-term oxidation of bronze artifacts, as well as the unavoidable point cloud noise during scanning, pose challenges to traditional Boolean modeling in handling these complex geometries. Unprocessed noise often leads algorithms to misjudge deep pitting as penetrable areas, generating cut surfaces that penetrate both the inner and outer walls, thus causing seepage and cracking risks during the casting stage. Therefore, there is an urgent need for a bronze artifact casting parting line design method based on Boolean modeling, which, through systematic point cloud processing and morphology optimization, ensures the accuracy of the parting line design and the safety of the casting process. This invention is designed for this purpose, constructing a full-link technical process from point cloud acquisition to block optimization through steps S1 to S5. Step S1 serves as the starting point, responsible for the initial processing of point cloud data, laying the foundation for subsequent steps.

[0025] The purpose of step S1 is to collect point cloud data of the bronze artifact's exterior and interior, perform density-adaptive resampling, implant temporal identifiers for outlier points, and generate a shape noise map. The following is a detailed description of the specific technical logic: S1.1, Point Cloud Acquisition: A high-precision laser scanner was used to comprehensively scan the exterior and interior of the bronze artifact to capture its three-dimensional geometric information. During the scanning process, the laser scanner recorded the positional data of each point on the artifact's surface, forming a raw point cloud dataset. This raw point cloud dataset contains detailed morphological information about the bronze artifact's surface; however, due to long-term oxidation and wear, the data may contain noise and outliers. These noises and outliers originate from equipment errors during the scanning process or physical defects on the artifact's surface.

[0026] S1.2, Density-Adaptive Resampling: The original point cloud dataset exhibits an uneven density distribution, with dense point clouds in detailed areas (such as pitting or inverted areas) and sparse point clouds in flat areas. To optimize the efficiency and accuracy of subsequent processing, an adaptive density resampling technique is employed to adjust the point cloud density distribution. Specifically: First, for each point in the original point cloud dataset, its local neighborhood radius is determined, defined as the maximum distance between the 20 nearest points to that point. Then, based on this local neighborhood radius, the local density is calculated, which is the number of points within a sphere centered at that point with a radius equal to the local neighborhood radius. Next, a density threshold range is set. If the local density of a point is lower than the lower limit of the density threshold range, new points are added in the neighborhood of that point through interpolation; if it is higher than the upper limit of the density threshold range, some points are deleted until the point cloud density tends to be balanced. After the adaptive density resampling process, the original point cloud dataset is transformed into a resampled point cloud dataset with a uniform density distribution, facilitating accurate identification of complex artifact morphologies in subsequent steps.

[0027] S1.3, Outlier scatter point detection: In the resampled point cloud dataset, anomalous scattered points deviating from the normal morphology are identified. These anomalous scattered points may be caused by scanning errors or surface damage to the object. The specific detection method is as follows: First, for each point in the resampled point cloud dataset, its local curvature is calculated. Local curvature is defined as the average distance from all points in the neighboring region to the fitted plane, reflecting the degree of curvature of the neighborhood. Then, the calculated local curvature is compared with a preset curvature threshold, which is determined based on the overall curvature characteristics of the object. If the local curvature of a point exceeds the curvature threshold, it is marked as a candidate anomalous scattered point. Next, the candidate anomalous scattered points are further verified by calculating the median distance between the point and all points in its neighborhood and comparing it with the global median distance of the resampled point cloud dataset. If the median distance significantly deviates from the global median distance, the point is confirmed as an anomalous scattered point. Through this anomalous scattered point detection process, a set of anomalous scattered points in the resampled point cloud dataset is identified, providing an accurate basis for subsequent processing.

[0028] S1.4, Timing Marker Implantation: Each anomalous point in the anomalous scatter plot is assigned a time-series identifier to record its temporal information during the scanning process. Specifically, the timestamp corresponding to each anomalous point is extracted from the acquisition sequence of the laser scanner, and this timestamp is embedded as a time-series identifier into the anomalous scatter plot data. The time-series identifier reflects the temporal order of the anomalous scatter plots' generation and is used to analyze the causes of anomalous scatter plots, such as external factors like equipment vibration or surface reflection. After embedding the time-series identifier, the anomalous scatter plot set contains not only location information but also temporal information, thus forming an anomalous scatter plot dataset that includes both location and temporal information, providing comprehensive data support for subsequent noise source analysis.

[0029] S1.5, Topography noise mapping generation: A topography-noise mapping is generated by combining a resampled point cloud dataset with an outlier dataset to reflect the noise distribution on the surface of bronze artifacts. The specific method is as follows: First, a three-dimensional mesh is constructed, with its resolution determined based on the average spacing of points in the resampled point cloud dataset. Then, each mesh cell is analyzed. If a mesh cell contains outliers, the noise intensity is calculated based on the local curvature of the outlier and the reciprocal of its median distance. If no outliers are found in the mesh cell, the noise intensity is recorded as zero. Finally, the noise intensity data from all mesh cells are integrated to output the topography-noise mapping, clearly demonstrating the noise distribution characteristics on the bronze artifact surface. This topography-noise mapping provides crucial prior information for the subsequent annotation and segmentation of complex topography.

[0030] In the design of bronze casting molds, the quality of point cloud data directly affects the accuracy of Boolean modeling and the casting process. Step S1 acquires the original point cloud data set through point cloud acquisition, optimizes the data distribution using density adaptive resampling, identifies noise points using outlier detection, analyzes the noise source by incorporating time-series identifiers, and finally generates a morphological noise map to display the noise distribution. By progressively improving data quality, it ensures that the resampled point cloud data set and the outlier set accurately reflect the morphological characteristics of the bronze, providing reliable data support for subsequent Boolean modeling and mold design, thereby achieving high-quality mold casting mold design and a safe casting process.

[0031] Step S1 provides a high-quality data foundation for subsequent Boolean modeling through point cloud acquisition, density adaptive resampling, anomaly detection, and the generation of a topographic noise map. The topographic noise map reflects the noise distribution on the bronze surface and provides prior information for the annotation of complex topographic features. However, features such as pitting and indentation on the bronze surface appear as highly irregular geometric shapes in the scanned data, making it difficult to accurately distinguish these features from noise using only the topographic noise map. Therefore, step S2 introduces a semantic slicing network, driven by the topographic noise map, to achieve accurate annotation of pitting and indentation areas and simultaneously generate curvature constraint matrices and danger zone masks, laying the foundation for the safety and accuracy of subsequent Boolean segmentation.

[0032] Step S2 processes the topography noise map and resampled point cloud generated in step S1, driving the semantic slicing network to label pitting and undercut regions, simultaneously generating curvature constraint matrices and hazard domain masks, and ensuring that all outputs correspond to the resampled point cloud. The following is a detailed description of the specific technical logic: S2.1, Semantic Slicing Network Driven: Using the topography noise map generated in step S1 as input, a pre-trained semantic slicing network is used to annotate the geometric features in the resampled point cloud dataset. The topography noise map is first transformed into a feature map, which contains information about the noise intensity and local curvature of each point in the resampled point cloud dataset. Noise intensity represents the degree of topography perturbation at each point in the resampled point cloud dataset, and local curvature reflects the bending characteristics of the geometry in the neighborhood of each point. The semantic slicing network adopts a multi-layer convolutional neural network architecture, extracting high-dimensional features from the feature map through multiple convolution and pooling operations. These high-dimensional features are then processed through fully connected layers to finally output the annotation results. The annotation results represent the category of each point in the resampled point cloud dataset, categorized into three types: pitting, indentation, and normal, representing the geometric feature type to which each point belongs.

[0033] The pitting and indentation features on the surface of bronze artifacts appear as highly irregular geometric shapes in scanned data, making them difficult to accurately distinguish using traditional manual methods or simple algorithms. Semantic slicing networks, utilizing deep learning techniques, learn complex geometric patterns through training, enabling them to automatically extract and identify features in resampled point cloud datasets. The annotation results generated by the semantic slicing network accurately reflect the geometric feature type of each point in the resampled point cloud dataset, providing accurate feature localization for subsequent Boolean segmentation operations and ensuring the segmentation process can be safely performed in complex geometric regions.

[0034] S2.2, Generation of curvature constraint matrix: Based on the annotation results of the semantic slicing network, the curvature change rate within the neighborhood of points labeled as pitting or indentation is calculated to generate a curvature constraint matrix. The curvature change rate is defined as the maximum difference in local curvature within the neighborhood of each point labeled as pitting or indentation, reflecting the degree of curvature fluctuation within the neighborhood. The curvature constraint matrix is ​​a diagonal matrix, with its diagonal elements determined by the following rules: for points labeled as pitting or indentation, the diagonal elements are the curvature change rate of that point; for points labeled as normal, the diagonal elements are zero. The calculation logic for generating the curvature constraint matrix is ​​as follows: first, points labeled as pitting or indentation are identified; then, the maximum difference in local curvature within the neighborhood of each point is calculated; this difference is filled into the corresponding position in the diagonal matrix, and other positions are filled with zero.

[0035] The curvature changes drastically in the pitting and undercut regions. Direct Boolean segmentation may result in discontinuous cut surfaces, affecting the assembly accuracy of the module. The curvature constraint matrix records the curvature variation characteristics of the pitting and undercut regions, providing constraints for Boolean segmentation. This matrix provides the basis for controlling the smoothness of the cut surfaces in the core Boolean segmentation process, ensuring a continuous transition in the pitting and undercut regions, thereby improving the assembly quality of the module.

[0036] S2.3, Generation of the danger zone mask: Based on the annotation results of the semantic slicing network and the topographic noise mapping, a hazard domain mask is generated to mark the parts with high noise intensity in the pitting and indentation regions. The hazard domain mask is a binary vector, whose elements indicate whether each point in the resampled point cloud dataset belongs to the hazard domain. The generation logic is as follows: for points labeled as pitting or indentation, check whether their noise intensity exceeds a preset threshold. If it does, the mask value for that point is 1, indicating that it belongs to the hazard domain; for points labeled as normal or whose noise intensity does not exceed the preset threshold, their mask value is 0. The hazard domain mask generation process is completed by comparing the noise intensity with the preset threshold point by point.

[0037] Points with high noise intensity in pitting and undercut areas are easily misjudged during Boolean segmentation, leading to the generation of cut surfaces that penetrate the inner and outer walls, increasing the risk during the casting process. The danger zone mask clearly identifies these high-noise-intensity areas, indicating the parts requiring special handling. The danger zone mask provides a location basis for the real-time generation of flexible buffer surfaces during Boolean segmentation, ensuring protective measures are taken in areas with high noise intensity and avoiding the risk of accidental cutting that could lead to penetration between the inner and outer walls.

[0038] S2.4, Maintain the correspondence: To ensure a one-to-one correspondence between each item in the semantic slicing network's annotation results, curvature constraint matrix, and hazard domain mask and the points in the resampled point cloud dataset, the following implementation method is employed: the index order of the annotation results, curvature constraint matrix, and hazard domain mask is completely consistent with the order of the points in the resampled point cloud dataset. The underlying principle is that, when generating the above data, the point order of the resampled point cloud dataset is always used as the benchmark, ensuring that for any point in the resampled point cloud dataset, its category, rate of curvature change, and hazard domain label can be directly obtained using the same index.

[0039] Step S2, based on the topography noise map generated in step S1, achieves accurate annotation of pitting and undercut regions in the resampled point cloud dataset by driving a semantic slicing network. The curvature constraint matrix generated based on the annotation results records the curvature variation characteristics of the pitting and undercut regions, providing support for Boolean segmentation to control the smoothness of the cut surface; the danger zone mask marks areas with high noise intensity, providing crucial information for the generation of flexible buffer surfaces. The annotation results, curvature constraint matrix, and danger zone mask maintain a strict correspondence with the points in the resampled point cloud dataset, ensuring that subsequent Boolean segmentation operations can safely and accurately handle complex topography regions, thereby improving the overall quality of the casting parting.

[0040] Step S2 uses a topography noise mapping-driven semantic slicing network to label pitting and undercut regions, generating a curvature constraint matrix and a danger zone mask, providing feature localization and risk control for Boolean segmentation. However, the segmentation process in Boolean modeling must ensure the continuity and safety of the cut surfaces, especially in complex areas such as pitting and undercuts, avoiding cut surfaces that penetrate the inner and outer walls. Step S3, based on this premise, uses the curvature constraint matrix and danger zone mask to achieve precise control of the segmentation surface and generate a flexible buffer surface, outputting a continuous fractal draft and related parameters to support subsequent optimization.

[0041] Step S3 takes the resampled point cloud, annotation results, curvature constraint matrix, and danger zone mask output from step S2 as input. It uses the curvature constraint matrix to trigger a Boolean segmentation core, generating a flexible buffer surface in real-time at the corresponding position of the danger zone mask. Finally, it outputs a continuous fractal draft and marks the buffer surface parameters. The detailed technical logic is as follows: S3.1, Boolean segmentation core trigger: Using the resampled point cloud dataset, annotation results, and curvature constraint matrix output from step S2 as input, the Boolean segmentation core is activated to perform block segmentation. The Boolean segmentation core employs an improved solid geometry construction algorithm, determining the position and orientation of the initial segmentation surface by analyzing the point positions and annotation results in the resampled point cloud dataset. Based on this, the normal vector of the segmentation surface is adjusted using the curvature constraint matrix. For points labeled as pitting or indentation, the corresponding curvature constraint values ​​are extracted from the curvature constraint matrix. The calculation logic for adjusting the segmentation surface normal vector is as follows: first, the initial normal vector is obtained; then, a weighted combination of the curvature constraint value and the surface gradient is calculated, where the weights are determined by the Frobenius norm of the curvature constraint matrix and a fixed adjustment coefficient; finally, the weighted combination result is added to the initial normal vector to obtain the adjusted normal vector. The surface gradient is calculated by fitting the local surface at each point in the resampled point cloud dataset. The fitting process determines the surface change trend based on the positional relationships within the point's neighborhood.

[0042] Due to the drastic curvature changes in the pitting and undercut regions, directly generating the cutting surface using traditional Boolean segmentation methods may result in uneven cut surfaces in these areas because it cannot adapt to local surface variations, leading to gaps or overlaps during mold assembly. Introducing a curvature constraint matrix to adjust the normal vector of the cutting surface allows for precise reflection of the local geometric features of the object's surface based on curvature constraint values ​​and surface gradients, enabling the cutting surface to better conform to curvature changes. Adjusting the normal vector of the cutting surface using the curvature constraint matrix ensures the smoothness and geometric consistency of the cutting surface in the pitting and undercut regions, thereby improving the assembly accuracy between molds and reducing assembly errors.

[0043] S3.2, Flexible buffer surface generated in real time: During the Boolean segmentation process, the intersection of the segmentation surface and the hazard mask is monitored in real time. Regions marked as 1 in the hazard mask represent high noise intensity areas. When the segmentation surface enters these regions, the generation mechanism of the flexible buffer surface is triggered. The flexible buffer surface consists of multiple parametric surface patches. The shape parameters of each surface patch are determined by the position of the point marked as 1 in the hazard mask and the noise intensity in the topography noise mapping. The generation logic of the surface patches is as follows: a quadratic surface model is constructed centered on the point marked as 1 in the hazard mask. The surface coefficients are determined by fitting the model to points in the resampled point cloud dataset within the neighborhood of that point. Then, the surface coefficients are modulated according to the noise intensity to adjust the surface shape. All surface patches are stitched together using boundary interpolation technology to ensure a smooth transition between adjacent surface patches, forming a continuous flexible buffer surface.

[0044] In the hazardous region mask, areas marked as 1 have high noise levels, and their point cloud data may contain significant deviations. Directly generating the cutting surface could lead to misjudgments of point locations, potentially cutting through the inner and outer walls and causing the block structure to fail. A flexible buffer surface, acting as a transition structure, adapts to the complex morphology of high-noise regions through parametric surface patches, providing additional geometric protection. Real-time generation of the flexible buffer surface effectively addresses the morphological complexity of high-noise regions, avoiding the problem of inner and outer wall penetration caused by erroneous cutting of the cutting surface, and ensuring the safety and integrity of the cutting process.

[0045] S3.3, Continuous parting draft output: The Boolean segmentation core integrates the adjusted segmentation surface with the flexible buffer surface to generate a continuous parting line draft. This continuous parting line draft consists of multiple sets of closed surfaces. The region marked as "1" in the danger zone mask is embedded with the flexible buffer surface, while the remaining regions use the standard segmentation surface. The process of embedding the flexible buffer surface involves connecting it to the standard segmentation surface in high-noise areas using boundary interpolation techniques. This ensures the continuity of the cut surface in pitting and undercut areas, while avoiding assembly gaps or through-hole problems. The integrated set of closed surfaces forms a complete continuous parting line draft, reflecting the parting line design of the object's surface.

[0046] If standard parting surfaces are used directly to process complex areas, discontinuities in the cut surfaces may occur due to curvature changes or noise interference, affecting the assembly quality of the mold blocks. Embedding flexible buffer surfaces can smoothly connect the transition between standard parting surfaces and high-noise areas, achieving geometric consistency in the cut surfaces. Continuous parting drafts, by integrating flexible buffer surfaces with standard parting surfaces, ensure the smoothness and continuity of the cut surfaces in pitting and undercut areas, improving the overall quality of the parting design and providing a reliable geometric model for subsequent casting processes.

[0047] S3.4, Buffer surface parameter marking: For each surface patch in the flexible buffer surface, its shape parameters and corresponding hazard point information are recorded. The shape parameters are the set of surface coefficients determined during the generation process of the surface patch. The buffer surface parameter set consists of the surface patch parameters corresponding to all points marked as 1 in the hazard mask. The specific method of parameter recording is as follows: a one-to-one correspondence is established between the surface coefficients of each surface patch and the points marked as 1 in the hazard mask, forming a structured parameter set for easy subsequent retrieval and analysis.

[0048] The surface coefficient of the flexible buffer surface directly determines its protective effect and geometric adaptability in high-noise areas. Recording these parameters can provide accurate reference for subsequent multi-scale displacement optimization, ensuring that the shape of the flexible buffer surface is adjusted in a targeted manner during the optimization process.

[0049] Step S3, based on the resampled point cloud data set, annotation results, curvature constraint matrix, and danger zone mask output from Step S2, performs block segmentation through a Boolean segmentation core. The curvature constraint matrix is ​​used to adjust the normal vector of the segmentation surface to ensure it adapts to the curvature changes in pitting and undercut regions. The danger zone mask triggers the generation of a flexible buffer surface to address the complex morphology of high-noise areas. The segmentation surface and the flexible buffer surface are integrated to output a continuous parting draft, while simultaneously recording the buffer surface parameter set. This process achieves smoothness, safety, and continuity of the segmentation surface in bronze casting parting design, providing a high-quality initial model and parameter support for multi-scale displacement optimization in Step S4, thereby improving the overall accuracy and stability of the casting parting design.

[0050] Step S3 takes the curvature constraint matrix and the danger zone mask as input, performs Boolean segmentation, and generates flexible buffer surfaces in the danger zone. It outputs a continuous parting line draft and buffer surface parameters, laying the initial model for subsequent optimization. However, although the continuous parting line draft achieves surface continuity in complex regions, potential assembly errors and penetration risks still exist, requiring further modification to meet the assemblability and safety requirements of the blocks. Step S4, based on this premise, performs multi-scale displacement optimization on the continuous parting line draft to eliminate penetration risks and improve assembly accuracy.

[0051] Step S4 takes the continuous fractal draft, buffer surface parameters, topography noise mapping and curvature constraint matrix output from step S3 as input, iteratively corrects the micro-displacement of the cut surface through a multi-scale displacement optimization process, and continuously checks the penetration risk with the curvature constraint matrix, finally generating an assemblable block mesh.

[0052] S4.1, Multi-scale displacement optimization process initiated: Using the continuous parting line draft, buffer surface parameters, and topography noise mapping output from step S3 as input, the multi-scale displacement optimization process is initiated. The continuous parting line draft serves as the initial geometric model, containing the preliminary continuity design of the cut surfaces; the buffer surface parameters describe the geometric characteristics of the flexible buffer surface; and the topography noise mapping provides noise distribution information for the point cloud. The multi-scale displacement optimization process decomposes the continuous parting line draft into two parts—a coarse-scale level and a fine-scale level—through progressive mesh subdivision technology. The coarse-scale level handles large-scale cut surface displacement adjustments, while the fine-scale level handles fine-tuning of local details, thereby achieving optimization control from the overall to the local level.

[0053] While continuous parting lines have achieved surface continuity in the pitting and undercut regions, there is a potential risk of assembly errors or through-wall penetration. By employing a multi-scale displacement optimization process to handle geometric features at different scales in layers, optimization can be performed separately for large-scale structural adjustments and local fine-tuning, thereby improving overall optimization efficiency and accuracy. This multi-scale displacement optimization process, through layered control, ensures both global structural adjustments and the accuracy of local details, guaranteeing stable optimization and convergent results. This provides technical support for generating high-quality, assemblable block meshes.

[0054] S4.2, Coarse-scale optimization: At the coarse-scale level, low-frequency geometric features are extracted from the continuous fractal draft to generate a coarsened mesh. The coarsened mesh simplifies the details of the continuous fractal draft while preserving key structural information. Based on this, a displacement field is calculated and applied to the coarsened mesh. The calculation logic for the displacement field is as follows: first, the buffer surface parameter norm and the smoothing factor are added together and their reciprocals are taken to obtain a scalar value. Then, this scalar value is multiplied element-wise by the displacement step size and the coarse-scale noise distribution to generate a displacement field vector of the same dimension as the coarsened mesh. Finally, this vector is projected along the direction of the cleavage normal vector to obtain the final displacement field. The displacement step size controls the magnitude of the optimization adjustment, and the smoothing factor is used to avoid anomalies in the calculation and adjust the smoothness of the displacement. The coarsened mesh is updated using the displacement field to generate the optimized coarse-scale mesh.

[0055] Continuous parting line drafts may contain large-scale geometric errors, affecting the efficiency of subsequent optimization. Coarse-scale optimization quickly corrects these errors by adjusting large-scale sectional displacements, reducing the processing burden of fine-scale optimization. The calculation of the displacement field, combined with buffer surface parameters and topographic noise mapping, ensures that the adjustment direction is consistent with the noise distribution on the object surface and the characteristics of the flexible buffer surface. Coarse-scale optimization improves the efficiency of the optimization process and reduces computational resource consumption by quickly correcting large-scale geometric errors. At the same time, the targeted design of the displacement field enhances the accuracy of the adjustment, ensuring that the optimized coarse-scale mesh meets the design requirements.

[0056] S4.3, Fine-scale optimization: The optimized coarse-scale mesh is interpolated back to the original mesh scale using mesh subdivision technology to generate an intermediate optimized mesh. The interpolation process restores the detailed information of the continuous fractal draft. Based on this, a local displacement field is applied to fine-tune the intermediate optimized mesh. The calculation logic of the local displacement field is as follows: First, the diagonal elements of the curvature constraint matrix are extracted to represent the curvature constraint strength at each point; then, the Frobenius norm of the curvature constraint matrix is ​​calculated to normalize the curvature constraint values; next, the tangential vector at each point is determined as the direction of local adjustment; the fine-tuning step size is multiplied by the normalized curvature constraint value to obtain a scalar, which is then multiplied by the tangential vector to obtain the vector value of the local displacement field at each point. This operation is repeated at all points of the mesh to generate the entire local displacement field. The fine-tuning step size is used to control the accuracy of the local adjustment. The intermediate optimized mesh is updated using the local displacement field to generate a refined optimized mesh.

[0057] Intermediate-scale optimization may result in localized misalignment of the cut surfaces in pitting and undercut areas. Fine-scale optimization fine-tunes these details using a local displacement field, precisely correcting the cut surface positions to ensure a good fit with the object's surface. The local displacement field is designed based on a curvature constraint matrix, adjusting areas with significant curvature variations to enhance the targeting of the optimization. Fine-scale optimization improves the accuracy of the cut surfaces by precisely fine-tuning local details, reducing assembly errors and the risk of penetration between inner and outer walls, ensuring that the refined optimized mesh meets the requirements of a high-quality, assemblable block mesh.

[0058] S4.4, Through-connection risk verification: After each displacement optimization, the risk of wall penetration is assessed using the curvature constraint matrix. The assessment logic is as follows: First, the difference between the curvature constraint value at each point and the average curvature constraint value in the neighborhood is calculated; then, the differences at all points are summed and averaged to obtain the neighborhood curvature consistency index. The neighborhood curvature consistency index reflects the smoothness of curvature changes in the mesh. If the neighborhood curvature consistency index is lower than a preset threshold, it indicates that the curvature changes are smooth and the risk of wall penetration is controllable; if it is higher than the preset threshold, it indicates that there is a curvature abrupt change, which may lead to wall penetration, requiring further optimization.

[0059] During optimization, changes in the curvature of the cut surface may lead to discontinuities or abrupt changes, increasing the risk of penetration between the inner and outer walls. Penetration risk verification quantifies the smoothness of the cut surface using a neighborhood curvature consistency index, ensuring that the optimized mesh maintains continuity in areas with significant curvature changes and avoiding penetration issues. Penetration risk verification provides clear evaluation criteria, ensuring the safety and reliability of the optimization process and improving the overall quality of the assemblable block mesh.

[0060] S4.5, Iterative Correction and Convergence: Repeat coarse-scale optimization and fine-scale optimization until the neighborhood curvature consistency index falls below a preset threshold or the maximum number of iterations is reached. After each iteration, the neighborhood curvature consistency index is calculated and evaluated. If the neighborhood curvature consistency index is below the preset threshold, the optimization result meets the requirements, and iteration stops; if the maximum number of iterations is reached but the requirements are not met, the current best result is used as the output. After optimization convergence, the refined optimized mesh is output as the fittable block mesh.

[0061] A single optimization may not completely eliminate geometric errors and connectivity risks. Iterative correction gradually approaches the optimal solution through multiple iterations, ensuring the stability of the optimization process and the convergence of results. Setting a maximum number of iterations prevents the optimization process from running indefinitely, improving computational efficiency. The iterative correction and convergence mechanism ensure the completeness of the optimization process, enabling the generated assemblable block mesh to meet design requirements in terms of assembly accuracy and safety.

[0062] Step S4 uses the continuous parting line draft, buffer surface parameters, and morphological noise mapping output from Step S3 as input, and systematically optimizes the continuous parting line draft through a multi-scale displacement optimization process. Coarse-scale optimization handles large-scale facet displacements, fine-scale optimization fine-tunes local details, a risk check ensures the safety of the optimization process, and an iterative correction and convergence mechanism guarantees the completeness of the results. After optimization, the output assemblable block mesh eliminates the risk of internal and external wall penetration, improves assembly accuracy, and provides a high-quality block model for the virtual casting simulation in the subsequent Step S5, ensuring the overall quality and safety of the bronze casting parting line design.

[0063] The assemblable template mesh output in step S4 already possesses assembly accuracy and geometric continuity, laying the foundation for virtual casting simulation. However, during actual casting, seepage risk and buffer surface performance directly affect the safety and stability of the template. Step S5 takes this as a starting point, evaluates seepage risk and buffer surface performance through physical simulation, and automatically adjusts the buffer surface parameters when the risk exceeds the limit, ensuring the safety and reliability of the final template file.

[0064] Step S5 takes the assemblable block mesh output from step S4 as input, performs physical simulation through a virtual casting simulation platform, extracts the seepage boundary hysteresis index and buffer elastic residual rate, and generates a buffer safety factor through a seepage instability early warning model. If the buffer safety factor is lower than a preset threshold, the buffer surface parameters are automatically adjusted, and the data is written back to the Boolean partitioning core and multi-scale displacement optimization module for synchronous optimization, finally outputting a final block file with controlled seepage risk. The specific technical logic is as follows: S5.1, Virtual Casting Simulation: The assemblable grid output from step S4 is used as input data and imported into a virtual casting simulation platform for processing. The virtual casting simulation platform simulates the actual casting process based on pre-set casting material properties, temperature conditions, and pressure parameters. Two types of data are generated through the simulation: pressure gradient data and buffer surface node deformation data. Pressure gradient data characterizes the pressure variation along the buffer surface during casting, while buffer surface node deformation data characterizes the deformation of each node on the buffer surface under thermal stress. When calculating the pressure gradient data, the virtual casting simulation platform determines the pressure change trend at each location on the buffer surface based on the pressure distribution generated by the casting material during flow. When calculating the buffer surface node deformation data, the virtual casting simulation platform records the deformation of each node on the buffer surface based on the node displacement caused by thermal stress.

[0065] Assembleable modular meshes may experience a decrease in structural stability during actual casting due to seepage and deformation effects. Virtual casting simulation allows for the evaluation of the performance of assembleable modular meshes under casting conditions before physical testing, thereby identifying potential structural risks. Virtual casting simulation provides an efficient and cost-effective risk assessment method, significantly reducing the number of physical tests and related costs, while providing necessary data support for subsequent seepage risk analysis and buffer surface performance evaluation.

[0066] S5.2, Calculation of seepage boundary hysteresis index: Pressure variation information along the buffer surface is extracted from the pressure gradient data generated by virtual casting simulation. Several sampling points are uniformly selected along the buffer surface. For each sampling point, its pressure gradient is calculated, defined as the rate of change of pressure per unit distance. Next, path integral processing is performed on the pressure gradient at each sampling point according to the flow direction of the casting material. The specific calculation method for the path integral is as follows: the dot product of the pressure gradient and the flow direction is accumulated along the flow path to obtain the hysteresis index component for each sampling point. Finally, the average value of the hysteresis index components of all sampling points is taken to obtain the seepage boundary hysteresis index, which is used to characterize the seepage inhibition ability of the buffer surface.

[0067] Seepage is a significant risk factor that can lead to the failure of assemblable mesh structures during the casting process. The seepage boundary hysteresis index, by quantifying the cumulative effect of the pressure gradient along the flow direction, can accurately assess the performance of the buffer surface in suppressing seepage. The seepage boundary hysteresis index provides a quantitative indicator for assessing the seepage risk of assemblable meshes, helping to determine the safety of assemblable meshes during the casting process and providing data for optimized design.

[0068] S5.3, Calculation of buffer elasticity residual rate: Relevant information is extracted from the deformation data of the buffer surface nodes generated by virtual casting simulation. For multiple nodes on the buffer surface, the displacement of each node under thermal stress and the original thickness at that node are recorded. The original thickness is defined as the initial thickness of the node before deformation, and the displacement is defined as the change in node position caused by thermal stress. For each node, its residual rate is calculated, which is defined as the ratio of displacement to original thickness. The average residual rate of all nodes is taken to obtain the buffer elastic residual rate, which is used to characterize the deformation recovery capability of the buffer surface.

[0069] The deformation of the buffer surface due to thermal stress during casting can affect its structural integrity. The buffer elastic residual rate, by quantifying the relative proportion of deformation to the original thickness, can effectively assess the elastic recovery performance of the buffer surface. The buffer elastic residual rate provides a quantitative indicator for evaluating the structural stability of assemblable meshes during casting, helping to determine the thermal shock resistance of the assemblable meshes and providing a reference for design improvements.

[0070] S5.4, Early warning model for osmotic instability: A seepage instability early warning model is constructed using the seepage boundary hysteresis index and buffer elastic residual rate as input data, combined with historical simulation data and experimental data. Historical data includes the probability distribution information of the seepage boundary hysteresis index and buffer elastic residual rate under safe and unsafe conditions. Using probabilistic calculation methods, the probability of an unsafe state occurring given the seepage boundary hysteresis index and buffer elastic residual rate is calculated. The probability of an unsafe state occurring is determined by the ratio of the joint conditional probability, prior probability, and normalization constant. The buffer safety factor is defined as 1 minus the probability of an unsafe state occurring, used to characterize the safety of the assemblable mesh in the current state.

[0071] The probability of an unsafe state occurring is calculated using Bayes' theorem. Specifically, it involves multiplying the likelihood of the observed indicators under unsafe conditions (i.e., the probability of observing these indicators under unsafe conditions) by the prior probability of the unsafe state (i.e., the initial probability of the unsafe state before considering any indicators), and then dividing by the total probability of observing these indicators (i.e., the probability of observing these indicators in all possible states). This calculation utilizes the Bayesian framework to update beliefs about unsafe states based on the evidence provided by the indicators. The likelihood reflects the probability of observing the indicators under unsafe conditions, the prior probability reflects the initial belief before considering the evidence, and the total probability serves as a normalization constant, ensuring that the sum of the posterior probabilities in all possible states is 1. In this way, the calculated posterior probability of an unsafe state can more accurately assess the risk of unsafety under a given indicator.

[0072] A comprehensive assessment of seepage phenomena and deformation risks requires consideration of multiple indicators simultaneously. The seepage instability early warning model combines the seepage boundary hysteresis index and buffer elastic residual rate with historical data using probabilistic calculation methods, providing a probability-based safety assessment approach. This model offers a quantitative indicator for the safety assessment of assemblable modular meshes, enabling the identification of potential risks before casting and ensuring the structural stability of the assemblable modular meshes and the safety of the casting process.

[0073] S5.5, Buffer surface parameters are automatically adjusted: Using the buffer safety factor and the current buffer surface parameters as input data, a predetermined safety threshold is set for comparison. If the buffer safety factor is lower than the predetermined safety threshold, multiple perturbation schemes are generated within the neighborhood of the current buffer surface parameters. Each perturbation scheme is obtained by superimposing a small perturbation vector on the current buffer surface parameters. For each perturbation scheme, the processing flow of virtual casting simulation, seepage boundary hysteresis index calculation, buffer elastic residual rate calculation, and seepage instability early warning model is re-executed sequentially to obtain a new buffer safety factor corresponding to each perturbation scheme. The scheme that increases the buffer safety factor the fastest is selected from all perturbation schemes and determined as the optimal buffer surface parameters. The optimal buffer surface parameters are then written back to the Boolean partitioning core and multi-scale displacement optimization module to synchronously update the assemblable block mesh. The above adjustment process is repeated until the buffer safety factor reaches or exceeds the predetermined safety threshold. If the buffer safety factor has reached or exceeded the predetermined safety threshold, the adjustment process is terminated, and the final version of the assemblable block mesh file is output.

[0074] When the buffer safety factor falls below a predetermined safety threshold, the assemblable mesh faces a risk of seepage, necessitating adjustments to the buffer surface parameters to enhance its safety. An automatic buffer surface parameter adjustment mechanism optimizes these parameters by generating perturbation schemes and evaluating their impact on the buffer safety factor. This mechanism, through a closed-loop feedback process, ensures the assemblable mesh meets design requirements, improving its structural stability and casting process safety while reducing the need for manual intervention and increasing design efficiency.

[0075] Step S5 uses the assemblable grid output from step S4 as input data to evaluate its seepage risk and buffer surface performance through virtual casting simulation. The seepage boundary hysteresis index is used to quantify the buffer surface's ability to suppress seepage, and the buffer elastic residual rate is used to quantify the buffer surface's deformation recovery ability. The seepage instability early warning model integrates the seepage boundary hysteresis index and the buffer elastic residual rate to evaluate the safety of the assemblable grid. If the buffer safety factor is lower than a predetermined safety threshold, the buffer surface parameters are optimized through an automatic adjustment mechanism, and the optimized parameters are written back to the Boolean partitioning core and multi-scale displacement optimization module to ensure that the final version of the assemblable grid file meets the safety requirements. This process forms a closed-loop feedback system from virtual simulation to parameter optimization, effectively improving the overall quality and safety of bronze casting parting design.

[0076] Example 2: Figure 2 This invention presents a bronze casting parting design system based on Boolean modeling, comprising: Mapping configuration module: After collecting point clouds of the object's exterior and interior, it performs density adaptive resampling and implants temporal identifiers for abnormal scattered points to generate a shape noise mapping.

[0077] The parsing and annotation module: Based on the topographic noise mapping, the semantic slicing network is used to annotate the pitting and undercut regions, and simultaneously generate the curvature constraint matrix and the danger domain mask and maintain the correspondence.

[0078] Segmentation Auxiliary Module: Uses curvature constraint matrix to trigger Boolean segmentation core, generates flexible buffer surfaces in real time at the corresponding positions of the danger domain mask, outputs continuous fractal drafts and marks buffer surface parameters.

[0079] Displacement control module: Input the buffer surface parameters and topography noise mapping into the multi-scale displacement optimization process, iteratively correct the micro-displacement of the cut surface and continuously check the connection risk with the curvature constraint matrix to obtain the assembly block mesh.

[0080] Stability control module: Import the assemblable template mesh into the virtual casting simulation platform, extract cross-domain indicators through physical simulation to generate risk coefficients to evaluate the performance of the buffer surface, and automatically adjust the buffer surface parameters to eliminate seepage risk when the evaluation is insufficient, and output the final template file with controlled seepage risk.

[0081] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0082] It should be noted that the system of the present invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting a variety of hardware environments and usage requirements.

[0083] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0084] It should be noted that, in this document, the use of relational terms such as "first" and "second" is merely to distinguish one entity or operation from another, and does not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0085] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for designing parting lines for bronze casting based on Boolean modeling, characterized in that, Including the following steps: S1: After collecting point clouds of the object's exterior and interior, density adaptive resampling is performed, and temporal identifiers are implanted for abnormal scattered points to generate a shape noise mapping. S2: Based on the topography noise mapping, the semantic slicing network is used to label the pitting and undercut regions, and simultaneously generate the curvature constraint matrix and the danger domain mask while maintaining the corresponding relationship. S3: Use the curvature constraint matrix to trigger the Boolean segmentation core, generate a flexible buffer surface in real time at the corresponding position of the danger domain mask, output a continuous fractal draft and mark the buffer surface parameters; S4: Input the buffer surface parameters and topology noise mapping into the multi-scale displacement optimization process, iteratively correct the micro-displacement of the cut surface and continuously check the connection risk with the curvature constraint matrix to obtain the assembleable block mesh. S5: Import the assemblable template mesh into the virtual casting simulation platform, extract cross-domain indicators through physical simulation to generate risk coefficients to evaluate the performance of the buffer surface, and automatically adjust the buffer surface parameters to eliminate seepage risk when the evaluation is insufficient, and output the final template file with controlled seepage risk.

2. The bronze casting parting design method based on Boolean modeling according to claim 1, characterized in that, Step S1 includes the following: The original point cloud data set of the bronze artifact's exterior and interior was collected using a laser scanner, and the point cloud density distribution of the original point cloud data set was adjusted using density adaptive resampling technology to generate a resampled point cloud data set.

3. The bronze casting parting design method based on Boolean modeling according to claim 2, characterized in that, Step S1 also includes the following: Anomaly point sets are identified based on the local curvature and median distance in the resampled point cloud dataset, and each anomaly point in the dataset is assigned a temporal identifier to record the time information during the scanning process. A three-dimensional mesh is constructed by combining the resampled point cloud dataset and the anomaly point set, and the noise intensity within the mesh cells in the three-dimensional mesh is calculated to generate a shape noise mapping that reflects the noise distribution characteristics of the bronze surface.

4. The bronze casting parting design method based on Boolean modeling according to claim 3, characterized in that, Step S2 includes the following: Using shape noise mapping as input, a pre-trained semantic slicing network is driven to annotate the geometric features in the resampled point cloud dataset and generate annotation results; Based on the annotation results, calculate the rate of curvature change in the neighborhood of the points marked as pitting or undercut, and generate a curvature constraint matrix; Based on the annotation results and morphological noise mapping, a hazardous area mask is generated to mark the parts of pitting corrosion and undercut regions where the noise intensity is higher than a preset threshold. Ensure that each item in the annotation results, curvature constraint matrix, and danger domain mask maintains a one-to-one correspondence with the points in the resampled point cloud dataset.

5. The bronze casting parting design method based on Boolean modeling according to claim 4, characterized in that, Step S3 includes the following: Using the resampled point cloud dataset, annotation results, curvature constraint matrix, and hazard domain mask as input, the Boolean segmentation core is initiated to perform block segmentation. The curvature constraint matrix is ​​used to adjust the normal vector of the segmentation surface to ensure the smoothness of the segmentation surface in the pitting and undercut regions. The generation of a flexible buffer surface is triggered by the hazard domain mask to deal with the complex morphology of the high-noise region. The segmentation surface and the flexible buffer surface are integrated to generate a continuous fractal draft. The shape parameters of each surface patch in the flexible buffer surface and the corresponding hazard domain point information are recorded to form the buffer surface parameters.

6. The bronze casting parting design method based on Boolean modeling according to claim 5, characterized in that, Step S4 includes the following: The continuous parting and drafting process is decomposed into coarse and fine scales through progressive mesh subdivision. At the coarse scale, the displacement field is calculated based on the buffer surface parameters and morphological noise mapping to adjust the large-scale displacement of the cutting surface. Then, it is interpolated to the original mesh scale, and at the fine scale, the local displacement field is applied to optimize the local details based on the curvature constraint matrix. After each round of optimization, the penetration risk of the cutting surface is evaluated using the curvature constraint matrix to ensure continuity and safety. Iterative optimization is performed until the neighborhood curvature consistency index is lower than the preset threshold or the maximum number of iterations is reached. Finally, the mesh of the assembled mold block is output.

7. The bronze casting parting design method based on Boolean modeling according to claim 6, characterized in that, Step S5 includes the following: The assemblable grid is imported into the virtual casting simulation platform. Casting material, temperature, and pressure parameters are set to generate pressure gradient data and deformation data of the buffer surface. The seepage boundary delay index is calculated using the pressure gradient data. By sampling multiple points along the buffer surface, the pressure gradient at each sampling point is calculated, and the delay index component at each point is obtained by integrating along the flow direction. The average value is taken as the seepage boundary delay index. The buffer elastic residual rate is calculated using the deformation data. By tracking multiple nodes on the buffer surface, the displacement and original thickness of each node are recorded, and the residual rate of each node is calculated as the ratio of displacement to original thickness. The average value is taken as the buffer elastic residual rate.

8. The bronze casting parting design method based on Boolean modeling according to claim 7, characterized in that, Step S5 also includes the following: The seepage boundary delay index and buffer elastic residual rate are input into the seepage instability early warning model. Based on historical simulation and experimental data, the probability distribution under safe and unsafe conditions is estimated. Bayes' theorem is applied to calculate the posterior probability of the unsafe condition. The buffer safety factor is obtained by subtracting the posterior probability from 1.

9. The bronze casting parting design method based on Boolean modeling according to claim 8, characterized in that, Step S5 also includes the following: When the buffer safety factor is lower than a predetermined threshold, multiple disturbance schemes are generated around the current buffer surface parameters. For each disturbance scheme, the simulation is rerun and the buffer safety factor is calculated. The scheme that maximizes the buffer safety factor is selected to update the buffer surface parameters. This process is iterated until the buffer safety factor reaches or exceeds the threshold. Finally, the final version block file with seepage risk control is output.

10. A bronze casting parting line design system based on Boolean modeling, used to implement the bronze casting parting line design method based on Boolean modeling as described in any one of claims 1-9, characterized in that, include: Mapping configuration module: After collecting point clouds of the object's exterior and interior cavities, it performs density adaptive resampling and implants temporal identifiers for abnormal scattered points to generate a shape noise mapping; The parsing and annotation module: Based on the topographic noise mapping, the semantic slicing network is used to annotate the pitting and undercut regions, and simultaneously generate the curvature constraint matrix and the danger zone mask and maintain the correspondence. Segmentation Auxiliary Module: Uses curvature constraint matrix to trigger Boolean segmentation core, generates flexible buffer surfaces in real time at the corresponding positions of the danger domain mask, outputs continuous fractal drafts and marks buffer surface parameters; Displacement control module: Input the buffer surface parameters and topography noise mapping into the multi-scale displacement optimization process, iteratively correct the micro-displacement of the cut surface and continuously check the penetration risk with the curvature constraint matrix to obtain the assemblable block mesh; Stability control module: Import the assemblable template mesh into the virtual casting simulation platform, extract cross-domain indicators through physical simulation to generate risk coefficients to evaluate the performance of the buffer surface, and automatically adjust the buffer surface parameters to eliminate seepage risk when the evaluation is insufficient, and output the final template file with controlled seepage risk.

Images