Automatic design method and system for special-shaped facing structure based on parametric modeling

Abstract

The invention relates to the technical field of building design and digitization, and discloses an automatic design method for a special-shaped facing structure based on parametric modeling. The method comprises the following steps: firstly, acquiring a geometrical characteristic parameter set of the special-shaped facing structure, wherein the geometrical characteristic parameter set comprises curved surface curvature distribution, boundary constraint conditions and a material attribute threshold value; thirdly, according to the curved surface curvature distribution, generating an initial geometric topological grid by using a dynamic subdivision algorithm, adjusting node positions according to boundary constraint conditions to obtain an optimized structural framework, and performing layered mapping on material attribute thresholds to generate a material distribution map; then, the input parametric modeling engine generates a three-dimensional parametric model, finally, a self-adaptive adjustment strategy graph is constructed through a multi-objective optimization algorithm, and a final design scheme is output. By means of the method, automation of design of the special-shaped facing structure is achieved, design efficiency and accuracy are improved, material utilization is optimized, scheme adaptability is enhanced, and design process collaboration is improved.

Description

Technical Field

[0001] The present invention relates to the field of architectural design and digital technology, and in particular to a method and system for automated design of special-shaped decorative structures based on parametric modeling. Background Art

[0002] In the field of modern architectural design, with the continuous improvement of people's aesthetic level and the continuous innovation of architectural concepts, special-shaped buildings have gradually become an important direction of architectural design. As a key component of special-shaped buildings, the design of special-shaped decorative structures faces many challenges.

[0003] Traditional architectural design methods have obvious limitations when dealing with special-shaped decorative structures. On the one hand, special-shaped decorative structures have complex geometric shapes, their surface curvature varies widely, and local concave-convex features and continuity indicators are difficult to accurately control. Designers relying on manual drawing or simple two-dimensional design software are unable to express these complex geometric features efficiently and accurately. They often need to spend a lot of time on repeated sketching and modification, and the design efficiency is extremely low. On the other hand, the boundary constraints of special-shaped decorative structures are complex, including the spatial connection relationship with the surrounding building structures and the load distribution information they bear. In traditional design, it is difficult to comprehensively and systematically consider the impact of these constraints on the structure, which can easily lead to problems such as structural instability and unreasonable connections in the actual construction or use of the design scheme.

[0004] From the perspective of material selection and application, different custom-shaped veneer structures have specific requirements for material properties, such as strength, durability, and aesthetics. However, traditional design methods struggle to rationally design material distribution based on material property thresholds. Designers often rely on experience to select materials, failing to fully utilize their performance advantages and potentially resulting in material waste and increased costs.

[0005] In terms of design process coherence and collaboration, traditional design models rely on independent design processes, lacking effective information sharing and integration mechanisms. This creates a disconnect between geometric design, structural design, and material design, leading to low overall design quality and significant costly modifications.

[0006] Furthermore, as construction projects continue to grow in scale and complexity, traditional design methods are unable to meet the demands of rapid design and multiple alternative comparisons. The limitations of traditional design methods become even more pronounced when faced with urgent projects or when multiple design options need to be provided to clients. Summary of the Invention

[0007] The purpose of the present invention is to provide an automated design method for special-shaped veneer structures based on parametric modeling to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] An automated design method for a special-shaped veneer structure based on parametric modeling, the method comprising:

[0010] Acquire a set of geometric characteristic parameters of the special-shaped decorative surface structure; the geometric characteristic parameters include surface curvature distribution, boundary constraints, and material property thresholds; the surface curvature distribution includes local concave-convex features and continuity indicators, and the boundary constraints include spatial connection relationships and load distribution information;

[0011] Based on the surface curvature distribution, an initial geometric topology mesh is generated by a dynamic subdivision algorithm, wherein the geometric topology mesh includes vertex density distribution and facet connection rules;

[0012] According to the boundary constraints, a constrained optimization model is used to adjust the node positions of the geometric topology grid to generate an optimized structural skeleton;

[0013] Performing layered mapping processing on the material attribute thresholds to generate a multi-level material distribution map;

[0014] Inputting the structural skeleton, geometric topological grid and multi-level material distribution map into a parametric modeling engine to generate a three-dimensional parametric model of the special-shaped veneer structure;

[0015] Based on the three-dimensional parameterized model, an adaptive adjustment strategy diagram is constructed through a multi-objective optimization algorithm to output a final design solution; the nodes of the adaptive adjustment strategy diagram represent design parameter modules, and the edges represent parameter adjustment sequences and optimization weight coefficients.

[0016] Preferably, generating the initial geometric topology mesh by a dynamic subdivision algorithm comprises:

[0017] Performing feature region segmentation on the surface curvature distribution to extract boundary markers of high curvature regions and low curvature regions;

[0018] Based on the preset subdivision rule library, the local grid density gradient is generated by adaptive subdivision times calculation;

[0019] The surface fitting algorithm is used to interpolate and encrypt the mesh vertices in the high curvature area, while keeping the mesh sparse in the low curvature area.

[0020] The grid density gradient and the patch connection rule are encoded as initial parameters of the geometric topology grid.

[0021] Preferably, the adjusting the node positions of the geometric topology grid using the constrained optimization model includes:

[0022] Performing spatial coordinate system transformation on the boundary constraint conditions to generate node displacement constraint equations;

[0023] The node position adjustment is iteratively calculated based on the gradient descent algorithm, and the Lagrange multiplier balance constraint and objective function are introduced;

[0024] Update the patch connection rules according to the adjusted node positions and verify the continuity indicators of the geometric topology mesh;

[0025] The geometric topological mesh that meets the continuity index is marked as the optimized structural skeleton.

[0026] Preferably, the parametric modeling engine includes a grid fusion module and a material mapping module, and the grid fusion module includes:

[0027] Performing parameterized encoding on the node positions of the structural skeleton to generate a first fusion vector;

[0028] Discretizing the patch connection rule of the geometric topological grid to generate a second fusion vector;

[0029] Performing feature alignment calculation on the multi-level material distribution map, extracting material distribution gradient features, and generating a third fusion vector;

[0030] The first fusion vector, the second fusion vector, and the third fusion vector are merged into a feature tensor of the three-dimensional parameterized model through a tensor fusion layer.

[0031] Preferably, the material mapping module includes:

[0032] Normalize the spatial dimension of the feature tensor to generate the material correlation matrix;

[0033] The association weights between material distribution and geometric features are extracted through convolutional neural networks to generate a weight mapping matrix;

[0034] Perform matrix dot multiplication on the material association matrix and the weight mapping matrix to generate the fused material distribution features;

[0035] The fused material distribution features are superimposed on the original feature tensor through residual connection, and the complete parameter sequence of the three-dimensional parametric model is output.

[0036] Preferably, the constructing of the adaptive adjustment strategy graph by a multi-objective optimization algorithm includes:

[0037] Initialize node attributes according to the design parameter module and generate edge weight matrix based on the optimized weight coefficient;

[0038] The parameter sequence of the three-dimensional parameterized model is used as the node state, and the edge weight matrix is composed of the time step of parameter adjustment and the optimization weight coefficient;

[0039] Iteratively update the multi-objective weights of each node through the Pareto frontier algorithm and adjust the edge weight matrix;

[0040] Generate the optimal parameter adjustment sequence covering all nodes according to the adjusted edge weight matrix.

[0041] Preferably, the method for constructing the segmentation rule base includes:

[0042] Collect standard subdivision samples of various typical surface types and extract the benchmark subdivision times and grid density distribution characteristics;

[0043] Perform curvature sensitivity analysis on the benchmark subdivision times to generate a multi-resolution subdivision model;

[0044] Classify subdivision models according to surface type and associate with standard parameter database;

[0045] The classified segmentation model is stored as a segmentation rule base, and the model is regularly updated based on newly collected samples.

[0046] Preferably, the parameter optimization method of the gradient descent algorithm includes:

[0047] Calculate the initial learning rate and momentum factor based on the historical grid-adjusted data distribution;

[0048] Through the grid convergence test, the parameter combinations are traversed and the parameters with the best balance between convergence speed and accuracy are selected;

[0049] Dynamically adjust the learning rate and momentum factor according to the balance degree to optimize the efficiency of node position adjustment.

[0050] Preferably, the weight updating method of the Pareto front algorithm includes:

[0051] The optimization weight between nodes is defined as the balance factor between the time step and the objective function weight;

[0052] Initialize the multi-objective weight of each node to zero and the starting weight to the preset initial value;

[0053] The non-inferior solution set is screened by iterative dominance relationship, and the cumulative optimization weight of each node is calculated;

[0054] The complete parameter adjustment sequence is generated by forward deduction based on the non-inferior solution set.

[0055] The present invention also includes an automated design system for special-shaped veneer structures based on parametric modeling, the system comprising:

[0056] Geometric parameter acquisition module: used to obtain a set of geometric characteristic parameters of the special-shaped decorative surface structure, the geometric characteristic parameters including surface curvature distribution, boundary constraints and material property thresholds; wherein the surface curvature distribution includes local concave-convex features and continuity indicators, and the boundary constraints include spatial connection relationships and load distribution information;

[0057] A topological mesh generation module is configured to generate an initial geometric topological mesh based on the surface curvature distribution through a dynamic subdivision algorithm, wherein the geometric topological mesh includes vertex density distribution and facet connection rules;

[0058] Structural optimization module: used to adjust the node positions of the geometric topological grid using the constraint optimization model according to the boundary constraint conditions to generate an optimized structural skeleton;

[0059] Material mapping module: performs layered mapping processing on the material attribute thresholds to generate a multi-level material distribution map;

[0060] Parametric modeling module: inputting the structural skeleton, geometric topological grid and multi-level material distribution map into the parametric modeling engine to generate a three-dimensional parametric model of the special-shaped veneer structure;

[0061] Optimization strategy generation module: used to construct an adaptive adjustment strategy diagram based on the three-dimensional parametric model through a multi-objective optimization algorithm and output the final design solution; the nodes of the adaptive adjustment strategy diagram represent the design parameter modules, and the edges represent the parameter adjustment sequence and optimization weight coefficients.

[0062] Compared with the prior art, the present invention has the following beneficial effects:

[0063] The automated design method for special-shaped decorative structures based on parametric modeling proposed in the present invention has brought significant improvements and positive benefits in many aspects to the field of architectural design. In terms of design efficiency, this method realizes the automation of the entire process from obtaining geometric feature parameters to outputting the final design scheme. By obtaining a set of geometric feature parameters of the special-shaped decorative structure and using a dynamic subdivision algorithm to quickly generate an initial geometric topological mesh, the design cycle is greatly shortened compared to traditional manual drawing and simple two-dimensional design software operations. For example, when dealing with complex hyperbolic special-shaped decorative structures, traditional design may take weeks or even months to complete the preliminary geometric design. However, with this method, only the relevant geometric feature parameters need to be input, and the initial topological mesh can be generated in a short time, which improves the design efficiency by several times or even dozens of times. At the same time, the subsequent constraint optimization, material mapping, parametric modeling, and multi-objective optimization links are closely connected and run automatically, avoiding the tedious processes and repetitive work caused by manual intervention, further improving the overall design efficiency.

[0064] From the perspective of design accuracy, the present invention can accurately handle the complex geometric features and boundary constraints of special-shaped veneer structures. By performing a detailed analysis of the surface curvature distribution, extracting local concave-convex features and continuity indicators, the generated initial geometric topological mesh can accurately reflect the geometric form of the special-shaped veneer. During the constraint optimization process, the boundary constraints are transformed into a spatial coordinate system, the node position adjustment amount is iteratively calculated using a gradient descent algorithm, and the Lagrange multiplier is introduced to balance the constraints and the objective function, ensuring that the optimized structural skeleton maintains good mechanical properties and geometric continuity while satisfying the boundary constraints. This makes the design solution safer and more reliable in practical applications, effectively reducing structural hazards and construction problems caused by inaccurate design.

[0065] In terms of material utilization, this method maps material property thresholds in layers to generate a multi-level material distribution map, enabling rational material selection and precise distribution. By extracting the weights associated with material distribution and geometric features through a convolutional neural network, it is possible to precisely configure materials based on factors such as the stress conditions and aesthetic requirements of different parts of the special-shaped decorative structure, fully leveraging the material's performance advantages. This not only improves the quality and performance of the building, but also effectively reduces material costs and waste. For example, in areas with high strength requirements, high-strength materials are selected and their distribution ratio is increased; in surface areas with high aesthetic requirements, materials with good decorative effects are selected, achieving a perfect match between material performance and design requirements.

[0066] To optimize and adapt the design, a multi-objective optimization algorithm was used to construct an adaptive adjustment strategy diagram based on a three-dimensional parametric model. This strategy diagram comprehensively considers multiple design objectives, such as structural performance, material cost, and construction difficulty. Using a Pareto frontier algorithm, the multi-objective weights of each node were iteratively updated to generate an optimal parameter adjustment sequence covering all nodes. This allows designers to quickly obtain a variety of design solutions that meet different requirements and flexibly adjust and optimize them based on actual project conditions, improving the adaptability and competitiveness of the design solutions.

[0067] This invention also promotes collaboration and information sharing within the architectural design process. Data exchange and processing across all design steps is based on a unified parametric model. Modules such as geometric parameter acquisition, topological mesh generation, structural optimization, material mapping, and parametric modeling are seamlessly integrated, eliminating the disconnected nature of traditional design models. This collaborative design model helps improve design team efficiency, reduce communication costs, and enhance the overall quality of design solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a working principle diagram of the automated design method for special-shaped veneer structures based on parametric modeling according to the present invention;

[0069] Figure 2 Flowchart for adjusting node positions for constrained optimization models;

[0070] Figure 3 This is the flow chart of the mesh fusion module of the parametric modeling engine;

[0071] Figure 4 A flowchart built for the segmentation rule base. DETAILED DESCRIPTION

[0072] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0073] See also Figure 1-Figure 4 The present invention provides an automated design method for special-shaped decorative structures based on parametric modeling, and its specific implementation method will be described in detail below.

[0074] The geometric feature parameters that need to be obtained include surface curvature distribution, boundary constraints and material property thresholds. Among them, the surface curvature distribution includes local concave-convex features and continuity indicators. These indicators can reflect the bending characteristics of the surface of the special-shaped decorative structure, such as whether the local area is convex or concave, and the smoothness of the surface transition at different positions. The boundary constraints include spatial connection relationships and load distribution information. The spatial connection relationship reflects the connection method between the special-shaped decorative structure and the surrounding building structure or components in space. The load distribution information indicates the distribution of various forces acting on the special-shaped decorative surface, such as the distribution law of wind load, self-weight load, etc. on the decorative surface. The material property threshold specifies the performance limits that the materials used to make special-shaped decorative structures must meet, such as the strength threshold of the material, the elastic modulus threshold, etc.

[0075] Using a dynamic subdivision algorithm, an initial geometric topology mesh is generated based on the surface curvature distribution. This geometric topology mesh includes vertex density distribution and facet connection rules. The vertex density distribution varies according to the surface curvature, with vertices denser in areas of greater curvature to better fit complex surface shapes. The facet connection rules determine the connection relationship between each facet, ensuring the integrity and stability of the entire mesh structure.

[0076] Based on the acquired boundary constraints, the constrained optimization model adjusts the node positions of the geometric topology mesh to generate an optimized structural skeleton. During this adjustment process, the spatial connectivity and load distribution information in the boundary constraints are fully considered, ensuring that the generated structural skeleton not only meets practical requirements but also has better mechanical properties and stability.

[0077] Material property thresholds are mapped layer by layer to generate a multi-level material distribution map. This map shows the appropriate distribution of materials within the irregular veneer structure at different levels based on the material property thresholds. For example, in certain areas with high stress, higher-strength materials are selected to improve the reliability of the entire structure.

[0078] The structural skeleton, geometric topological mesh, and multi-level material distribution maps are input into the parametric modeling engine. After a series of internal processing, a 3D parametric model of the special-shaped veneer structure is generated. This model can intuitively display the exterior shape, internal structure, and material distribution of the special-shaped veneer structure, facilitating observation and analysis by designers.

[0079] Based on the generated 3D parametric model, an adaptive adjustment strategy graph was constructed using a multi-objective optimization algorithm. The nodes of the adaptive adjustment strategy graph represent design parameter modules, while the edges represent the parameter adjustment sequence and optimization weight coefficients. Through analysis and calculation of the strategy graph, a final design solution was output. This solution comprehensively considers multiple design objectives, such as structural performance, material cost, and construction difficulty, achieving overall optimization results.

[0080] The technical solution of the present invention is further described in detail below with reference to specific embodiments.

[0081] Example 1:

[0082] When generating the initial geometric topological mesh, the surface curvature distribution is first segmented into characteristic regions. Using image processing techniques and mathematical algorithms, the boundaries of high and low curvature regions are identified. For example, using a segmentation method based on curvature thresholds, when the curvature value of a surface point exceeds a set high curvature threshold, the area containing that point is initially identified as a high curvature region; when the curvature value is less than a set low curvature threshold, it is identified as a low curvature region. In this way, regions with different curvature characteristics are accurately demarcated.

[0083] Based on the preset subdivision rule library, adaptive subdivision times are calculated. The subdivision rule library is constructed by collecting standard subdivision samples of various typical surface types. When collecting samples, common surface types such as spheres, cylinders, and hyperboloids are covered, and the baseline subdivision times and mesh density distribution characteristics of each sample are extracted. Then, a curvature sensitivity analysis is performed on the baseline subdivision times to generate a multi-resolution subdivision model. For example, for spherical samples, different subdivision times are set in different curvature areas, and the fitting effect of the mesh on the spherical shape after subdivision is observed, so as to determine the relationship between different curvatures and subdivision times. The subdivision models are classified according to the surface type, and associated with the standard parameter database, and the classified subdivision models are stored as a subdivision rule library. In practical applications, according to the surface type of the current special-shaped finishing structure, a suitable subdivision model is selected from the subdivision rule library to calculate the local mesh density gradient.

[0084] A surface fitting algorithm is used to interpolate and encrypt mesh vertices in high-curvature areas. Taking the cubic spline interpolation algorithm as an example, given the coordinates of several vertices on the boundary of a high-curvature region, cubic spline interpolation is used to calculate the coordinates of the vertices that need to be encrypted within the boundary, allowing the encrypted mesh to more accurately fit the surface shape of the high-curvature region. At the same time, the mesh distribution in low-curvature areas is kept sparse, avoiding unnecessary increases in mesh density and improving computational efficiency. Finally, the mesh density gradient and patch connection rules are encoded as the initial parameters of the geometric topological mesh, completing the generation of the initial geometric topological mesh.

[0085] Example 2:

[0086] Detailed implementation of the constrained optimization model to adjust the positions of geometric topology mesh nodes.

[0087] First, the boundary constraints are transformed into a spatial coordinate system. Assuming that the special-shaped decorative structure is in a three-dimensional spatial coordinate system, the spatial connection relationship and load distribution information in the boundary constraints are more complicated to express in the original coordinate system. Through the coordinate transformation matrix T, the boundary constraints are transformed into a coordinate system that is easier to calculate. The point coordinates (x, y, z) involved in the spatial connection relationship are transformed by T to obtain the new coordinates (x ′ ,y ′ ,z ′ ), and generate node displacement constraint equations based on these new coordinates. For example, if the connection relationship between a node and an adjacent structure requires that its displacement in a certain direction is zero, the corresponding displacement constraint equation u can be established i =0(u i represents the displacement of the i-th node in a specific direction).

[0088] The node position adjustment is iteratively calculated based on the gradient descent algorithm. The initial learning rate α0 and momentum factor β are calculated based on the historical grid adjustment data distribution. Different combinations of α0 and β are traversed through the grid convergence test, and the parameters with the highest balance between convergence speed and accuracy are selected. In the iterative calculation process, the objective function J(θ) (θ represents the node position parameter) is defined, and the gradient of the objective function with respect to the node position parameter is calculated. According to the formula (θ k represents the node position parameter at the kth iteration, and α is the dynamically adjusted learning rate) to update the node position. At the same time, the Lagrangian multiplier λ is introduced to balance the constraints and the objective function to construct the Lagrangian function (g i (θ) represents the i-th constraint function, and m is the number of constraints).

[0089] Update the patch connection rules based on the adjusted node positions. Check whether the updated patch connections meet the continuity index requirements of the geometric topology mesh. Continuity can be measured by calculating the angle deviation between adjacent patches and the length deviation of common edges. If the angle deviation is within the allowable range and the common edge length deviation meets the set standards, the geometric topology mesh is considered to have good continuity. The geometric topology mesh that meets the continuity index is marked as the optimized structural skeleton to ensure that the structural skeleton meets the design requirements in terms of mechanical properties and geometric shape.

[0090] Example 3:

[0091] In the mesh fusion module, the node positions of the structural skeleton are parameterized. Assume that there are n nodes in the structural skeleton and the position coordinates of each node are (x i ,y i ,z i )(i=1,2,…,n), convert the coordinates of each node into a vector representation using a specific encoding function E(x,y,z) to generate the first fused vector v1. For example, one-hot encoding can be used to encode nodes based on their spatial location range, so that the encoded vector of each node uniquely identifies its location information.

[0092] Discretize the patch connection rules of the geometric topological mesh. A geometric topological mesh consists of multiple patches, each connected to other patches by edges. Convert the patch connection rules into a matrix, where the elements in the matrix represent whether there is a connection relationship between the patches and the type of connection. Discretize this matrix and convert it into a vector representation to generate the second fusion vector v2. For example, if patch A and patch B are connected by an edge, the corresponding position in the matrix is recorded as 1, otherwise it is recorded as 0. The matrix is then expanded into vectors in a certain order.

[0093] Perform feature alignment calculations on the multi-level material distribution map. The multi-level material distribution map contains material distribution information at different levels, and extracts material distribution gradient features. Assuming the material property value of the material distribution map in a certain direction is M(x), the material distribution gradient is calculated by calculating the material property difference between adjacent positions ΔM(x) = M(x + Δx) - M(x) (Δx is a small position increment). These gradient features are integrated and encoded to generate the third fusion vector v3.

[0094] The first fused vector v1, the second fused vector v2, and the third fused vector v3 are combined into the feature tensor T of the three-dimensional parametric model through the tensor fusion layer. The tensor fusion layer can use tensor operations in deep learning, such as tensor product and tensor addition, to fuse the three vectors together according to certain rules. This allows the feature tensor T to fully contain information about the structural skeleton, geometric topological grid, and multi-level material distribution map, laying the foundation for the subsequent generation of the three-dimensional parametric model.

[0095] Example 4:

[0096] Detailed implementation of the material mapping module in the parametric modeling engine: In the material mapping module, the spatial dimension normalization processing is performed on the feature tensor T. Assuming that the value range of the feature tensor T in the spatial dimension is different, in order to facilitate subsequent calculations, it needs to be normalized to a unified range. The normalization formula is used (T ij Represents the element in row i and column j of the feature tensor T, T min and T max are the minimum and maximum values of the characteristic tensor T in the spatial dimension respectively), and generate the material association matrix M.

[0097] The convolutional neural network (CNN) is used to extract the association weights between material distribution and geometric features. A CNN model consisting of convolutional layers, pooling layers, and fully connected layers is constructed. The material association matrix M is used as input, and the convolution kernel in the convolution layer is used to slide convolution on the matrix to extract features of different scales. For example, the size of the convolution kernel K is 3×3, and the convolution operation is performed on the material association matrix M with a stride of 1. ( is the result after convolution). After multiple layers of convolution and pooling operations, a weight mapping matrix W is output through the fully connected layer. The elements in this matrix represent the association weights between material distribution and geometric features.

[0098] Perform matrix dot multiplication on the material association matrix M and the weight mapping matrix W to generate the fusion material distribution feature. The matrix dot multiplication formula is F ij =M ij ×Wij (F ij is the element in the i-th row and j-th column of the fusion material distribution feature matrix), and the fusion material distribution feature matrix F is obtained.

[0099] The fused material distribution features are superimposed on the original feature tensor T through residual connection. The residual connection formula is T new = T + F, outputting a complete parameter sequence for the 3D parametric model. This preserves the important information in the original feature tensor while incorporating the fused material distribution features, allowing the generated 3D parametric model to more accurately reflect the material distribution and geometric characteristics of the special-shaped veneer structure.

[0100] Example 5:

[0101] When constructing the adaptive adjustment strategy graph, the node attributes are initialized according to the design parameter module. The design parameter module contains multiple design parameters, such as structural size parameters, material selection parameters, etc. Each node is assigned a corresponding initial attribute value. For example, node A corresponds to the length parameter in the structural size parameter, and the initial value is set to l0. The edge weight matrix E is generated based on the optimization weight coefficient. Assume that there are multiple optimization weight coefficients, corresponding to different design goals, such as the structural strength target weight w1, the material cost target weight w2, etc. The elements E in the edge weight matrix ij It is calculated based on the time step Δt of parameter adjustment and the optimized weight coefficient, for example (s is the number of design targets, f k (p i ,p j ) is the function related to the kth design goal, p i and p j are the design parameters corresponding to nodes i and j).

[0102] The parameter sequence of a 3D parametric model is used as the node state. This parameter sequence contains information such as the structural skeleton, geometric topology mesh, and material distribution, which changes as design parameters are adjusted. The edge weight matrix, consisting of the time step for parameter adjustment and the optimization weight coefficient, determines the connection strength between nodes and the priority of parameter adjustment.

[0103] The multi-objective weight of each node is iteratively updated through the Pareto frontier algorithm. The optimization weight between nodes is defined as the balance factor between the time step and the objective function weight. The multi-objective weight of each node is initialized to zero, and the starting weight is the preset initial value w start. In each iteration, the objective function values of different nodes under various design objectives are calculated, and the dominance relationship between the nodes is determined by comparing these values. If node i is not worse than node j in all design objectives, and is better than node j in at least one design objective, then node i is said to dominate node j. The non-inferior solution set is screened out through iterative dominance relationships, and the nodes in the non-inferior solution set achieve a certain balance between multiple design objectives. The cumulative optimization weight of each node is calculated, and a complete parameter adjustment sequence is generated by forward deduction based on the non-inferior solution set. During the derivation process, the design parameters of the node are gradually adjusted according to the parameter adjustment order and optimization weight coefficient determined by the edge weight matrix, and finally the optimal parameter adjustment sequence covering all nodes is obtained, and an adaptive adjustment strategy graph is constructed.

[0104] Example 6:

[0105] When building a subdivision rule library, we collect standard subdivision samples for various typical surface types. Using specialized measurement equipment and modeling software, we acquire subdivision sample data for common surfaces, such as spheres, cylinders, and cones. From these samples, we extract benchmark subdivision orders and mesh density distribution characteristics. For example, for a spherical sample with a radius of R, we determine the benchmark subdivision order n0 for different accuracy requirements through experimentation and calculation. We also record the density distribution of the mesh after subdivision on the sphere, such as the difference in mesh density near the poles and the equator.

[0106] Perform a curvature sensitivity analysis on the benchmark subdivision order. Change the curvature value of the surface and observe the effect of the benchmark subdivision order on the surface fitting effect. Taking a sphere as an example, adjust the radius R to change the curvature and calculate the error between the subdivided mesh and the ideal sphere under different curvatures. Generate a multi-resolution subdivision model based on the analysis results, and determine the appropriate subdivision order and mesh density adjustment strategy for different curvature ranges. Classify the subdivision models according to surface type and associate them with a standard parameter database. The collected subdivision models of various surface types are divided into different categories, such as spheres and cylinders. Each category is associated with a corresponding standard parameter database, which contains relevant parameter information for surfaces in that category, such as curvature range and benchmark subdivision order range. The classified subdivision models are stored as a subdivision rule library and regularly updated based on newly collected samples to ensure the accuracy and timeliness of the subdivision rule library.

[0107] In terms of parameter optimization for the gradient descent algorithm, the initial learning rate α0 and momentum factor β are calculated based on the distribution of historical grid adjustment data. The convergence speed and accuracy of grid adjustments in the historical data are analyzed, and appropriate initial values are determined through statistical methods. For example, the rate of change of the objective function value at each grid adjustment in the historical data is calculated, and the initial values of α0 and β are determined based on the distribution of the rate of change. A grid convergence test iterates through parameter combinations, selecting the parameters that best balance convergence speed and accuracy. During the test, different combinations of α0 and β are set, and grid adjustment calculations are performed on the same special-shaped veneer structure. The convergence speed and final accuracy of the different combinations are compared, and the optimal combination is selected. The learning rate and momentum factor are dynamically adjusted based on the balance to optimize the efficiency of node position adjustment. As the grid adjustment process progresses, the values of α and β are dynamically adjusted based on the changes in the objective function value and the convergence trend, allowing the algorithm to converge to the optimal solution faster and more accurately.

[0108] In terms of weight update of the Pareto frontier algorithm, the optimization weight between nodes is defined as the balance factor between the time step and the objective function weight. The multi-objective weight of each node is initialized to zero, and the starting weight is the preset initial value w start . During the iteration process, by calculating the objective function values of different nodes under multiple design objectives, the dominance relationship between the nodes is determined, and the non-inferior solution set is screened out. The cumulative optimization weight of each node is calculated, and a complete parameter adjustment sequence is generated by forward deduction based on the non-inferior solution set. For example, in a scenario with two design objectives, structural strength and material cost, the score of each node under these two objectives is calculated, and the scores between the nodes are compared to determine the dominance relationship, and then the non-inferior solution set is screened out. Finally, the optimal parameter adjustment sequence is determined based on the non-inferior solution set to achieve multi-objective optimization.

[0109] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0110] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An automated design method for special-shaped veneer structures based on parametric modeling, characterized in that: include: Obtaining a set of geometric feature parameters of the special-shaped veneer structure; The geometric characteristic parameters include surface curvature distribution, boundary constraints and material property thresholds; the surface curvature distribution includes local concave-convex features and continuity indicators, and the boundary constraints include spatial connection relationships and load distribution information; Based on the surface curvature distribution, an initial geometric topology mesh is generated by a dynamic subdivision algorithm, wherein the geometric topology mesh includes vertex density distribution and facet connection rules; According to the boundary constraints, a constrained optimization model is used to adjust the node positions of the geometric topology grid to generate an optimized structural skeleton; Performing layered mapping processing on the material attribute thresholds to generate a multi-level material distribution map; Inputting the structural skeleton, geometric topological grid and multi-level material distribution map into a parametric modeling engine to generate a three-dimensional parametric model of the special-shaped veneer structure; Based on the three-dimensional parameterized model, an adaptive adjustment strategy diagram is constructed through a multi-objective optimization algorithm to output a final design solution; the nodes of the adaptive adjustment strategy diagram represent design parameter modules, and the edges represent parameter adjustment sequences and optimization weight coefficients.

2. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 1, characterized in that: The generating of the initial geometric topology mesh by the dynamic subdivision algorithm includes: Performing feature region segmentation on the surface curvature distribution to extract boundary markers of high curvature regions and low curvature regions; Based on the preset subdivision rule library, the local grid density gradient is generated by adaptive subdivision times calculation; The surface fitting algorithm is used to interpolate and encrypt the mesh vertices in the high curvature area, while keeping the mesh sparse in the low curvature area. The grid density gradient and the patch connection rule are encoded as initial parameters of the geometric topology grid.

3. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 1, characterized in that: The method of adjusting the node positions of the geometric topology grid using the constraint optimization model includes: Performing spatial coordinate system transformation on the boundary constraint conditions to generate node displacement constraint equations; The node position adjustment is iteratively calculated based on the gradient descent algorithm, and the Lagrange multiplier balance constraint and objective function are introduced; Update the patch connection rules according to the adjusted node positions and verify the continuity indicators of the geometric topology mesh; The geometric topological mesh that meets the continuity index is marked as the optimized structural skeleton.

4. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 1, characterized in that: The parametric modeling engine includes a grid fusion module and a material mapping module. The grid fusion module includes: Performing parameterized encoding on the node positions of the structural skeleton to generate a first fusion vector; Discretizing the patch connection rule of the geometric topological grid to generate a second fusion vector; Performing feature alignment calculation on the multi-level material distribution map, extracting material distribution gradient features, and generating a third fusion vector; The first fusion vector, the second fusion vector, and the third fusion vector are merged into a feature tensor of the three-dimensional parameterized model through a tensor fusion layer.

5. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 4, characterized in that: The material mapping module includes: Normalize the spatial dimension of the feature tensor to generate the material correlation matrix; The association weights between material distribution and geometric features are extracted through convolutional neural networks to generate a weight mapping matrix; Perform matrix dot multiplication on the material association matrix and the weight mapping matrix to generate the fused material distribution features; The fused material distribution features are superimposed on the original feature tensor through residual connection, and the complete parameter sequence of the three-dimensional parametric model is output.

6. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 1, characterized in that: The method of constructing an adaptive adjustment strategy graph through a multi-objective optimization algorithm includes: Initialize node attributes according to the design parameter module and generate edge weight matrix based on the optimized weight coefficient; The parameter sequence of the three-dimensional parameterized model is used as the node state, and the edge weight matrix is composed of the time step of parameter adjustment and the optimization weight coefficient; Iteratively update the multi-objective weights of each node through the Pareto frontier algorithm and adjust the edge weight matrix; Generate the optimal parameter adjustment sequence covering all nodes according to the adjusted edge weight matrix.

7. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 2, characterized in that: The method for constructing the segmentation rule base includes: Collect standard subdivision samples of various typical surface types and extract the benchmark subdivision times and grid density distribution characteristics; Perform curvature sensitivity analysis on the benchmark subdivision times to generate a multi-resolution subdivision model; Classify subdivision models according to surface type and associate with standard parameter database; The classified segmentation model is stored as a segmentation rule base, and the model is regularly updated based on newly collected samples.

8. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 3, characterized in that: The parameter optimization method of the gradient descent algorithm includes: Calculate the initial learning rate and momentum factor based on the historical grid-adjusted data distribution; Through the grid convergence test, the parameter combinations are traversed and the parameters with the best balance between convergence speed and accuracy are selected; Dynamically adjust the learning rate and momentum factor according to the balance degree to optimize the efficiency of node position adjustment.

9. The method for automated design of special-shaped veneer structures based on parametric modeling according to claim 6, characterized in that: The weight updating method of the Pareto frontier algorithm includes: The optimization weight between nodes is defined as the balance factor between the time step and the objective function weight; Initialize the multi-objective weight of each node to zero and the starting weight to the preset initial value; The non-inferior solution set is screened by iterative dominance relationship, and the cumulative optimization weight of each node is calculated; The complete parameter adjustment sequence is generated by forward deduction based on the non-inferior solution set.

10. An automated design system for special-shaped veneer structures based on parametric modeling, characterized in that: include: Geometric parameter acquisition module: used to obtain a set of geometric characteristic parameters of the special-shaped decorative surface structure, the geometric characteristic parameters including surface curvature distribution, boundary constraints and material property thresholds; wherein the surface curvature distribution includes local concave-convex features and continuity indicators, and the boundary constraints include spatial connection relationships and load distribution information; A topological mesh generation module is configured to generate an initial geometric topological mesh based on the surface curvature distribution through a dynamic subdivision algorithm, wherein the geometric topological mesh includes vertex density distribution and facet connection rules; Structural optimization module: used to adjust the node positions of the geometric topological grid using the constraint optimization model according to the boundary constraint conditions to generate an optimized structural skeleton; Material mapping module: performs layered mapping processing on the material attribute thresholds to generate a multi-level material distribution map; Parametric modeling module: inputting the structural skeleton, geometric topological grid and multi-level material distribution map into the parametric modeling engine to generate a three-dimensional parametric model of the special-shaped veneer structure; Optimization strategy generation module: used to construct an adaptive adjustment strategy diagram based on the three-dimensional parametric model through a multi-objective optimization algorithm and output the final design solution; the nodes of the adaptive adjustment strategy diagram represent the design parameter modules, and the edges represent the parameter adjustment sequence and optimization weight coefficients.

Images