Rigid origami recommendation system

Abstract

The invention relates to the technical field of intelligent recommendation systems, and discloses a rigid origami recommendation system, comprising: a knowledge graph management module for constructing and maintaining a knowledge graph of rigid origami, including a product entity, a patent entity and a geometric feature entity, and dynamically updating attributes and relationships of the entities; the user behavior capturing and feedback module is used for recording behavior data of the user and dynamically adjusting recommendation parameters according to user feedback; the multi-objective optimization recommendation algorithm module is used for generating a recommendation result based on multi-objective optimization of energy minimization, geometric conflict minimization and path complexity minimization of the folding path; and the dynamic physical modeling module is used for verifying the physical consistency of the recommendation result based on the geometric and physical characteristics of the rigid origami. According to the method, high precision, dynamic adaptability and practical feasibility of rigid origami path recommendation are realized through dynamic knowledge graph expansion, a multi-objective optimization algorithm and a physical verification and feedback mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent recommendation systems, and specifically to a rigid origami recommendation system. Background Art

[0002] As a new interdisciplinary technology, rigid origami is gradually being widely applied in fields such as spacecraft deployment structures, architectural design, and industrial product manufacturing due to its outstanding advantages in structural foldability, space utilization efficiency, and material savings. However, the design and application of rigid origami require in-depth calculations and verifications at multiple levels such as geometric modeling, physical verification, and optimization recommendation. Designing a folding path that conforms to geometric laws and satisfies physical constraints often takes a lot of time and professional knowledge. Therefore, an intelligent recommendation system is needed to help users quickly find the optimal folding path among a wide range of design options.

[0003] In the prior art, certain progress has been made in the research of recommendation systems. In some systems, knowledge graphs are used to manage and organize relevant knowledge in the field, and their structured data can significantly improve the organization efficiency of information. In addition, the application of multi-objective optimization algorithms enables the recommendation system to be optimized based on specific single objectives (such as energy minimization or path simplification). These technical solutions have certain advantages in meeting specific requirements and simplifying the design process, especially when dealing with standardized problems or known requirements, showing high reliability and operability.

[0004] However, there are still many deficiencies in the prior art, which have become the main obstacles restricting the wide application of rigid origami recommendation systems. First of all, most existing knowledge graphs are static structures and cannot be updated in real time according to user behavior, making it difficult for them to adapt to dynamic requirements or integrate new knowledge. Secondly, most existing recommendation algorithms are mainly based on single-objective optimization and do not comprehensively consider multiple factors such as path energy, geometric conflicts, and complexity, making it difficult to meet users' requirements for multi-dimensional performance of folding paths. In addition, the geometric and physical property verification mechanism is insufficient. In the prior art, the verification of folding paths mostly stays at the geometric calculation level and lacks support for actual mechanical properties and dynamic parameter adjustment, resulting in possible problems in the actual application of recommendation results. Finally, the user interaction function in existing systems is limited, and it is unable to fully capture user behavior data for real-time feedback optimization, resulting in an obvious deviation between the recommendation results and user requirements. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides a rigid origami recommendation system, which solves the problems of insufficient knowledge dynamic expansion, lack of multi-objective optimization, imperfect physical verification, and insufficient utilization of user feedback in existing rigid origami recommendation systems.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: A rigid origami recommendation system, comprising:

[0007] A knowledge graph management module, used to construct and maintain the knowledge graph of rigid origami, including product entities, patent entities, and geometric feature entities, and dynamically update the attributes and relationships of the entities;

[0008] A user behavior capture and feedback module, used to record the user's behavior data and dynamically adjust the recommendation parameters according to the user's feedback;

[0009] A multi-objective optimization recommendation algorithm module, used to generate recommendation results based on the multi-objective optimization of energy minimization, geometric conflict minimization, and path complexity minimization of the folding path;

[0010] A dynamic physical modeling module, used to verify the physical consistency of the recommendation results based on the geometric and physical characteristics of rigid origami;

[0011] A user interaction and visualization interface module, used to provide a search function and display the recommendation results, and receive the user's real-time feedback.

[0012] Preferably, the knowledge graph management module includes:

[0013] An entity definition module, used to define product entities, patent entities, and geometric feature entities in the knowledge graph;

[0014] An attribute update module, used to dynamically add unmatched keywords to the attributes of the entities through the user behavior data;

[0015] A relationship update module, used to update the relationship weights between entities according to the user click times and keyword relevance.

[0016] Preferably, the user behavior capture and feedback module includes:

[0017] A behavior record module, used to record the user's search times and click times;

[0018] A parameter adjustment module, used to dynamically adjust the optimization weight factor in the recommendation algorithm according to the user's click behavior;

[0019] A data transmission module, used to transmit the user behavior data to the knowledge graph management module and the recommendation algorithm module.

[0020] Preferably, the multi-objective optimization recommendation algorithm module includes:

[0021] A path energy calculation module, used to calculate the energy of each folding path;

[0022] A geometric conflict detection module, used to detect the geometric conflict nodes in the path;

[0023] A path complexity analysis module for calculating the total number of nodes in a path;

[0024] A comprehensive score calculation module for calculating a comprehensive score based on the weights of energy, geometric conflict, and path complexity;

[0025] An optimization model execution module for generating recommended results by screening non-dominated solution paths.

[0026] Preferably, the dynamic physical modeling module includes:

[0027] A geometric consistency verification module for verifying the geometric consistency of the folding path based on a rigid rotation matrix and an angle theorem;

[0028] A parameter dynamic adjustment module for dynamically adjusting the crease stiffness parameter and the target folding angle parameter based on user behavior;

[0029] A physical constraint verification module for verifying the folding constraint conditions in the recommended path.

[0030] Preferably, the user interaction and visualization interface module includes:

[0031] A search interface module for providing a user to input search content and select an entity type;

[0032] A recommended result display module for grouping and displaying recommended content according to optimization goals;

[0033] A real-time feedback module for receiving a user's score for the recommended result and transmitting it to the feedback module.

[0034] Preferably, the attribute update module is completed through the following steps:

[0035] Obtain the search content input by the user and perform word segmentation;

[0036] Extract unmatched keywords and record the click times and detail times of the keyword;

[0037] Calculate the relevance of the keyword, and when the relevance exceeds a preset threshold, add the keyword to the attributes of the corresponding entity in the knowledge graph.

[0038] Preferably, the comprehensive score calculation module is completed through the following steps:

[0039] Calculate the deviation between the current angle and the target angle for each crease of the path;

[0040] Calculate the total energy of the path according to the crease stiffness;

[0041] Detect geometric conflict nodes in the path and record the number of conflicts;

[0042] Calculate the total number of nodes in the path;

[0043] Based on the path energy, the number of conflicting nodes, and the total number of nodes, calculate a comprehensive score according to preset weights.

[0044] Preferably, the parameter dynamic adjustment module is completed through the following steps:

[0045] Dynamically adjust the crease stiffness parameter according to the user's stay time and click behavior on the recommendation result page;

[0046] Update the target folding angle parameter according to the search content input by the user and the angular deviation of the associated path;

[0047] Apply the adjusted parameters to the physical constraint verification module.

[0048] Preferably, the recommendation result display module is completed through the following steps:

[0049] Filter the recommended content according to the entity type selected by the user;

[0050] Group and sort the recommended content according to the optimization goal;

[0051] Display the detailed information of the recommended content on the recommendation result interface, including path complexity, geometric conflict situation, and relevant crease angles.

[0052] The present invention provides a rigid origami recommendation system. It has the following beneficial effects:

[0053] 1. Through the technical solution of dynamically expanding the knowledge graph, the present invention can update the attributes and relationship weights of entities in real time, adapting to the changes in user behavior and search needs. This design makes the system more agile in learning and integrating new knowledge. Compared with the problem of lagging updates caused by the fixed structure of the knowledge graph in the prior art, the present invention effectively solves the technical shortcomings of poor adaptability and insufficient coverage.

[0054] 2. The present invention adopts a multi-objective optimization algorithm, which comprehensively considers multiple key factors such as path energy, geometric conflict, and path complexity to generate a more scientific and reasonable recommended path. This way not only makes the recommended content more in line with actual needs but also takes into account the balance of multi-dimensional performance. Compared with the one-sidedness of single-objective optimization recommendations in the prior art, it overcomes the deficiency that the diverse needs of users cannot be met.

[0055] 3. Through the dynamic physical modeling module, the present invention adds rigid rotation verification and geometric consistency verification to the recommended path to ensure that the path design conforms to the theoretical constraints of rigid origami. At the same time, dynamically adjusting the crease stiffness and angle parameters makes the recommended path highly feasible. Compared with the prior art lacking physical verification, the present invention significantly improves the application ability of the recommended content in actual scenarios.

[0056] 4. The present invention combines a user behavior capture and feedback module to dynamically adjust the recommendation logic by recording search, click, and rating data, enabling seamless matching of recommended content with user needs. This design optimized based on real-time data effectively overcomes the problem that traditional systems cannot flexibly respond to changes in user interests, making the recommendation results more personalized and accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is the system architecture diagram of the present invention;

[0058] Figure 2 is the module structure diagram of the knowledge graph management module of the present invention;

[0059] Figure 3 is the module structure diagram of the user behavior capture and feedback module of the present invention;

[0060] Figure 4 is the module structure diagram of the multi-objective optimization recommendation algorithm module of the present invention;

[0061] Figure 5 is the module structure diagram of the dynamic physical modeling module of the present invention;

[0062] Figure 6 is the module structure diagram of the user interaction and visualization interface module of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the specification of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0064] Please refer to the attached Figure 1 - attached Figure 6 , the embodiment of the present invention provides a rigid origami recommendation system, including:

[0065] A knowledge graph management module for constructing and maintaining a knowledge graph of rigid origami, including product entities, patent entities, and geometric property entities, and dynamically updating the attributes and relationships of the entities;

[0066] The knowledge graph management module provides the necessary knowledge support and data sources. The dynamic nature of the knowledge graph enables the system to continuously update according to real-time user needs to meet complex and diverse recommendation scenarios. The following describes this module in detail in combination with specific implementation manners.

[0067] In this embodiment, the knowledge graph management module is mainly used to construct and maintain the knowledge graph in the field of rigid origami, and improve the adaptability of the knowledge graph through a dynamic extension mechanism.

[0068] Generally, a knowledge graph consists of two parts: entities and relationships. Entities include origami models, patent information, and geometric properties, and relationships are used to describe the associations between these entities. For example, there may be a relationship of "having a certain property" between an origami model and geometric properties, and a relationship of "involving a certain property" between patent information and geometric properties.

[0069] As an option, the entities and relationships of the knowledge graph can be stored in the form of a graph structure, where nodes represent entities and edges represent relationships. Specifically, each entity can have multiple attributes, and each edge can have a weight value to represent the association strength between entities.

[0070] In a possible implementation, the construction and dynamic extension of the knowledge graph include the following parts:

[0071] In this embodiment, the entity construction of the knowledge graph includes the following:

[0072] The origami model entity includes, but is not limited to, the following attributes:

[0073] Product name, used to identify a specific origami model.

[0074] Configuration type, such as planar folding, spatial folding.

[0075] Material, such as metal sheet, polymer film.

[0076] The geometric property entity includes the following:

[0077] Crease type, such as mountain fold, valley fold.

[0078] Folding method, such as rigid folding, flexible folding.

[0079] Processing method, such as laser cutting, mechanical pressing.

[0080] As an option, the weight value of the relationship can be calculated by the following formula:

[0081] R(k1,k2) = (S1·r + S2)·E(k1,k2)

[0082] Where: R(k1,k2) represents the relationship weight between entity k1 and entity k2; S1 represents the number of entities searched for the same keyword; S2 represents the number of entities not searched for the same keyword; r is a correction coefficient used to adjust the influence of keyword search results on the relationship weight, usually taking a value of 2; E(k1,k2) represents the number of user click behavior statistics.

[0083] In this embodiment, the dynamic expansion of the knowledge graph includes the following:

[0084] When the user inputs a keyword for searching, the system will first retrieve in the knowledge graph whether there is such a keyword. If the keyword is not matched, the system will record this keyword as an unmatched keyword, and at the same time count its click times C(N2l,k1) and the click times D(N2l,k1) of the details page.

[0085] Specifically, the calculation formula for the relevance degree of the keyword is:

[0086]

[0087] Where: G(N2l,k1) represents the relevance degree between keyword N2 and entity k1; D(N2l,k1) represents the number of times the user clicks the "original text link" on the details page; C(N2l,k1) represents the total number of times the user clicks keyword N2 and entity k1.

[0088] Generally, when the relevance degree of the keyword exceeds the preset threshold, the system will dynamically add this keyword to the attributes of entity k1. For example, when the user searches for "flexible folding" and it is not matched in the knowledge graph, but after clicking on relevant origami models multiple times, the system adds "flexible folding" to the geometric property attributes of the corresponding model.

[0089] In this embodiment, the update process of the knowledge graph includes the following steps:

[0090] First, the system extracts entity and attribute information from the keywords input by the user. If the unmatched keywords have a large number of click times, the system marks them as candidate attributes and dynamically updates them to the knowledge graph.

[0091] Then, the relationship weights are adjusted according to the user's click and search behaviors. The higher the click times, the greater the relationship weight. This dynamic adjustment mechanism ensures that the relationship structure of the knowledge graph can reflect the real needs of users.

[0092] Finally, the system automatically supplements the possible missing content in the knowledge graph through a real-time expansion mechanism. Such a mechanism not only improves the integrity of the knowledge graph but also enhances the system's adaptability to domain knowledge.

[0093] Through the above implementation methods, the knowledge graph management module realizes the comprehensive modeling and dynamic expansion of the core data in the field of rigid origami, providing basic support for the efficient operation of the recommendation system.

[0094] The user behavior capture and feedback module is used to record the user's behavior data and dynamically adjust the recommendation parameters according to the user's feedback;

[0095] The main function of the user behavior capture and feedback module is to capture the user's interaction behaviors in real time and optimize the weights of the recommendation algorithm through a dynamic feedback mechanism, thereby improving the accuracy and personalization level of the recommended content. This module directly depends on the entity and relationship data provided by the knowledge graph management module, and at the same time provides quantitative data support for the multi-objective optimization recommendation algorithm module. The functions and implementation of this module are described in detail below in combination with specific implementation methods.

[0096] In this embodiment, the user behavior capture and feedback module includes three main functional parts: behavior recording, parameter dynamic adjustment, and data transmission.

[0097] Generally, this module generates behavior variables for optimizing the recommendation algorithm by recording interaction data such as the user's search times, click times, and dwell time. After being processed, the behavior data is transmitted to the knowledge graph management module and the recommendation algorithm module for dynamically adjusting the recommendation logic.

[0098] As an option, the capture of user behavior can be divided into explicit behavior and implicit behavior. Explicit behavior includes the user's search and click operations, while implicit behavior is captured through detailed data such as dwell time and page scrolling. Specifically, after being statistically analyzed and processed, these data form user behavior variables.

[0099] In this embodiment, the specific implementation of the behavior recording module includes the following:

[0100] In a possible implementation, after the user enters a keyword in the system, the system will record the following behavior data:

[0101] Search times A(N1m), which represents the number of times the keyword N1 is searched by the user.

[0102] Click times B(N1m), which represents the number of times the user clicks on the content related to the keyword N1 in the recommendation results.

[0103] Open times L(N1m), which represents the total number of visits by the user to the page related to a certain keyword. The calculation formula is:

[0104] L(N1m) = A(N1m) + B(N1m)

[0105] Where: L(N1m) represents the total number of views of the recommendation page corresponding to the keyword N1m; A(N1m) is the search times of the keyword; B(N1m) is the click times of the keyword.

[0106] In some embodiments, the system will also record the user's implicit behavior, such as the dwell time T of the user on the page stay . This data is statistically analyzed by a timer and collected at regular intervals.

[0107] In this embodiment, the specific implementation of the parameter dynamic adjustment module includes the following:

[0108] As an option, the system dynamically adjusts the optimization weight factor of the recommendation algorithm according to user behavior variables to make the recommendation results more in line with user needs. The logic of weight adjustment can be described as follows:

[0109] Specifically, the objective function of the recommendation algorithm includes three main optimization objectives: path energy U, geometric conflict C, and path complexity L. The comprehensive score formula is:

[0110] F(p i ) = αU(p i ) + βC(p i ) + γL(p i )

[0111] Where: F(p i ) represents the comprehensive score of path p i ; U(p i ) represents the total energy of path p i ; C(p i ) represents the number of geometric conflict nodes of path p i ; L(p i ) represents the complexity of path p i ;

[0112] α, β, and γ are the weight factors of path energy, geometric conflict, and path complexity respectively.

[0113] In a possible implementation, the adjustment of the weight factor is based on the following rules:

[0114] When the user clicks on low-energy paths more frequently, the system increases the value of the weight factor α, making the recommendation results more inclined to energy minimization.

[0115] When the user shows a higher interest in paths with fewer geometric conflicts, the system dynamically increases the value of β.

[0116] For users who prefer simple paths, the system increases the weight value of γ.

[0117] In this embodiment, the specific implementation of the data transmission module includes the following:

[0118] As an option, the user behavior data will be transmitted to the knowledge graph management module after being recorded for updating entity attributes and relationship weights. For example, when the click count C(N2l, k1) and the detail click count D(N2l, k1) of the user on content related to a certain keyword N2l reach a certain threshold, the system will add this keyword to the attributes of the corresponding entity k1 in the knowledge graph.

[0119] In another possible implementation, the behavior data is also transmitted to the multi-objective optimization recommendation algorithm module for dynamically updating the weight factors in the objective function.

[0120] Through the above implementation, the user behavior capture and feedback module realizes the comprehensive recording and dynamic feedback of user behavior data.

[0121] The multi-objective optimization recommendation algorithm module is used to generate recommendation results based on the multi-objective optimization of energy minimization, geometric conflict minimization, and path complexity minimization of the folding path;

[0122] Generally, a recommendation system needs to meet multiple optimization objectives. For example, in rigid origami recommendation, it is necessary to balance the relationship between path energy, geometric conflict, and path complexity. There are often contradictions between these objectives. Through the multi-objective optimization recommendation algorithm module, the optimal recommendation path is generated through comprehensive score calculation.

[0123] As an option, this module adopts the Pareto optimization method to screen out non-dominated solution paths and form a Pareto front. Specifically, based on the keywords input by the user, this module extracts the set of folding paths associated with them from the knowledge graph, calculates the optimization objective scores of each path, and finally generates the recommendation results sorted by priority.

[0124] In this embodiment, the specific calculation of the multi-objective optimization objectives includes the following:

[0125] In a possible implementation, the optimization objectives include path energy U, geometric conflict C, and path complexity L:

[0126] The path energy U is used to measure the energy consumption of the folding path, and the calculation formula is:

[0127]

[0128] Where: U(p i ) represents the total energy of path p i ; k j is the stiffness coefficient of crease j, reflecting the mechanical properties required during the folding process; θ j is the current angle of crease j; is the target angle of crease j; m is the total number of creases in the path.

[0129] The geometric conflict C represents the number of geometric conflicts occurring at the folding nodes in the path, and the calculation formula is:

[0130]

[0131] Where: C(p i ) represents path pi The number of geometric conflicts; δ j is a binary variable. When a geometric conflict occurs at node j, δ j = 1; otherwise, δ j = 0.

[0132] The path complexity L represents the total number of nodes in the path, and the calculation formula is:

[0133] L(p i ) = n

[0134] where: L(p i ) represents the complexity of path p i ; n is the total number of nodes in the path.

[0135] In this embodiment, the specific implementation of the comprehensive objective function includes the following:

[0136] As an option, the comprehensive objective function combines the above optimization objectives and is calculated through the following formula:

[0137] F(p i ) = αU(p i ) + βC(p i ) + γL(p i )

[0138] where: F(p i ) is the comprehensive score of path p i ; α, β, and γ are the weight factors of path energy, geometric conflict, and path complexity respectively, used to adjust the priorities of different optimization objectives.

[0139] In a possible implementation, the initial values of the weight factors are set by the system and dynamically adjusted through the user behavior capture and feedback module. For example, when the user pays more attention to path energy, the system will appropriately increase the value of α; when the user prefers a simple path, the system will increase the value of γ.

[0140] In this embodiment, the specific implementation of the Pareto optimization method includes the following:

[0141] Generally, there are multiple non-dominated solution paths for a multi-objective optimization problem. As an option, this module uses the Pareto optimization method to screen non-dominated solution paths, and the specific process is as follows:

[0142] Extract the path set P = {p1, p2,..., p k} associated with the user input keywords from the knowledge graph, and calculate the comprehensive score F(p i ) of each path. The condition for defining path p i as a non-dominated solution path is:

[0143] and F(p j ) ≠ F(p i )

[0144] where: p j represents any path in the path set p; F(p j ) represents the comprehensive score of path p j .

[0145] In some embodiments, the system also dynamically adjusts the sorting rules of the Pareto front according to the user's click behavior. For example, paths with higher historical user preferences are preferentially displayed.

[0146] In this embodiment, the generation and display of the recommendation results include the following:

[0147] As an option, the system generates recommendation results based on the non-dominated solution paths in the Pareto front and sorts them according to the comprehensive score. On the recommendation result interface, the system will display key information such as the energy value, geometric conflict number, and complexity of each path.

[0148] Specifically, when the user selects a recommended path, the system will call the dynamic physical modeling module to further verify it and update the verification result to the recommendation interface in real time.

[0149] Through the above implementation manner, the multi-objective optimization recommendation algorithm module realizes the comprehensive optimization of the recommended content.

[0150] The dynamic physical modeling module is used to verify the physical consistency of the recommendation results based on the geometric and physical characteristics of rigid origami;

[0151] The dynamic physical modeling module is used to verify the physical consistency of the recommendation results and dynamically adjust the parameters of the origami model. This module takes the recommended paths generated by the multi-objective optimization recommendation algorithm module as input, and combines the geometric characteristic data provided by the knowledge graph to verify whether the geometric and physical characteristics of the origami paths meet the requirements of rigid origami theory. At the same time, this module optimizes the actual feasibility of the paths by dynamically adjusting parameters such as crease stiffness and folding angle, thereby improving the accuracy and practicality of the recommendation results.

[0152] In this embodiment, the dynamic physical modeling module includes three main parts: geometric consistency verification, dynamic adjustment of physical parameters, and physical constraint verification.

[0153] Generally, the folding paths of rigid origami need to meet geometric consistency and mechanical rationality. For example, the folding surface needs to maintain rigid rotation characteristics, and the crease angle needs to meet the constraints of Kawasaki's theorem. As an option, this module realizes these verification and adjustment functions through mathematical modeling and dynamically optimizes the parameters of the recommended paths in combination with user behavior data.

[0154] In a possible implementation, the system calls this module to perform geometric and physical verification on the recommended paths one by one, eliminates the paths that do not meet the constraint conditions, and generates a better path through parameter optimization.

[0155] In this embodiment, the specific implementation of the geometric consistency check includes the following:

[0156] Specifically, this module first verifies the rigid rotation characteristics of each folding surface in the origami path. The rigid rotation of the folding surface can be verified by the following matrix formula:

[0157] M i RM j = M j

[0158] Where: M i represents the i-th folding surface in the folding path, and represents the position and direction of the surface in matrix form; M j represents the j-th folding surface; R is the rigid rotation matrix, which describes the rotation relationship between two folding surfaces.

[0159] As an option, this module utilizes the characteristics of matrix calculation to ensure the rigid condition of the rotation transformation by verifying whether R is an orthogonal matrix and its determinant is 1.

[0160] Generally, the sum of the crease angles needs to satisfy the Kawasaki theorem. Specifically, this module verifies whether each folding node in the path meets the following relationship:

[0161]

[0162] Where: θ k represents the angle of the k-th crease in the folding path; the odd angles and even angles respectively point to the sets of angles on both sides of the folding node.

[0163] Through the above verification, the system can eliminate the paths that do not meet the geometric consistency, thus ensuring the scientific nature of the recommended content.

[0164] In this embodiment, the specific implementation of the dynamic adjustment of physical parameters includes the following:

[0165] As an option, the system dynamically adjusts the crease stiffness k j and the target angle according to the user behavior and path characteristics. The specific adjustment rules are as follows:

[0166] Generally, the crease stiffness k j reflects the mechanical characteristics required during the folding process, and its adjustment formula is:

[0167]

[0168] Wherein: is the initial crease stiffness; Δk j is the stiffness increment dynamically adjusted according to user behavior, and the calculation formula is:

[0169] Δk j = η·T stay

[0170] Wherein: η is the adjustment coefficient, usually taking a value of 0.1; T stay is the residence time of the user on the recommended path page.

[0171] Target angle The adjustment is based on the current deviation of the recommended path, and the calculation formula is:

[0172]

[0173] Wherein: θ j is the current crease angle; Δθ j is the dynamically adjusted angle, which is determined according to the geometric conflict situation of the nodes in the folding path.

[0174] In a possible implementation manner, when the user selects a certain folding path type multiple times, the system will make a greater adjustment to the parameters of this path to better fit the user's preferences.

[0175] As an option, this module re-verifies the physical constraint conditions of the path based on the dynamically adjusted path parameters. For example, the system recalculates the total energy and the number of geometric conflicts of the adjusted path.

[0176] The verification method of geometric conflict is the same as the aforementioned geometric consistency verification. If the physical constraint verification of the path fails, the system will automatically eliminate it and hide the relevant content in the recommendation interface.

[0177] In this embodiment, the physical optimization of the recommendation result includes the following content:

[0178] In the optimized path, the system will preferentially display the paths with lower total energy and fewer geometric conflicts. Specifically, when the user clicks on a certain recommended path, the system will call the dynamic physical modeling module to verify it and update the physical parameters of the recommendation content in real time.

[0179] In some embodiments, the system also supports the user to manually adjust the crease stiffness and the target angle to verify the feasibility of the folding path under different parameters.

[0180] Through the above implementation, the dynamic physical modeling module realizes the geometric and physical consistency verification of the recommended path and dynamic parameter adjustment, which not only ensures the scientific nature and practical usability of the recommendation results, but also provides users with more accurate recommended content.

[0181] The user interaction and visualization interface module is used to provide a search function and display of recommendation results, and receive real-time feedback from users;

[0182] The user interaction and visualization interface module provides users with a search entry, display of recommended content, and real-time feedback function, and presents the core information of the recommendation results in an intuitive visualization manner. This module is closely associated with the knowledge graph management module, the multi-objective optimization recommendation algorithm module, and the dynamic physical modeling module. By calling the data and results of these modules, it dynamically adjusts the content and recommendation logic of the user interface.

[0183] In this embodiment, the user interaction and visualization interface module includes three main functional parts: a search interface, a recommended result display interface, and a real-time feedback interface.

[0184] Generally, users input keywords through the search interface, and the system calls the knowledge graph management module to retrieve relevant entities, and combines with the multi-objective optimization recommendation algorithm module to generate preliminary recommendation results. The recommendation results are grouped and displayed through the recommended result display interface. The user's interaction behaviors (such as clicks, ratings, etc.) are recorded in real time and fed back to the user behavior capture and feedback module to further optimize the system recommendation logic.

[0185] As an option, the user interface is centered on modular design, which can not only adapt to the usage habits of different users, but also flexibly adjust the structure and content of the interface.

[0186] In a possible implementation, the system realizes the efficient interaction between user behavior data and recommendation logic through a dynamic response mechanism. For example, when a user selects a certain recommended path, the system can immediately update the display priority of other recommended paths.

[0187] In this embodiment, the specific implementation of the search interface includes the following:

[0188] As an option, the search interface provides two main interaction methods: keyword input and entity type filtering.

[0189] Specifically, users can input keywords in the search box, and the system performs word segmentation on the input content through natural language processing technology to extract the keyword set N = {N1, N2, …, N n}. If the user selects an entity type, the search scope will be limited within that type. For example, when the user selects "geometric characteristics" as the entity type, the system only retrieves the entities related to geometric characteristics in the knowledge graph.

[0190] In a possible implementation, the system provides a keyword completion function for users. When the user enters some keywords, the system preferentially recommends keywords with higher popularity according to the keyword popularity L(N1m). The calculation formula for keyword popularity is:

[0191] L(N1m) = A(N1m) + B(N1m)

[0192] where: L(N1m) is the popularity of keyword N1m; A(N1m) is the number of searches for keyword N1m; B(N1m) is the number of clicks on keyword N1m.

[0193] Generally, the higher the keyword popularity, the higher the user's attention, and the system will preferentially display these keywords.

[0194] In this embodiment, the specific implementation of the recommended result display interface includes the following content:

[0195] The recommended result display interface sorts the recommended content according to the comprehensive score generated by the multi-objective optimization recommendation algorithm module, and groups and displays the recommended results.

[0196] As an option, the recommended results can be grouped and displayed according to three dimensions: path energy, geometric conflict, and path complexity. For example, the recommended results can be divided into three major categories: "low-energy path recommendation", "low-conflict path recommendation", and "simple path recommendation", and the recommended results within each category are sorted in descending order of the comprehensive score.

[0197] Specifically, each recommended result is displayed in the form of a card, and the following information is included in the card:

[0198] The core parameters of the folded path, including path energy U(p i ), geometric conflict number C(p i ) and path complexity L(p i );

[0199] The relevance G(N1m,k1) to the user's search keyword, and its calculation formula is:

[0200]

[0201] where: G(N1m,k1) is the relevance between keyword N1m and recommended content k1; D(N1m,k1) is the number of clicks on the detail page; C(N1m,k1) is the total number of clicks on the recommended content.

[0202] Generally, the recommended content with higher relevance is preferentially displayed at the top of the list.

[0203] In a possible implementation, the user can further adjust the display order of the recommended results through the filtering button on the right. For example, the user can select "Prioritize the display of paths with the fewest geometric conflicts", and the system will reorder and update the recommendation interface.

[0204] In this embodiment, the specific implementation of the real-time feedback interface includes the following:

[0205] As an option, the real-time feedback interface allows users to rate, comment on, and mark the recommended content.

[0206] Specifically, the user can click the rating button on the right side of each recommended path to rate the path from 1 to 5. The rating data will be transmitted to the user behavior capture and feedback module in real time and used to dynamically adjust the recommendation logic. For example, when the average rating of a certain path is high, the system will increase the display priority of this path.

[0207] In some embodiments, the system also supports users to mark the paths that do not meet the requirements. The marked paths will be removed from the recommended results and trigger the knowledge graph management module to retrieve and update the relevant entities of this path.

[0208] In a possible implementation, the real-time feedback of the user will also trigger the dynamic physical modeling module to re-verify the physical consistency of the path. For example, when the user feedbacks that a certain path is not feasible in actual application, the system will check the geometric and physical parameters of this path to confirm the problem.

[0209] Through the above implementation, the user interaction and visualization interface module realizes the full-process closed-loop design of search, recommended result display, and real-time feedback, which not only improves the user experience but also optimizes the recommendation logic of the system through real-time data interaction.

[0210] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Rigid origami recommendation system, characterized in that include: The knowledge graph management module is used to build and maintain the knowledge graph of rigid origami, including product entities, patent entities, and geometric property entities, and dynamically update the attributes and relationships of entities; User behavior capture and feedback module, used to record user behavior data and dynamically adjust recommendation parameters based on user feedback; Multi-objective optimization recommendation algorithm module, which is used to generate recommendation results based on multi-objective optimization of folding path energy minimization, geometric conflict minimization and path complexity minimization; A dynamic physical modeling module to verify the physical consistency of the recommended results based on the geometric and physical properties of rigid origami; The user interaction and visualization interface module is used to provide search functions and recommendation result display, and receive real-time feedback from users.

2. The rigid origami recommendation system according to claim 1, characterized in that: The knowledge graph management module includes: The entity definition module is used to define product entities, patent entities, and geometric property entities in the knowledge graph; The attribute update module is used to dynamically add unmatched keywords to the attributes of the entity through user behavior data; The relationship update module is used to update the relationship weights between entities based on the number of user clicks and keyword relevance.

3. The rigid origami recommendation system according to claim 1, characterized in that: The user behavior capture and feedback module includes: Behavior recording module, used to record the number of searches and clicks by users; The parameter adjustment module is used to dynamically adjust the optimization weight factor in the recommendation algorithm according to the user's click behavior; The data transmission module is used to transmit user behavior data to the knowledge graph management module and the recommendation algorithm module.

4. The rigid origami recommendation system according to claim 1, characterized in that: The multi-objective optimization recommendation algorithm module includes: A path energy calculation module is used to calculate the energy of each folding path; A geometric conflict detection module is used to detect geometric conflict nodes in the path; Path complexity analysis module, used to calculate the total number of nodes in the path; A comprehensive score calculation module, used to calculate the comprehensive score based on the weights of energy, geometric conflict and path complexity; The optimization model execution module is used to generate recommendation results through non-inferior solution path screening.

5. The rigid origami recommendation system according to claim 1, characterized in that: The dynamic physical modeling module includes: A geometric consistency checking module is used to check the geometric consistency of the folding path based on the rigid rotation matrix and the angle theorem; A parameter dynamic adjustment module is used to dynamically adjust the crease stiffness parameters and target folding angle parameters based on user behavior; The physical constraint verification module is used to verify the folding constraints in the recommended path.

6. The rigid origami recommendation system according to claim 1, characterized in that: The user interaction and visualization interface module includes: The search interface module is used to allow users to input search content and select entity types; The recommendation result display module is used to display the recommended content in groups according to the optimization target; The real-time feedback module is used to receive the user's rating of the recommendation results and pass it to the feedback module.

7. The rigid origami recommendation system according to claim 2, characterized in that: The attribute update module is completed by the following steps: Get the search content entered by the user and perform word segmentation; Extract unmatched keywords and record the number of clicks and details of the keyword; Calculate the relevance of the keyword. When the relevance exceeds the preset threshold, add the keyword to the attribute of the corresponding entity in the knowledge graph.

8. The rigid origami recommendation system according to claim 4, characterized in that: The comprehensive score calculation module is completed by the following steps: For each crease of the path, calculate the deviation between the current angle and the target angle; The total energy of the path is calculated based on the crease stiffness; Detect geometric conflict nodes in the path and record the number of conflicts; Calculate the total number of nodes in the path; Based on the path energy, the number of conflicting nodes and the total number of nodes, the comprehensive score is calculated according to the preset weights.

9. The rigid origami recommendation system according to claim 5, characterized in that: The parameter dynamic adjustment module is completed by the following steps: Dynamically adjust the fold stiffness parameters based on the user's stay time and click behavior on the recommendation result page; Update the target folding angle parameter according to the search content input by the user and the angle deviation of the associated path; Apply the adjusted parameters to the physical constraint verification module.

10. The rigid origami recommendation system according to claim 6, characterized in that: The recommendation result display module is completed by the following steps: Filter recommendations by entity type selected by the user; Group and sort recommended content according to optimization goals; The recommended results interface displays detailed information about the recommended content, including path complexity, geometric conflicts, and related crease angles.

Images