Intelligent Pattern Generation and Material Adaptation System and Method for Intangible Cultural Heritage Crafts

Abstract

This invention relates to the field of intangible cultural heritage protection technology, specifically to an intelligent pattern generation and material adaptation system and method for intangible cultural heritage crafts. The system includes a knowledge graph construction module, a hierarchical perceptual adversarial network module, a material adaptation module, a digital manufacturing module, and a virtual prototyping module. By collecting intangible cultural heritage craft pattern data, a knowledge graph containing pattern features, material properties, and their relationships is constructed to generate intelligent patterns that conform to intangible cultural heritage aesthetics. The material adaptation module analyzes the characteristic parameters of the intelligent patterns and predicts the appropriate material parameters. The digital manufacturing module generates digital manufacturing instructions, and the virtual prototyping module performs physical property simulation. This method establishes a complete digital link from pattern design to material manufacturing, solving the problem of systematic preservation and application of traditional craft knowledge, and promoting the inheritance and development of intangible cultural heritage crafts.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intangible cultural heritage protection, and particularly relates to an intelligent pattern generation and material adaptation system for intangible cultural heritage crafts and a method thereof, and especially relates to intelligent generation of traditional intangible cultural heritage patterns, material property prediction and digital manufacturing using artificial intelligence technology. BACKGROUND

[0002] Intangible cultural heritage is a precious treasure of human civilization, and the traditional crafts and aesthetic concepts contained therein represent the creativity and wisdom of a particular nation. However, with social changes and industrialization, many traditional crafts are at risk of a heritage gap. For example, traditional lacquer craft pattern design and material configuration require decades of experience accumulation, and the experience of traditional craftsmen is difficult to record and pass on completely. Although digital technology provides new means for intangible cultural heritage protection, existing technologies mainly focus on digital archiving and static display, lacking a deep understanding of the internal creative logic of traditional crafts and intelligent reproduction.

[0003] The existing technology has the following main problems: first, the digital representation of traditional craft knowledge is insufficient, making it difficult to convert implicit craft knowledge into explicit computable models; second, existing pattern generation techniques cannot truly understand and reproduce the aesthetic characteristics of traditional patterns; third, traditional material properties and modern digital manufacturing technologies cannot be effectively combined; fourth, there is a lack of systematic solutions that combine pattern design and material properties. These problems seriously limit the digital protection and innovative development of intangible cultural heritage crafts. SUMMARY

[0004] The purpose of the present application is to provide an intelligent pattern generation and material adaptation system for intangible cultural heritage crafts and a method thereof, which realizes intelligent generation of traditional intangible cultural heritage patterns and material adaptation by constructing a hierarchical perception adversarial network and a material property mapping model, providing a solution for the digital protection and innovative development of intangible cultural heritage crafts.

[0005] The present application proposes an intelligent pattern generation and material adaptation system for intangible cultural heritage crafts, comprising:

[0006] A knowledge graph construction module for collecting intangible cultural heritage craft pattern data and constructing a knowledge graph containing pattern features, material properties and their associated relationships;

[0007] A hierarchical perception adversarial network module connected to the knowledge graph construction module for receiving intangible cultural heritage craft pattern feature data sent by the knowledge graph construction module and generating intelligent patterns consistent with intangible cultural heritage aesthetics based on the intangible cultural heritage craft pattern feature data, wherein the hierarchical perception adversarial network module includes a texture perception subnetwork, a color perception subnetwork and a topology perception subnetwork;

[0008] a material adaptation module, connected with the hierarchical perception GAN module, configured to receive the intelligent pattern generated by the hierarchical perception GAN module, analyze feature parameters of the intelligent pattern, and predict material parameters adapted to the intelligent pattern based on the feature parameters;

[0009] a digital manufacturing module, connected with the material adaptation module, configured to receive the material parameters predicted by the material adaptation module, and generate digital manufacturing instructions based on the material parameters and the intelligent pattern;

[0010] and a virtual prototype module, connected with the hierarchical perception GAN module and the material adaptation module, configured to perform physical property simulation on the intelligent pattern generated by the hierarchical perception GAN module and the material parameters predicted by the material adaptation module.

[0011] Preferably, the knowledge graph construction module comprises:

[0012] a data acquisition unit, configured to acquire image data of non-heritage craft patterns, process flow description data, and material property data;

[0013] a data preprocessing unit, connected with the data acquisition unit, configured to perform denoising, geometric correction, and color balance processing on the image data, semantic analysis on the process flow description data, and parameter standardization on the material property data;

[0014] and a knowledge graph generation unit, connected with the data preprocessing unit, configured to construct a triple knowledge graph based on the preprocessed data, wherein the triple knowledge graph comprises entity nodes, relationship edges, and attribute values, the entity nodes comprise pattern units, material types, and process flows, and the relationship edges represent the association relationship between the entity nodes.

[0015] Preferably, the texture perception subnetwork comprises:

[0016] a texture feature extractor, configured to extract multi-scale texture features from non-heritage craft patterns and generate a texture feature vector;

[0017] a texture semantic encoder, connected with the texture feature extractor, configured to map the texture feature vector to a high-dimensional semantic space and generate a texture semantic representation;

[0018] and a texture generator, connected with the texture semantic encoder, configured to generate a texture image conforming to non-heritage aesthetics based on the texture semantic representation.

[0019] Preferably, the color perception subnetwork comprises:

[0020] a color feature extractor, configured to extract color distribution features from non-heritage craft patterns and generate a color feature vector;

[0021] a color semantic mapper, connected with the color feature extractor, configured to establish a mapping relationship between color features and cultural semantics, and generate a color semantic representation;

[0022] and a color scheme generator, connected with the color semantic mapper, configured to generate a color scheme in line with the non-heritage aesthetics based on the color semantic representation.

[0023] Preferably, the topology-aware subnetwork comprises:

[0024] a topology structure analyzer, configured to extract a skeleton structure from the non-heritage craft pattern and construct a topology relationship graph;

[0025] a golden spiral constructor, connected with the topology structure analyzer, configured to construct a multi-level spiral reference framework based on a Fibonacci sequence;

[0026] and a topology optimizer, connected with the golden spiral constructor, configured to perform topology optimization on the pattern structure based on the multi-level spiral reference framework, maintain the topology relationship between pattern elements, and optimize the spatial distribution.

[0027] Preferably, the material adaptation module comprises:

[0028] a pattern feature analysis unit, configured to analyze the visual features, structural features and aesthetic features of the intelligent pattern, and generate pattern feature parameters;

[0029] a material property database, configured to store the physical properties, chemical properties and processing properties of the non-heritage craft materials;

[0030] a feature-material mapping unit, connected with the pattern feature analysis unit and the material property database, configured to construct a mapping relationship between the pattern feature parameters and the material properties, and predict optimal material parameters;

[0031] and a material parameter optimization unit, connected with the feature-material mapping unit, configured to comprehensively optimize the optimal material parameters, and generate a material formula and processing parameters adapted to the intelligent pattern.

[0032] Preferably, the digital manufacturing module comprises:

[0033] a gradient material design unit, configured to design a gradient material structure based on the material parameters, and generate a gradient material distribution scheme;

[0034] a 3D printing parameter prediction unit, connected with the gradient material design unit, configured to predict the layer thickness, temperature curve and printing speed parameters of 3D printing based on the gradient material distribution scheme;

[0035] a shrinkage compensation unit, connected with the 3D printing parameter prediction unit, configured to analyze the solidification shrinkage characteristics of the material, calculate a geometric compensation parameter, and pre-compensate the printing model;

[0036] and a manufacturing instruction generation unit, connected with the shrinkage compensation unit, configured to convert the 3D printing parameters and the geometric compensation parameter into digital manufacturing instructions.

[0037] As preferred, the virtual prototype module comprises:

[0038] a three-dimensional model construction unit configured to construct a three-dimensional virtual model based on the intelligent pattern and the material parameters;

[0039] a finite element analysis unit, connected with the three-dimensional model construction unit, configured to analyze the mechanical properties of the three-dimensional virtual model, and predict the stress distribution and deformation behavior;

[0040] a material rendering unit, connected with the three-dimensional model construction unit, configured to visually render the three-dimensional virtual model based on the material parameters, and simulate the visual effects of the material;

[0041] and a dynamic simulation unit, connected with the finite element analysis unit and the material rendering unit, configured to simulate the dynamic changes of the manufacturing process, and generate a process flow animation.

[0042] As preferred, the method further comprises:

[0043] a human-computer interaction module, connected with the hierarchical perception adversarial network module, the material adaptation module, and the virtual prototype module, configured to provide a parameter adjustment interface, receive user input adjustment instructions, deliver the adjustment instructions to the corresponding modules, and display the adjusted results;

[0044] and a knowledge updating module, connected with the knowledge graph construction module and the human-computer interaction module, configured to update the knowledge graph based on the user input adjustment instructions and the adjusted results.

[0045] An intelligent pattern generation and material adaptation method for non-heritage crafts, comprising:

[0046] constructing a non-heritage craft knowledge graph, including collecting non-heritage craft pattern data, extracting pattern features, material attributes, and their associated relationships, and constructing a triple knowledge graph;

[0047] generating an intelligent pattern, including training a hierarchical perception adversarial network based on the non-heritage craft knowledge graph, wherein the hierarchical perception adversarial network comprises a texture perception sub-network, a color perception sub-network, and a topology perception sub-network, and generating an intelligent pattern conforming to non-heritage aesthetics through the hierarchical perception adversarial network;

[0048] Predicting material parameters, including analyzing characteristic parameters of the intelligent pattern, constructing a mapping relationship between pattern characteristics and material attributes, and predicting material parameters adapted to the intelligent pattern based on the mapping relationship;

[0049] Generating digital manufacturing instructions, including designing gradient material structures based on the material parameters, predicting 3D printing parameters, calculating geometric compensation parameters, and converting the 3D printing parameters and the geometric compensation parameters into digital manufacturing instructions;

[0050] And performing virtual prototype simulation, including constructing a three-dimensional virtual model based on the intelligent pattern and the material parameters, performing mechanical performance analysis and visual rendering on the three-dimensional virtual model, and simulating dynamic changes in the manufacturing process.

[0051] The present application realizes the following beneficial effects by deeply integrating traditional intangible cultural heritage process knowledge with modern artificial intelligence technology:

[0052] 1. A complete digital link from pattern design to material manufacturing is established, solving the problem of systematic preservation and application of traditional process knowledge;

[0053] 2. A hierarchical perception adversarial network is innovatively designed, realizing deep understanding and re-creation of traditional patterns in three dimensions of texture, color and topological structure;

[0054] 3. An intelligent material adaptation mechanism is developed, establishing a mapping relationship between pattern characteristics and material attributes, solving the technical problem of combining traditional patterns with modern materials;

[0055] 4. An integrated solution from design to manufacturing is provided, which reduces the technical threshold of intangible cultural heritage process digital innovation through virtual prototype and digital twin technologies;

[0056] 5. It provides technical support for the protection and innovative development of intangible cultural heritage, and helps to promote the creative transformation and innovative development of traditional culture. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The overall architecture of the intelligent pattern generation and material adaptation system for intangible cultural heritage process of the present application is shown in the figure;

[0058] Figure 2 The structure diagram of the knowledge graph construction module of the present application is shown in the figure;

[0059] Figure 3 The structure diagram of the hierarchical perception adversarial network module of the present application is shown in the figure;

[0060] Figure 4 The structure diagram of the texture perception subnetwork of the present application is shown in the figure;

[0061] Figure 5 This is a schematic diagram of the color perception subnetwork of the present invention;

[0062] Figure 6 This is a schematic diagram of the topology-sensing subnetwork of the present invention;

[0063] Figure 7 This is a schematic diagram of the material adaptation module of the present invention;

[0064] Figure 8 This is a schematic diagram of the structure of the digital manufacturing module of the present invention;

[0065] Figure 9 This is a schematic diagram of the structure of the virtual prototype module of the present invention;

[0066] Figure 10 This is a flowchart of the intelligent pattern generation and material adaptation method for intangible cultural heritage crafts according to the present invention. Detailed Implementation

[0067] Please refer to the attached document. Figures 1-10 The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Those skilled in the art should understand that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention.

[0068] Reference Figure 1 This invention provides an intelligent pattern generation and material adaptation system for intangible cultural heritage crafts, including a knowledge graph construction module 1, a hierarchical perceptual adversarial network module 2, a material adaptation module 3, a digital manufacturing module 4, a virtual prototype module 5, a human-computer interaction module 6, and a knowledge update module 7.

[0069] The knowledge graph construction module 1 is used to collect intangible cultural heritage (ICH) craft pattern data and construct a knowledge graph containing pattern features, material properties, and their relationships. The hierarchical perceptual adversarial network module 2 is connected to the knowledge graph construction module 1 and receives ICH craft pattern feature data from the knowledge graph construction module 1. Based on this data, it generates intelligent patterns that conform to ICH aesthetics. The material adaptation module 3 is connected to the hierarchical perceptual adversarial network module 2 and receives the intelligent patterns generated by the hierarchical perceptual adversarial network module 2. It analyzes the characteristic parameters of the intelligent patterns and predicts the material parameters that match the intelligent patterns based on these parameters. The digital manufacturing module 4 is connected to the material adaptation module 3 and receives the material parameters predicted by the material adaptation module 3. Based on these material parameters and the intelligent patterns, it generates digital manufacturing instructions. The virtual prototype module 5 is connected to the hierarchical perceptual adversarial network module 2 and the material adaptation module 3 and performs physical property simulation on the intelligent patterns generated by the hierarchical perceptual adversarial network module 2 and the material parameters predicted by the material adaptation module 3.

[0070] Each module is interconnected through standardized data interfaces, forming a complete technology chain. During the data flow process, the knowledge graph construction module 1 first transmits the intangible cultural heritage craft characteristic data to the hierarchical perceptual adversarial network module 2. The intelligent patterns and pattern feature data generated by the hierarchical perceptual adversarial network module 2 are then transmitted to the material adaptation module 3. The material parameters predicted by the material adaptation module 3 are then transmitted to the digital manufacturing module 4. At the same time, the intelligent patterns and material parameters are also transmitted to the virtual prototype module 5 for simulation verification.

[0071] Reference Figure 2 The knowledge graph construction module 1 includes a data acquisition unit 11, a data preprocessing unit 12, and a knowledge graph generation unit 13. The data acquisition unit 11 is used to collect image data, process flow description data, and material property data of intangible cultural heritage craft patterns. Preferably, image data acquisition uses a high-resolution scanning device with a resolution of no less than 1200 DPI to ensure complete capture of pattern details. Process flow description data is obtained from intangible cultural heritage inheritors through structured interviews, recording the traditional craft's production steps, technical points, and quality control standards. Material property data includes physical properties (such as hardness, elastic modulus, shrinkage rate, etc.), chemical properties (such as composition, pH value, reactivity, etc.), and processing properties (such as curing time, operability, etc.).

[0072] The data preprocessing unit 12 is connected to the data acquisition unit 11 and is used to perform denoising, geometric correction, and color balancing on image data, semantic analysis on process flow description data, and parameter standardization on material property data. During image preprocessing, a multi-scale denoising algorithm is used to effectively filter out noise generated during acquisition while preserving the detailed features of the pattern. Geometric correction uses affine transformation based on feature point matching to correct image deformation and rotation. Color balancing uses a standard color chart calibration method to ensure the accuracy of color information. Semantic analysis of the process flow description data uses natural language processing technology to extract process terms, operating steps, and causal relationships. Standardization of material property data converts parameters from different sources and units into a unified standard format.

[0073] The knowledge graph generation unit 13 is connected to the data preprocessing unit 12 and is used to construct a triplet knowledge graph based on the preprocessed data. The triplet knowledge graph includes entity nodes, relation edges, and attribute values. Entity nodes include pattern units, material types, and process flows, and relation edges represent the associations between entity nodes. This invention uses an HBase database to store the triplet knowledge graph and uses a Neo4j graph database to realize knowledge association and reasoning. Preferably, the categories of entity nodes include pattern units, material types, process flows, tools, and aesthetic rules, etc.; the categories of relation edges include composition relationships, evolutionary relationships, adaptation relationships, and process relationships, etc.; and attribute values ​​include category, shape, size, color, and cultural connotation, etc.

[0074] A concrete example of knowledge graph construction is modeling the pattern knowledge in traditional lacquerware craftsmanship. First, typical pattern units, such as the meander pattern, cloud pattern, and dragon pattern, are extracted from multiple traditional lacquer artworks, recording their morphological characteristics, proportional relationships, and combination methods. Then, characteristic data of traditional lacquer materials are collected, including the drying time, shrinkage rate, and color characteristics of lacquer from different origins. Finally, the process flow is recorded, including the specific operational methods and quality control standards for steps such as polishing, lacquering, and decoration. This information is organized in triplets, such as (dragon pattern, commonly used in, imperial artifacts), (raw lacquer, drying time, 24 hours), (polishing, preceding process, lacquering), etc. In this way, implicit process knowledge is transformed into explicit, computable models.

[0075] Reference Figure 3 The hierarchical perceptual adversarial network module 2 includes a texture-perceptual subnetwork 21, a color-perceptual subnetwork 22, and a topology-perceptual subnetwork 23. These three subnetworks work together to analyze and generate intangible cultural heritage patterns from three dimensions: texture, color, and topology.

[0076] Reference Figure 4 The texture-aware sub-network 21 includes a texture feature extractor 211, a texture semantic encoder 212, and a texture generator 213. The texture feature extractor 211 is used to extract multi-scale texture features from intangible cultural heritage craft patterns and generate texture feature vectors. Preferably, the texture feature extractor 211 employs a multi-level convolutional structure, setting receptive fields of different scales, increasing from 3×3 to 15×15, to cover texture features at different scales. In practical implementation, an improved VGG network structure can be used to extract feature maps at five different levels, capturing complete information from low-level texture details to high-level structural features.

[0077] The texture semantic encoder 212 is connected to the texture feature extractor 211 and is used to map texture feature vectors to a high-dimensional semantic space to generate texture semantic representations. In this invention, the texture semantic encoding adopts an autoencoder structure, compressing the feature vectors into a 128-dimensional latent space to form a semantic representation of the texture. This mapping can associate visual features with cultural semantics, for example, establishing a mapping relationship between semantic concepts such as delicate smoothness and rough boldness and corresponding texture features.

[0078] Texture generator 213 is connected to texture semantic encoder 212 and is used to generate texture images that conform to the aesthetics of intangible cultural heritage based on texture semantic representation. Texture generator 213 adopts a transposed convolutional network structure, and through multi-level transposed convolution operations, it gradually restores low-dimensional semantic representation into high-resolution texture images. Preferably, texture generator 213 also includes residual connection structure to maintain the effective transmission of detailed information in deep networks and avoid information loss during the generation process.

[0079] A key technological innovation in texture generation is the design of the texture-aware loss function. Traditional adversarial loss functions primarily focus on pixel-level differences, making it difficult to capture the structural characteristics of textures. This invention designs a multi-scale texture-aware loss function, the mathematical expression of which is:

[0080] ,

[0081] in: The total texture-aware loss is used to evaluate the degree of matching between the generated texture and the target style; This is a Gram matrix-based texture representation loss used to measure differences in texture styles; To combat loss, it is used to evaluate the realism of the generated texture; To reconstruct the loss and ensure content consistency; , and These are weighting coefficients, which control the contribution ratio of each loss term. Their values ​​typically range from [0,1]. In this embodiment, they are preferably set to... .

[0082] Gram matrix loss The calculation method is as follows:

[0083] ,

[0084] in: Representing an image In the network The Gram matrix is ​​calculated from the feature map of the layer. The specific calculation method is to flatten the feature map and then perform matrix multiplication. This represents the generated texture image; Represents a real reference texture image; For the first The weight coefficient of a layer controls the degree of influence of features at different levels. Generally, the weight of deeper features is greater than that of shallower features. This represents the Frobenius norm, used to calculate the distance between matrices; This represents the summation across multiple layers of the network, typically using feature maps from layers 2, 4, 7, and 10. The Gram matrix effectively captures the statistical characteristics of textures and is a powerful tool for characterizing texture styles.

[0085] In practical applications, the texture-aware subnetwork 21 can learn from traditional lacquer patterns and generate new patterns with similar aesthetic characteristics. For example, for the traditional coiled dragon pattern, the network can capture its smooth curves and strong lines, and retain these characteristics when generating new patterns, while introducing innovative variations.

[0086] Reference Figure 5 The color perception sub-network 22 includes a color feature extractor 221, a color semantic mapper 222, and a color scheme generator 223. The color feature extractor 221 extracts color distribution features from intangible cultural heritage craft patterns to generate color feature vectors. Preferably, the color feature extractor 221 first converts the RGB color space to the HSV space, which is more consistent with human color perception. Then, it constructs a color histogram to analyze color distribution patterns, including features such as dominant hue, color richness, and color harmony. Finally, it extracts color emotional features and establishes a correlation model between color and emotional expression.

[0087] A color semantic mapper 222 is connected to a color feature extractor 221 to establish a mapping relationship between color features and cultural semantics, generating a color semantic representation. This invention constructs a color vocabulary specific to intangible cultural heritage crafts, such as cinnabar red, pine soot black, and gold in traditional lacquerware, and establishes a connection between color and cultural semantics through a color-semantic mapping matrix. This mapping relationship enables the system to understand the symbolic meaning and aesthetic value of color in traditional culture.

[0088] A color scheme generator 223 is connected to a color semantic mapper 222 and is used to generate color schemes that conform to the aesthetics of intangible cultural heritage based on color semantic representation. The color scheme generator 223 employs a conditional generation model, capable of generating corresponding color schemes based on semantic descriptions. Preferably, the color scheme generator 223 also includes a color harmony evaluation module to ensure that the generated color schemes have aesthetic harmony.

[0089] A key innovation in the color generation process is the design of the color perception adversarial loss function. The mathematical expression for this loss function is:

[0090] ,

[0091] in: The total color perception loss is used to evaluate the quality of the generated color scheme; Color resistance loss is used to assess the realism of generated colors; This is a color semantic consistency loss, used to ensure the consistency between color and target semantics; The loss of color harmony is used to assess the degree of coordination of a color scheme. , and The weighting coefficient controls the contribution ratio of each loss term, and its value ranges from [0,1]. In this embodiment, it is preferably set to... , , Color semantic consistency loss The calculation method is as follows:

[0092] ,

[0093] in: This is a color semantic mapping function that maps color schemes to semantic space; The generated color scheme is represented as a set of RGB values; The target semantic vector describes the desired color style and emotional expression; This represents the L2 norm, used to calculate the Euclidean distance between vectors. (Color semantic mapping function) The color scheme is converted into a 64-dimensional semantic vector using a pre-trained neural network.

[0094] Color harmony loss The calculation is based on color harmony theory to evaluate the degree of harmony of color schemes. In traditional lacquerware craftsmanship, commonly used color harmony models include complementary color harmony and analogous color harmony. The system quantifies the harmony of color schemes by calculating the distribution and spacing of colors on the color wheel. The specific calculation method considers factors such as hue differences, saturation balance, and brightness contrast to construct a comprehensive scoring function.

[0095] In practical applications, the color perception subnetwork 22 can generate color schemes that are both aesthetically pleasing and innovative, based on the color principles of traditional lacquerware craftsmanship. For example, based on the classic combination of traditional vermilion red and pine soot black, the system can generate new color schemes that maintain similar color relationships but have subtle variations, meeting modern aesthetic needs.

[0096] Reference Figure 6The topology-aware subnetwork 23 includes a topology analyzer 231, a golden spiral constructor 232, and a topology optimizer 233. The topology analyzer 231 is used to extract the skeleton structure from intangible cultural heritage craft patterns and construct a topology graph. Preferably, the topology analyzer 231 employs an improved skeleton extraction algorithm to retain the main structural features of the patterns and analyzes the connection relationships and topological properties between nodes using graph theory methods.

[0097] The golden spiral constructor 232 is connected to the topology analyzer 231 and is used to construct a multi-level spiral reference frame based on the Fibonacci sequence. The Fibonacci sequence is a mathematical pattern widely found in nature and traditional art, represented as 0, 1, 1, 2, 3, 5, 8, 13, 21, ..., where each number is the sum of the two preceding numbers. Based on this sequence, a golden spiral can be constructed as a reference frame for the pattern layout. Preferably, the golden spiral constructor 232 sets up a multi-level spiral structure, progressively distributed from the center outwards, forming a harmonious spatial layout.

[0098] Topology optimizer 233 is connected to golden spiral constructor 232 and is used to perform topology optimization on the pattern structure based on a multi-level spiral reference frame, maintaining the topological relationships between pattern elements and optimizing the spatial distribution. The goal of topology optimization is to optimize the aesthetics and stability of the structure while preserving the cultural characteristics of the pattern. Preferably, topology optimizer 233 constructs an objective function that comprehensively considers aesthetics and structural stability, and gradually adjusts the topology of the pattern through an iterative optimization method.

[0099] The objective function for topology optimization is designed as follows:

[0100] ,

[0101] in: The objective function for topology optimization is defined by a smaller value, indicating better optimization performance. An aesthetic scoring function is used to evaluate the aesthetic value of the pattern layout. Use the structural stability function to evaluate the physical stability of the structure; Use a cultural feature fidelity function to evaluate the degree of preservation of cultural features before and after optimization; , and The weighting coefficients control the importance of each objective function, and their values ​​range from [0,1] and satisfy the following conditions: In this embodiment, it is preferably set to 0.3, .

[0102] Aesthetic scoring function Based on composition theory, the balance, symmetry, and rhythm of pattern layout are evaluated. Specific calculations involve quantitative indicators of visual center of gravity distribution, axial symmetry, and spatial rhythm. Structural stability function. The stability of a structure is assessed by analyzing its connectivity and integrity. Specifically, graph theory methods are used to compute graph connectivity, node degree distribution, and critical path identification. A cultural feature fidelity function is also included. By comparing the topological features before and after optimization, it is ensured that the optimization process does not alter the cultural characteristics of the pattern. This involves calculating feature vector similarity and evaluating the preservation of key node positions.

[0103] In traditional lacquerware craftsmanship, the optimization of topological structure is of great significance. For example, for coiled dragon patterns, traditional techniques often employ a rotationally symmetrical layout. Through topological optimization, the curvilinear shape and spatial distribution of the dragon's body can be optimized while maintaining this symmetry, resulting in a more harmonious overall composition.

[0104] The three subnetworks of the hierarchical perceptual adversarial network module 2 achieve overall optimization through a collaborative training mechanism. The training process consists of three stages: pre-training, collaborative training, and fine-tuning. In the pre-training stage, each subnetwork trains independently to establish basic model capabilities. In the collaborative training stage, a joint loss function is constructed, integrating the evaluation results of the three subnetworks to achieve overall optimization. In the fine-tuning stage, expert evaluation feedback is introduced to adjust network parameters and improve the quality of the generated results.

[0105] Another innovation is the global-local collaborative control during the generation process. The system is designed with global structural control parameters and local detail control parameters. The former adjusts the overall layout and proportions, while the latter adjusts the morphological characteristics of the pattern units. Through the global-local linkage mechanism, the overall coordination of the generated patterns is ensured. At the same time, semantic guidance vectors control the generation process, enabling intention-driven creation.

[0106] Reference Figure 7 The material adaptation module 3 includes a pattern feature analysis unit 31, a material property database 32, a feature-material mapping unit 33, and a material parameter optimization unit 34. The pattern feature analysis unit 31 analyzes the visual, structural, and aesthetic features of the intelligent pattern to generate pattern feature parameters. Preferably, visual features include the pattern's color distribution, texture complexity, and line characteristics; structural features include the pattern's topological structure, spatial distribution, and proportional relationships; and aesthetic features include the pattern's stylistic tendencies, emotional expression, and cultural connotations.

[0107] The material properties database 32 is used to store the physical, chemical, and processing properties of materials used in intangible cultural heritage crafts. This invention constructs a comprehensive database of intangible cultural heritage material properties, including not only traditional materials such as lacquer, gold leaf, and silver leaf, but also modern materials such as epoxy resin and acrylate, providing data support for material selection and parameter prediction.

[0108] The feature-material mapping unit 33 is connected to the pattern feature analysis unit 31 and the material property database 32 to construct the mapping relationship between pattern feature parameters and material properties, and to predict the optimal material parameters. This invention uses a deep neural network to construct a feature-material mapping model to achieve the prediction of material parameters from pattern features. The mathematical expression of the mapping model is:

[0109] ,

[0110] in: This is a material parameter vector, including information such as material type, proportion, and process parameters, with dimensions of [missing information]. (In this embodiment) ; The pattern feature vector includes visual features, structural features, and aesthetic features, with dimensions of [dimensions missing]. (In this embodiment) ; The mapping function is implemented using a deep neural network, which will... The pattern features of the dimension are mapped as Material parameters of the dimension.

[0111] Mapping function The network structure comprises multiple fully connected layers. The number of nodes in the input layer matches the pattern feature dimension (128 nodes), and the number of nodes in the output layer matches the material parameter dimension (20 nodes). The intermediate layers employ the ReLU activation function to enhance the network's nonlinear expressive power. Preferably, the network uses a residual connection structure to mitigate the gradient vanishing problem in deep networks. Network training utilizes the backpropagation algorithm, with the mean squared error loss function and the Adam optimizer. The initial learning rate is set to 0.001, and a learning rate decay strategy is employed.

[0112] The material parameter optimization unit 34 is connected to the feature-material mapping unit 33, and is used to comprehensively optimize the optimal material parameters to generate a material formula and processing parameters adapted to the smart pattern. Preferably, the material parameter optimization considers multiple factors, including visual effect, physical properties, manufacturing difficulty, and cost. The optimization process uses a multi-objective optimization algorithm to balance the weights of different factors and generate the comprehensive optimal material parameters.

[0113] In traditional lacquerware craftsmanship, material matching is a complex process that requires consideration of multiple factors, including the lacquer's color, transparency, and drying time. This invention, through its material matching module 3, can predict the most suitable lacquer material parameters, such as film thickness, lacquer liquid ratio, and curing conditions, based on the characteristics of the generated pattern. For example, for delicate and refined patterns, the system predicts higher film gloss and lower film thickness; for bold and unrestrained patterns, it predicts lower gloss and thicker film thickness.

[0114] Reference Figure 8 The digital manufacturing module 4 includes a gradient material design unit 41, a 3D printing parameter prediction unit 42, a shrinkage compensation unit 43, and a manufacturing instruction generation unit 44. The gradient material design unit 41 is used to design a gradient material structure based on material parameters and generate a gradient material distribution scheme. Preferably, the gradient material design adopts a multi-layered structure, gradually transitioning from the bottom layer to the surface layer, simulating the material layering in traditional processes.

[0115] The 3D printing parameter prediction unit 42 is connected to the gradient material design unit 41 and is used to predict the layer thickness, temperature profile, and printing speed parameters of 3D printing based on the gradient material distribution scheme. This invention uses a BP neural network to construct the 3D printing parameter prediction model. The network structure includes an input layer, a hidden layer, and an output layer, with 6, 8, and 6 nodes respectively. Input parameters include material hardness, tensile strength, color parameters, and curing characteristics; output parameters include printing layer thickness, temperature profile, printing speed, and cooling time.

[0116] Temperature profile parameters are key parameters in 3D printing. This invention designs a temperature profile control model:

[0117] ,

[0118] in: For the first The printing temperature of the layer, in degrees Celsius; The reference temperature is generally set as the lowest operating temperature of the material, and the unit is degrees Celsius. This represents the temperature range, indicating the maximum temperature change from the bottom layer to the top layer, expressed in degrees Celsius. This is a temperature gradient function that controls the shape of the curve representing temperature changes. This is a layer index, starting from 0 for counting; represents the total number of layers in the printed model. In this embodiment, for a large paint-based material, The preferred setting is 60°C. The preferred setting is 40°C, total number of layers The number of layers is determined by the model complexity, and is generally 30-100.

[0119] Preferably, the temperature gradient function An S-shaped curve is used to achieve a smooth temperature transition:

[0120] ,

[0121] in: This represents the normalized temperature ratio, with a value range of [0,1]. The normalized floor height position has a value range of [0,1], representing the relative position of the current floor in the total height; This parameter controls the steepness of the S-shaped curve. The value range is usually [5, 15]. The larger the value, the steeper the curve. In this embodiment, it is preferably set to 10. The base of the natural logarithm is approximately 2.71828. This S-shaped curve results in faster temperature changes in the middle layer and slower changes in the bottom and top layers, which is beneficial for material stability and print quality.

[0122] The shrinkage compensation unit 43 is connected to the 3D printing parameter prediction unit 42 and is used to analyze the curing shrinkage characteristics of the material, calculate geometric compensation parameters, and pre-compensate the printed model. In traditional lacquer processing, shrinkage occurs during the curing process, resulting in discrepancies between the finished product dimensions and the design. This invention uses the shrinkage compensation unit 43 to predict the material's shrinkage rate and pre-compensate the printed model, ensuring the dimensional accuracy of the final product.

[0123] The manufacturing instruction generation unit 44 is connected to the shrinkage compensation unit 43 and is used to convert 3D printing parameters and geometric compensation parameters into digital manufacturing instructions. Preferably, the manufacturing instructions adopt the standard G-code format, which is compatible with mainstream 3D printing equipment and ensures the reliability and stability of the manufacturing process.

[0124] In the digital manufacturing of traditional lacquerware, the digital manufacturing module 4 can generate precise 3D printing instructions based on the material parameters predicted by the material adaptation module 3. For example, for patterns requiring multiple layers of lacquer film, the system will generate multi-layered gradient printing parameters to simulate the layered effect of traditional lacquerware; for gold leaf decoration, the system will adjust the printing parameters to ensure the adhesion of the gold leaf material.

[0125] Reference Figure 9 The virtual prototype module 5 includes a 3D model construction unit 51, a finite element analysis unit 52, a material rendering unit 53, and a dynamic simulation unit 54. The 3D model construction unit 51 is used to construct a 3D virtual model based on intelligent patterns and material parameters. Preferably, the 3D model construction process includes extracting boundaries and feature surfaces from the pattern, performing region segmentation based on color and features, meshing the segmented regions, and constructing a 3D surface rendering model.

[0126] The finite element analysis unit 52 is connected to the three-dimensional model construction unit 51, and is used to perform mechanical property analysis on the three-dimensional virtual model, predicting stress distribution and deformation behavior. This invention uses the finite element analysis method to establish a mechanical model of the material and analyze the stress distribution and deformation of the patterned structure under external forces. The core of finite element analysis is to construct the mechanical equations of the material:

[0127] ,

[0128] in, For stress tensor, it represents the stress state at various points inside the material and is a 3×3 symmetric tensor; The strain tensor represents the degree of material deformation and is also a 3×3 symmetric tensor. Let be the material stiffness matrix, representing the elastic properties of the material. For isotropic materials, it is a 6×6 matrix, derived from Young's modulus. Compared to Poisson Confirmed. Regarding lacquer, a commonly used material in intangible cultural heritage crafts, Typically, it falls within the range of [500MPa, 2000MPa]. It is usually in the range [0.25, 0.35].

[0129] Preferably, for composite material structures, a layered modeling method is adopted, setting different mechanical parameters for each layer of material to more accurately simulate the mechanical behavior of real materials. Through finite element analysis, the mechanical properties of the patterned structure can be predicted, potential structural problems can be identified in a timely manner, and the design scheme can be optimized.

[0130] The material rendering unit 53 is connected to the 3D model construction unit 51 and is used to perform visual rendering of the 3D virtual model based on material parameters to simulate the visual effects of the material. Preferably, the material rendering adopts physically based rendering (PBR) technology, which achieves highly realistic visual effects by simulating the physical interaction between light and the material surface. Rendering parameters include diffuse color, specular reflectivity, roughness, metallicity, etc., which are directly related to the physical properties of the material.

[0131] The dynamic simulation unit 54 is connected to the finite element analysis unit 52 and the material rendering unit 53 to simulate the dynamic changes in the manufacturing process and generate process flow animation. Preferably, the dynamic simulation adopts a multi-frame sequence method to simulate the traditional manufacturing process. The specific implementation method is: set the number of frames in the process animation, divide the process into multiple consecutive frames, and use the deformation data of the previous frame as the initial condition for the next frame to form a continuous dynamic process.

[0132] In Virtual Prototyping Module 5, the digital twin-driven mechanism is an innovative feature. The system constructs a physical-virtual mapping system to achieve two-way data interaction. Changes in the virtual model reflect the state of the physical model, and operations on the physical model update the virtual model. This two-way mapping mechanism enables the virtual prototype to more accurately reflect the real manufacturing process, providing a reliable verification platform for the digital innovation of intangible cultural heritage crafts.

[0133] The present invention also includes a human-computer interaction module 6 and a knowledge update module 7. The human-computer interaction module 6 is connected to the hierarchical perceptual adversarial network module 2, the material adaptation module 3, and the virtual prototype module 5, and is used to provide a parameter adjustment interface, receive adjustment commands input by the user, transmit the adjustment commands to the corresponding modules, and display the adjusted results. The knowledge update module 7 is connected to the knowledge graph construction module 1 and the human-computer interaction module 6, and is used to update the knowledge graph based on the adjustment commands input by the user and the adjusted results.

[0134] Reference Figure 10 This invention also provides a method for intelligent pattern generation and material adaptation for intangible cultural heritage crafts, including the following steps:

[0135] Step S1: Construct a knowledge graph of intangible cultural heritage crafts, including collecting data on intangible cultural heritage craft patterns, extracting pattern features, material properties and their relationships, and constructing a triplet knowledge graph;

[0136] Step S2, generating intelligent patterns, including training a hierarchical perceptual adversarial network based on the knowledge graph of intangible cultural heritage crafts, wherein the hierarchical perceptual adversarial network includes a texture perception subnetwork, a color perception subnetwork and a topology perception subnetwork, and generating intelligent patterns that conform to the aesthetics of intangible cultural heritage through the hierarchical perceptual adversarial network.

[0137] Step S3, predict material parameters, including analyzing the feature parameters of the smart pattern, constructing a mapping relationship between pattern features and material properties, and predicting material parameters that are compatible with the smart pattern based on the mapping relationship;

[0138] Step S4: Generate digital manufacturing instructions, including designing gradient material structures based on material parameters, predicting 3D printing parameters, calculating geometric compensation parameters, and converting 3D printing parameters and geometric compensation parameters into digital manufacturing instructions.

[0139] Step S5 involves performing virtual prototype simulation, including building a three-dimensional virtual model based on intelligent patterns and material parameters, performing mechanical performance analysis and visual rendering on the three-dimensional virtual model, and simulating the dynamic changes in the manufacturing process.

[0140] This invention deeply integrates traditional intangible cultural heritage (ICH) craft knowledge with modern artificial intelligence technology, achieving intelligent generation and material adaptation of traditional patterns, and providing a systematic solution for the digital protection and innovative development of ICH crafts. Compared with existing technologies, this invention has the following significant advantages: First, it establishes a complete digital link from pattern design to material manufacturing, solving the problem of systematic preservation and application of traditional craft knowledge; second, it innovatively designs a hierarchical perceptual adversarial network, realizing a deep understanding and re-creation of traditional patterns in three dimensions: texture, color, and topological structure; third, it develops an intelligent material adaptation mechanism, establishing a mapping relationship between pattern features and material properties, solving the technical challenge of combining traditional patterns with modern materials; fourth, it provides an integrated solution from design to manufacturing, lowering the technical threshold for digital innovation of ICH crafts through virtual prototyping and digital twin technology.

[0141] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An intelligent pattern generation and material adaptation system for non-heritage crafts, characterized in that, Comprise: a knowledge graph construction module for collecting intangible cultural heritage process pattern data and constructing a knowledge graph containing pattern features, material attributes, and their associated relationships; a hierarchical perception adversarial network module connected to the knowledge graph construction module for receiving intangible cultural heritage process pattern feature data sent by the knowledge graph construction module and generating intelligent patterns conforming to intangible cultural heritage aesthetics based on the intangible cultural heritage process pattern feature data, wherein the hierarchical perception adversarial network module includes a texture perception subnetwork, a color perception subnetwork, and a topology perception subnetwork; a material adaptation module connected to the hierarchical perception adversarial network module for receiving intelligent patterns generated by the hierarchical perception adversarial network module, analyzing the feature parameters of the intelligent patterns, and predicting material parameters adapted to the intelligent patterns based on the feature parameters; a digital manufacturing module connected to the material adaptation module for receiving material parameters predicted by the material adaptation module and generating digital manufacturing instructions based on the material parameters and the intelligent patterns; and a virtual prototype module connected to the hierarchical perception adversarial network module and the material adaptation module for simulating the physical properties of intelligent patterns generated by the hierarchical perception adversarial network module and material parameters predicted by the material adaptation module; the texture perception subnetwork includes: a texture feature extractor for extracting multi-scale texture features from intangible cultural heritage process patterns and generating a texture feature vector; a texture semantic encoder connected to the texture feature extractor for mapping the texture feature vector to a high-dimensional semantic space and generating a texture semantic representation; and a texture generator connected to the texture semantic encoder for generating texture images conforming to intangible cultural heritage aesthetics based on the texture semantic representation; the color perception subnetwork includes: a color feature extractor for extracting color distribution features from intangible cultural heritage process patterns and generating a color feature vector; a color semantic mapper connected to the color feature extractor for establishing a mapping relationship between color features and cultural semantics and generating a color semantic representation; and a color scheme generator connected to the color semantic mapper for generating color schemes conforming to intangible cultural heritage aesthetics based on the color semantic representation; the topology perception subnetwork includes: a topology structure analyzer for extracting skeleton structures from intangible cultural heritage process patterns and constructing a topology relationship graph; a golden spiral constructor connected to the topology structure analyzer for constructing a multi-level spiral reference framework based on the Fibonacci sequence; and a topology optimizer connected to the golden spiral constructor for topologically optimizing pattern structures based on the multi-level spiral reference framework, maintaining the topological relationships between pattern elements and optimizing spatial distribution.

2. The intangible craft process oriented intelligent pattern generation and material adaptation system according to claim 1, wherein, the knowledge graph construction module includes: a data acquisition unit for acquiring image data, process flow description data, and material property data of intangible cultural heritage process patterns; a data preprocessing unit connected to the data acquisition unit for denoising, geometric correction, and color balance processing the image data, performing semantic analysis on the process flow description data, and parameter standardization on the material property data; And a knowledge graph generation unit connected with the data preprocessing unit, configured to construct a triple knowledge graph based on the preprocessed data, the triple knowledge graph comprising entity nodes, relation edges and attribute values, the entity nodes comprising pattern units, material types and process flows, and the relation edges representing the association between the entity nodes. 3.The smart pattern generation and material adaptation system for intangible cultural heritage process according to claim 1, wherein, The material adaptation module comprises: a pattern feature analysis unit configured to analyze the visual features, structural features and aesthetic features of the intelligent pattern, and generate pattern feature parameters; a material property database configured to store the physical properties, chemical properties and processing properties of the non-heritage craft material; a feature-material mapping unit connected with the pattern feature analysis unit and the material property database, configured to construct a mapping relationship between the pattern feature parameters and the material properties, and predict optimal material parameters; and a material parameter optimization unit connected with the feature-material mapping unit, configured to comprehensively optimize the optimal material parameters, and generate a material formula and processing parameters adapted to the intelligent pattern.

4. The intangible craft process oriented smart pattern generation and material adaptation system as claimed in claim 1, wherein, The digital manufacturing module comprises: a gradient material design unit configured to design a gradient material structure based on the material parameters, and generate a gradient material distribution scheme; a 3D printing parameter prediction unit connected with the gradient material design unit, configured to predict the layer thickness, temperature curve and printing speed parameters of 3D printing based on the gradient material distribution scheme; a shrinkage compensation unit connected with the 3D printing parameter prediction unit, configured to analyze the solidification shrinkage properties of the material, calculate geometric compensation parameters, and pre-compensate the printing model; and a manufacturing instruction generation unit connected with the shrinkage compensation unit, configured to convert the 3D printing parameters and the geometric compensation parameters into digital manufacturing instructions.

5. The intangible craft process oriented smart pattern generation and material adaptation system as claimed in claim 1, wherein, The virtual prototype module comprises: a three-dimensional model construction unit configured to construct a three-dimensional virtual model based on the intelligent pattern and the material parameters; a finite element analysis unit connected with the three-dimensional model construction unit, configured to analyze the mechanical properties of the three-dimensional virtual model, and predict the stress distribution and deformation behavior; a material rendering unit connected with the three-dimensional model construction unit, configured to visually render the three-dimensional virtual model based on the material parameters, and simulate the visual effect of the material; and a dynamic simulation unit connected with the finite element analysis unit and the material rendering unit, configured to simulate the dynamic changes of the manufacturing process, and generate a process flow animation.

6. The intangible craft process oriented smart pattern generation and material adaptation system as claimed in claim 1, wherein, Further comprising: a human-computer interaction module connected with the hierarchical perception adversarial network module, the material adaptation module and the virtual prototype module, configured to provide a parameter adjustment interface, receive user input adjustment instructions, deliver the adjustment instructions to the corresponding modules, and display the adjusted results; and a knowledge update module connected with the knowledge graph construction module and the human-computer interaction module, configured to update the knowledge graph based on the user input adjustment instructions and the adjusted results.

7. The method for intelligent pattern generation and material adaptation for non-heritage crafts, using the system of any one of claims 1-6, characterized in that, Comprise: constructing a non-heritage craft knowledge graph, comprising collecting non-heritage craft pattern data, extracting pattern features, material properties and their association, and constructing a triple knowledge graph; The method comprises the following steps: generating an intelligent pattern, including training a hierarchical perception adversarial network based on the non-heritage craft knowledge graph, wherein the hierarchical perception adversarial network comprises a texture perception subnetwork, a color perception subnetwork and a topological perception subnetwork, and the intelligent pattern conforming to the non-heritage aesthetics is generated through the hierarchical perception adversarial network; predicting material parameters, including analyzing feature parameters of the intelligent pattern, constructing a mapping relationship between pattern features and material attributes, and predicting material parameters adapted to the intelligent pattern based on the mapping relationship; generating digital manufacturing instructions, including designing gradient material structures based on the material parameters, predicting 3D printing parameters, calculating geometric compensation parameters, and converting the 3D printing parameters and the geometric compensation parameters into digital manufacturing instructions; and performing virtual prototype simulation, including constructing a three-dimensional virtual model based on the intelligent pattern and the material parameters, performing mechanical performance analysis and visual rendering on the three-dimensional virtual model, and simulating the dynamic changes of the manufacturing process.

Images