A Material Creation Method and System Based on Intelligent Recommendation of Similar Materials

Abstract

This invention discloses a material creation method and system based on intelligent recommendation of similar materials, relating to the fields of information technology application and data processing. The method includes: acquiring multi-type attribute information of the material to be created; performing a two-stage similarity calculation and sorting of existing materials in the database (coarse sorting and AI assistant fine sorting); and filtering out recommended similar materials. The coarse sorting involves filtering a candidate set from the database using single-attribute filtering and similarity calculation based on an improved edit distance algorithm. The AI ​​assistant fine sorting involves performing differentiated similarity calculations for different types of attributes on each existing material in the candidate set, obtaining a comprehensive similarity through dynamic weight fusion, and then performing a secondary sorting based on the priority of existing material status. The method also includes: filtering the recommended results according to user permissions and providing feedback; and responding to the user's selection operation command by executing a direct reference, reference creation, or continued creation process based on the selected similar materials to complete the material creation.

Description

Technical Field

[0001] This invention relates to the fields of information technology applications and data processing, and in particular to a material creation method and system based on intelligent recommendation of similar materials. Background Technology

[0002] In the process of digital transformation in the manufacturing industry, Product Lifecycle Management (PLM) systems have become the core platform for enterprises' digital material management. Material creation, as the initial stage of product development, directly impacts product development cycles and enterprise operating costs due to its efficiency and data standardization. Currently, when users create new materials in the system, facing tens of thousands of material data entries, how to fully utilize existing materials to create new materials while avoiding the duplication of identical materials is a long-standing problem for manufacturing enterprises and PLM vendors. Issues such as duplicate material creation and chaotic coding directly lead to increased costs and reduced R&D efficiency for manufacturing enterprises, resulting in problems such as inventory backlog, increased procurement costs, and redundant design.

[0003] To avoid duplicate material creation, existing technologies propose intelligent recommendation strategies for material creation. These strategies calculate the descriptive language similarity between the material to be created and existing materials in the system, determining whether to create the material repeatedly or provide similar materials for selection based on the similarity score. However, existing solutions typically use only a single-stage, brief description similarity comparison for matching. This results in low response efficiency in scenarios with massive amounts of material data, poor matching accuracy, and a tendency to encounter issues such as duplicate creation, matching failures, or matching materials with significant differences, ultimately impacting the efficiency of material creation. Summary of the Invention

[0004] To address the shortcomings of the existing technologies, this invention provides a material creation method and system based on intelligent recommendation of similar materials. Through a two-stage similarity calculation and sorting process involving coarse database ranking and fine ranking by an AI assistant, it quickly provides high-precision similar materials to assist in material creation. This solves the problems of easy duplication of material creation, slow response and poor accuracy of similar material recommendations in the existing technologies, which lead to low material creation efficiency and poor results.

[0005] In a first aspect, the present invention provides a material creation method based on intelligent recommendation of similar materials, comprising: Respond to the user's material creation command and obtain multi-type attribute information of the material to be created; Based on the information obtained, a two-stage similarity calculation and sorting process is performed on the existing materials in the database: coarse sorting and fine sorting by AI assistant, to filter out recommended results for similar materials. The coarse sorting process involves filtering candidate materials from the database through single attribute filtering and similarity calculation based on an improved edit distance algorithm. The fine sorting process involves performing differential similarity calculations on each existing material in the candidate set for different types of attributes, obtaining a comprehensive similarity through dynamic weight fusion, and then performing a second sorting based on the priority of the existing material status. The recommended results are filtered based on user permissions and then fed back. Responding to the user's selected operation command, the system executes the process of direct reference, reference creation, or continued creation based on the selected similar materials to complete the material creation.

[0006] A further technical solution is that the multiple types of attributes include: Text-based attributes include name, material description, and function description; Numerical attributes include weight, length, width, height, dimensions, and specifications. Category-based attributes include type, material group, and lifecycle status; Collective attributes include tag sets, applicable scenario sets, and process feature sets; Material attributes include material grade, material category, and material standard number.

[0007] A further technical solution involves setting up an attribute missing tolerance mechanism, which allows intelligent recommendations of similar materials to be triggered by acquiring only some attributes.

[0008] A further technical solution involves the following coarse-grained database sorting process: Based on the type and attribute information of the material to be created input by the user, the existing materials in the database are initially filtered; An improved edit distance algorithm integrated with CLR is used to calculate the multi-attribute similarity between existing materials and materials to be created after initial filtering. The calculation results are converted into similarity percentages and then weighted and summed. Specifically, for name, material description, functional description, type, and material attributes, the edit distance algorithm is used to calculate the minimum number of edit operations between two strings to obtain the edit distance. Then, according to the calculation formula: similarity = 100 × (1 - edit distance / max(string 1 length, string 2 length)), the edit distance is converted into a similarity percentage. The top N existing materials are selected as the material candidate set based on their calculated scores in descending order.

[0009] Further technical solutions, the AI ​​assistant's fine-sorting process is as follows: For each existing material in the candidate set, calculate the differential similarity between it and the material to be created for different types of attributes; among them, textual attributes are calculated using a fuzzy matching algorithm, numerical attributes are calculated based on numerical differences, categorical attributes are calculated based on data value matching, set attributes are calculated using the Jaccard similarity coefficient, and material attributes are calculated based on the material equivalence knowledge base. Based on the preset weight configuration table, the similarity of different types of attributes is dynamically weighted and fused to obtain the comprehensive similarity. The existing materials in the candidate set are sorted based on the comprehensive similarity. When the interpolation of the comprehensive similarity is less than a preset threshold, the existing materials are sorted in conjunction with their state priority to obtain the final sorting result. The top M existing materials are then selected as the recommended similar materials.

[0010] Further technical solutions, including dynamic weight configuration, include: Configure attribute weight coefficients according to material type, business characteristics, and project stage; When an attribute is missing, the weights of the acquired attributes of the material to be created are automatically normalized. The priority ranking rules for material status include: Lifecycle status priority: Released > Under Design > Under Review > Obsolete; Frequency priority under the same conditions: high frequency use > low frequency use > no use; Creation time priority under the same conditions: Recently created > Early created.

[0011] Further technical solutions include filtering the recommendation results based on user permissions and providing feedback, including: Obtain the current user's role, department, and project group permissions, and simultaneously perform batch verification of the access permissions for each existing material in the recommendation results; After matching user and material permissions, materials without permissions are removed from the recommended results, and materials with permissions are added from subsequent rankings. The system then returns and displays similar materials that the current user has the right to access.

[0012] Secondly, the present invention provides a material creation system based on intelligent recommendation of similar materials, comprising: The data acquisition module is used to respond to the user's material creation command and obtain multi-type attribute information of the material to be created; The intelligent recommendation module for similar materials is used to perform a two-stage similarity calculation and sorting of existing materials in the database based on the acquired information, namely, coarse sorting and AI assistant fine sorting, to filter out the recommended results of similar materials. The coarse sorting is: filtering candidate sets from the database through single attribute filtering and similarity calculation based on an improved edit distance algorithm; the AI ​​assistant fine sorting is: performing differential similarity calculations for different types of attributes on each existing material in the candidate set, obtaining a comprehensive similarity through dynamic weight fusion, and then performing a second sorting based on the status priority of existing materials. The display module is used to filter the recommendation results based on user permissions and provide feedback. The material creation module is used to respond to the user's selected operation command, and execute the process of direct reference, reference creation or continued creation based on the selected similar materials to complete the material creation.

[0013] Thirdly, the present invention also provides an electronic device, comprising: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the above-described material creation method based on intelligent recommendation of similar materials.

[0014] Fourthly, the present invention also provides a computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the above-described material creation method based on intelligent recommendation of similar materials.

[0015] Fifthly, the present invention also provides a computer program product comprising executable instructions stored in a computer-readable storage medium; wherein, when the processor of an electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the aforementioned material creation method based on intelligent recommendation of similar materials is implemented.

[0016] The above one or more technical solutions have the following beneficial effects: This invention provides a material creation method and system based on intelligent recommendation of similar materials. It employs a two-stage intelligent similarity calculation and sorting approach combining database coarse ranking and AI fine ranking. This, coupled with material equivalence knowledge base recognition, multi-dimensional dynamic weight configuration, attribute missing tolerance, full-process permission filtering, and deep reuse of engineering data in a reference creation mode, significantly shortens the recommendation response time for massive amounts of material data. It also greatly improves the accuracy of similar material recommendations and material matching precision, solving the problems of easy duplication of material creation, slow response time for similar material recommendations, and poor accuracy leading to low material creation efficiency and poor results in existing technologies. This invention can adapt to the diverse needs of different enterprises, different material types, and different R&D stages, supporting effective recommendations throughout the entire process from conceptual design to detailed design. It also strictly ensures data access security and greatly shortens the new material design cycle, improves design data reuse rate and R&D efficiency through mechanisms such as intelligent document copying / reference, BOM structure relationship reuse, and rapid inheritance of processes and attributes. This avoids duplicate creation and multiple codes for a single item from the source, reduces inventory backlog, procurement and management costs, and optimizes enterprise material data governance.

[0017] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0018] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0019] Figure 1 This is an overall flowchart of the material creation method based on intelligent recommendation of similar materials in Embodiment 1 of the present invention. Detailed Implementation

[0020] It should be noted that the following detailed descriptions are exemplary and are intended only to describe specific embodiments and to provide further explanation of the invention, and are not intended to limit the scope of exemplary embodiments of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0021] Example 1 This embodiment provides a material creation method based on intelligent recommendation of similar materials, such as... Figure 1 As shown, the specific steps include: Step S1: Respond to the user's material creation command and obtain the multi-type attribute information of the material to be created.

[0022] Specifically, the PLM system responds to the user's material creation command by displaying an interface with multiple types and attribute fields for the user to fill in the relevant information of the material to be created; the system receives at least one key attribute information input by the user, which includes, but is not limited to, name, material description, function description, number, type, weight, material, etc.

[0023] As one implementation method, this embodiment sets multiple types of attributes, including text-based, numerical, categorical, set-based, and material-based attributes. Text-based attributes include name, material description, and functional description; numerical attributes include weight, length, width, height, size, and specification parameters; categorical attributes include type, material group, and lifecycle status; set-based attributes include tag set, applicable scenario set, and process feature set; and material-based attributes include material grade, material category, and material standard number.

[0024] This embodiment supports an attribute missing tolerance mechanism, meaning users can fill in only partial attribute information. Obtaining at least one attribute is sufficient to trigger intelligent recommendations for similar materials. The PLM system will calculate similarity based on the filled-in attributes, while missing attributes will not be included in the calculation. This mechanism allows users to obtain effective recommendation results even in the early stages when material requirements are not fully clear.

[0025] Step S2: Based on the obtained information, perform a two-stage similarity calculation and sorting of the existing materials in the database, including coarse sorting and fine sorting by the AI ​​assistant, and filter out the recommended results of similar materials.

[0026] Specifically, considering that the PLM system stores tens of thousands of material data, to improve matching efficiency, this embodiment designs a two-stage sorting and filtering process: database coarse sorting and AI assistant fine sorting. First, coarse sorting filters existing materials in the database based on a single, clear attribute (such as type attribute) and similarity calculation using an improved edit distance algorithm, narrowing the matching range and improving calculation and matching efficiency. Then, AI assistant fine sorting calculates differentiated similarity for each existing material in the candidate set based on different type attributes. A comprehensive similarity is obtained through dynamic weight fusion, and a second sort is performed based on the priority of existing material status. Based on the final sorting result, similar material recommendations are selected, improving the accuracy of similar material selection. The specific process of this step is as follows: Step S2.1: Coarse sorting of the database.

[0027] First, based on a specific attribute information filled in by the user, such as the "type" attribute, the existing materials in the database are initially filtered to narrow down the matching range.

[0028] Then, within the filtered data range, the similarity of specific attributes is calculated using the database's own similarity algorithm; or, if the database does not support a similarity algorithm, a custom similarity algorithm integrated into the Common Language Runtime (CLR) is introduced into the database for similarity calculation.

[0029] In this embodiment, the custom similarity algorithm integrated by CLR is an improved edit distance algorithm. The basic edit distance algorithm is as follows: the Levenshtein distance algorithm (or edit distance algorithm) is used to calculate the minimum number of edit operations between two strings, including insertion, deletion, and replacement. Based on this, the algorithm is used to calculate the multi-attribute similarity between the existing materials obtained from the initial filtering and the materials to be created. For example, the edit distance between the materials to be created and the existing materials is calculated for attributes such as name, material description, function description, type, and material. The calculation result (i.e., edit distance) is then converted into a similarity percentage. The similarity percentage conversion formula is: similarity (%) = 100 × (1 - edit distance / max(string 1 length, string 2 length)). Finally, the edit distances corresponding to each attribute are weighted and summed to generate a preliminary comprehensive similarity score.

[0030] Finally, the preliminary candidate materials are sorted in descending order based on the initial comprehensive similarity scores, and the first N records (e.g., 200 records) are usually selected for the fine sorting stage.

[0031] The coarse sorting stage described above makes full use of the database's computing power to quickly filter out highly relevant candidate sets from massive amounts of material data (potentially tens of thousands to hundreds of thousands of records), significantly improving the system's response speed.

[0032] Step S2.2: AI assistant fine sorting.

[0033] Specifically, the coarsely sorted candidate material list is sent to the AI ​​assistant (or AI module). The AI ​​assistant uses a more precise fuzzy matching algorithm (such as the FuzzyWuzzy algorithm) or other natural language processing algorithms to perform secondary calculations and fine sorting of the candidate material attributes.

[0034] First, for each existing material in the candidate set, a differentiated calculation strategy is adopted for attributes of different data types to calculate the differentiated similarity between its attributes and those of the material to be created, specifically including: 1) For numerical attributes (such as weight, length, width, height, size, specifications, etc.): First, normalize them to eliminate the influence of dimensions, and then calculate the similarity based on the difference between the standardized values. The calculation formula is: similarity = 1 - |value1 - value2| / max(value1, value2).

[0035] 2) For categorical attributes (such as type, material group, lifecycle status, etc.): similarity is determined based on the exact match of its data values. If there is a perfect match, the similarity is 1.0, otherwise it is 0.

[0036] 3) For set-type (or tag-type) attributes (such as tag set, applicable scenario set, process feature set, etc.): Jaccard similarity coefficient is used to quantify the degree of overlap between two sets. The calculation formula is: Jaccard = |intersection| / |union|.

[0037] 4) For text-based attributes (such as names, material descriptions, and functional descriptions): Fuzzy matching algorithms, such as FuzzyWuzzy's token_set_ratio method, are used to calculate the similarity of character matching. This algorithm can handle situations such as word order changes and missing words.

[0038] 5) For material attributes (such as material grade, material category, material standard number, etc.): A material equivalence knowledge base is introduced. This knowledge base maintains multiple representations of materials and equivalence relationships. Similarity is calculated based on the material equivalence knowledge base. For example: 304 stainless steel ≈ SUS304 ≈ 0Cr18Ni9 ≈ 1.4301. When calculating material similarity, if two materials are equivalent in the knowledge base, the similarity = 1.0; if they belong to the same material category, such as both being stainless steel, the similarity = 0.7; if they belong to different material categories but have similar characteristics, such as both being high-strength steel, the similarity = 0.4; if they are completely different categories, the similarity is calculated based on text edit distance, and the similarity < 0.3.

[0039] Preferably, the aforementioned material knowledge base can be expanded and maintained according to the actual usage of the enterprise, supporting material representation methods of different industries and different standard systems.

[0040] By introducing a material equivalence knowledge base for similarity calculation in the manufacturing PLM field, this embodiment can accurately identify the equivalence, category, and characteristic relationships of materials through knowledge base mapping, compared to traditional text matching methods that cannot identify "304 stainless steel" and "SUS304" as the same material. This further improves the accuracy of material matching. The introduction of this knowledge base significantly improves the recommendation effect in material-intensive industries (such as machinery manufacturing and chemical industry).

[0041] Secondly, based on a preset weight configuration table, the similarity of different types of attributes is dynamically weighted and fused to obtain a comprehensive similarity score. In this embodiment, the weight coefficients of each attribute are not fixed values, but are flexibly managed and dynamically configured through the weight configuration table. Specifically, the dynamic weight configuration mechanism includes: 1) Assign different weights based on material type: Different types of materials (such as products, parts, materials, standard parts, etc.) have different attributes of importance. For example: Standard parts: Specifications (0.4) > Material (0.3) > Name (0.3); Customized parts: Name (0.5) > Functional description (0.3) > Dimensions (0.2); Material: Material (0.6) > Specifications (0.3) > Name (0.1).

[0042] 2) Configure according to customer business characteristics: Different enterprises and industries have different focuses on material attributes. The system allows customers to customize weights according to their own business characteristics.

[0043] 3) Material selection is configured according to project stage: The focus of material selection differs at different stages of product development. For example: in the conceptual design stage, functional description has a higher weight; in the detailed design stage, size and material have a higher weight; and in the mass production stage, cost and supplier have a higher weight.

[0044] Preferably, the priority (or hierarchy) of the above multiple weight configuration methods is preset, and the weight configuration is executed sequentially according to the preset priority or hierarchy. That is: based on the information of the material to be created, the system first determines whether there is project stage information. If there is, the weight is configured according to the corresponding project stage information. If not, the system then determines whether there is a preset customized weight. If the weight is pre-customized according to the customer's business characteristics, the weight is configured accordingly. Otherwise, the system proceeds to the next level of configuration, that is, the weight is configured according to the material type.

[0045] Through the multi-dimensional and customizable weight configuration system designed above, it supports differentiated configuration based on material type, customer business characteristics, and project stage. It also automatically performs weight normalization adjustment when the user only fills in some attributes. This mechanism enables the recommendation system to adapt to the diverse needs of different enterprises, industries, and application scenarios, and has strong versatility and flexibility.

[0046] As one implementation method, when attributes are missing, the dynamic weight adjustment is as follows: when the user only fills in part of the attributes, the system automatically normalizes the weights of the filled attributes to ensure that the total weight is 1.0. For example, if the user only fills in the name and material attributes, the weights of these two attributes are automatically amplified proportionally to keep the total weight at 1.0.

[0047] Based on the clearly defined weight coefficients of each attribute, the similarity scores of all individual attributes are weighted and summed according to their preset business weights. The calculation formula is as follows: Overall similarity = Σ(similarity of attribute i × weight of attribute i); Where i represents all attributes involved in the calculation, and the sum of their weights is 1.0.

[0048] Finally, the existing materials in the candidate set are ranked based on the comprehensive similarity to obtain the final recommended ranking result.

[0049] As a further implementation, a material state priority strategy is introduced in the above sorting process. That is, when the overall similarity is the same or close (the difference is less than a preset threshold, such as 0.05), the material state priority is further used for sorting, including: 1) Sorted by lifecycle status: Released > Under Design > Under Review > Obsolete; 2) Sorted by usage frequency: High-frequency use > Low-frequency use > Unused; 3) Sort by creation time: Recently created > Early created (reflecting the timeliness of the design).

[0050] Preferably, when sorting according to the priority of material status, the above three sorting strategies can be implemented simultaneously. For example, the OrderBy sorting method (i.e., sorting the set in ascending / descending order by a specified key) can be used during the query to sort in descending order by life cycle status, usage frequency, and creation time. Alternatively, one or two of the conditions can be retained for sorting according to customization. This embodiment does not specifically limit the conditions, which can be customized according to the actual situation.

[0051] The above prioritization strategy ensures that the recommended results are not only highly similar, but also more in line with the usage habits in engineering practice, giving priority to mature and stable material data.

[0052] The above-described architecture, which combines a database-level improved edit distance algorithm for coarse ranking and an AI application-level multi-algorithm fusion for fine ranking, fully utilizes the computing power of the database to quickly filter large-scale data (processing tens of thousands of data entries), and then uses AI for precise sorting. This approach balances response speed and recommendation accuracy. Compared to single-level similarity calculation methods, its response time can be shortened by more than 60%, while the recommendation accuracy can be improved by 30%.

[0053] Step S3: Filter the recommendation results according to user permissions and provide feedback.

[0054] Specifically, the refined material list (i.e., the recommendation results) is paginated, and the system batch checks whether the current user has permission to access each recommended material. The system only displays the material data that the user has permission to view, ranked from highest to lowest similarity, on the front-end interface. The permission verification mechanism is as follows: First, the system queries the user's role, department, project team, and other permission information, and batch verifies the access permission settings for each recommended similar material, such as confidentiality level and departmental restrictions. Then, after matching user and material permissions, materials that the user does not have permission to access are removed from the recommendation results, and authorized material entries are added from subsequent rankings. The system then returns and displays the similar materials that the current user has permission to access, or directly returns the actual number of visible materials. The system only displays the material data that the user has permission to view, ranked from highest to lowest similarity, on the front-end interface.

[0055] As a further implementation, the system also displays or provides the following in the recommendation results: similarity percentage (e.g., 85% match), a list of matched attributes (which attributes match and which do not), a preview of key information about the material (code, name, drawing thumbnail, etc.), and the status and usage of the material (published, number of times used, etc.).

[0056] Step S4: Respond to the user's selected operation command, and execute the direct reference, reference creation, or continue creation process based on the selected similar materials to complete the material creation.

[0057] Specifically, users make judgments and take actions based on the received recommendation list, including: 1) Direct Reference: If any of the recommended materials fully meet the requirements, the user can choose to directly reference the material. At the same time, the system will interrupt the creation process of new materials and apply the existing material selected by the user to the current business scenario (such as adding it to the BOM structure, associating it with drawings, etc.).

[0058] 2) Reference Creation: If none of the recommended materials are completely matching, but highly similar materials exist, users can select "Reference Creation." In reference creation mode, the system provides an intelligent data reuse mechanism, including: 2.1) Document Classification, Recognition, and Differentiated Processing. The system automatically identifies the nature of documents based on document type tags, as follows: Related documents (such as design drawings, 3D models, technical specifications, etc.): These documents are closely related to the specific materials. New materials usually require modification of these documents. Therefore, the default method is to use "copy" to create a copy of the document for the user to modify. Reference documents (such as national standards, industry specifications, inspection standards, general processes, etc.): These documents are universal and can be shared by multiple materials. Therefore, the "reference" method is used by default. New materials are directly associated with the original document without the need to create a copy.

[0059] Ideally, users can view the system's recommended methods on the interface and adjust them according to their actual needs.

[0060] 2.2) BOM Structure Relationship Copying Mechanism. The system copies the "structural relationships" of the BOM, rather than specific material objects. The copied relationships include: parent-child hierarchy (which materials are subordinate components), quantity relationships (the usage of each subordinate material), location relationships (position number in the assembly), and substitution relationships (substitutable materials). Preferably, subordinate material objects themselves still point to the original materials, without creating copies; users can selectively replace certain subordinate materials in the BOM editing interface of the new material.

[0061] 2.3) Selective reuse of process data. This reuse includes: for process cards, users can choose to copy or reference them; for process routes, the route structure can be copied, but the operations can be adjusted individually; for time quotas, they can be selectively inherited.

[0062] 2.4) Pre-population of attribute data. This pre-population includes: automatically pre-populating most of the attribute values ​​of the reference material in the new material creation form; requiring users to modify only the differentiated attributes (such as size, specifications, etc.); and highlighting which attributes have been pre-populated to prompt users to check.

[0063] Based on the above-mentioned intelligent data reuse mechanism, the complete process of reference creation is as follows: First, the user selects a highly similar material from the recommended list and clicks the "Reference Creation" button; then, the system pops up the "Reference Creation Configuration" interface, displaying a list of associated data for the referenced material; the user selects a processing method (copy / reference / not import) for each type of data, and the system provides a default recommendation; after the user confirms, the system jumps to the new material creation interface, which is pre-filled with a large amount of information; the user modifies the differentiating attributes and completes the creation; finally, the system automatically establishes the association relationship and copies the relevant data.

[0064] The aforementioned reference establishment mechanism allows users to selectively and deeply reuse associated engineering data of similar materials when creating new materials. For example, it automatically determines whether to copy or reference based on document type tags, avoiding unnecessary document copies and potential document consistency issues. BOM relationship copying only copies structural information and not objects, ensuring design flexibility while avoiding data redundancy. This achieves a leap from "avoiding duplication" to "efficient reuse," enabling designers to quickly reuse design documents and process data of existing mature materials, greatly shortening the design cycle of new materials.

[0065] 3) Continue creating: If the user believes that none of the recommended materials meet their needs, they can ignore the recommendations and continue to create entirely new materials according to the normal process.

[0066] The superiority of the proposed solution is further verified through the following specific example. Specifically, the user enters the "Create Component" function module in the PLM system. The system displays an interface containing multiple attribute fields, such as: name (e.g., hex bolt), number (e.g., automatically generated or manually entered), type (enumerated value, select "Part"), category (e.g., fastener), weight (e.g., 0.05kg), and material (e.g., 304 stainless steel). The user fills in the four attributes—type, category, name, and material—according to their needs, and then clicks the "Smart Recommendation" button.

[0067] Next, the similarity calculation and ranking are performed in two stages: coarse ranking and fine ranking. In the coarse ranking stage, after receiving the four attribute values ​​input by the user, the system first performs a rapid filter at the database level, including: The first step is to perform initial filtering based on the "Type" attribute. The system executes a query in the material master data table to filter out all records where "Type = Part," excluding other material types such as products and materials. Assuming there are 100,000 material records in the database, 50,000 part records will remain after filtering.

[0068] The second step involves calling the database's built-in CLR custom similarity function to execute a text similarity algorithm based on edit distance. This includes calculating the Levenshtein edit distance (the minimum number of operations required to convert one string into another through insertion, deletion, and replacement) between two strings for each of the user-input attributes: name, category, and material. The edit distance is then converted into a similarity percentage. The similarity percentages of each attribute are then weighted and summed according to preset weights. For the "parts-fasteners" category, the weights are configured as follows: name weight 0.5, category weight 0.3, and material weight 0.2. Finally, the system calculates a comprehensive similarity score for each of the 50,000 parts data entries.

[0069] The third step is to sort the records in descending order of similarity score and select the top 200 records as the candidate material list. For example: the material ranked 1st is M8×20 hex bolt, made of 304 stainless steel, with a comprehensive score of 92.5; the material ranked 2nd is M8 external hex bolt, made of SUS304, with a comprehensive score of 88.3; and the material ranked 3rd is hex head bolt, made of stainless steel, with a comprehensive score of 85.7.

[0070] As another implementation method, instead of using CLR custom functions, the key attributes of the materials (such as classification, material, key performance parameters) can be tagged and stored in a dedicated index table. In the coarse sorting stage, the candidate set can be quickly filtered by performing efficient intersection operations on the tag index table (for example, materials that match multiple tags such as "fasteners" and "304 stainless steel" are given priority). This can also achieve the goal of quickly reducing the data range. Then, the candidate set is input into the AI ​​assistant for fine sorting.

[0071] Through the database coarse sorting stage, the high-performance computing capabilities of the database are utilized to quickly complete the screening and preliminary sorting of 50,000 data entries.

[0072] During the AI ​​assistant's fine-tuning stage, the system sends the list of 200 candidate materials after the initial rough sorting to the AI ​​assistant service for further refinement, including: The first step is for the AI ​​assistant to perform more precise similarity calculations on various attributes of each candidate material: For text-based attributes, the `token_set_ratio` method of the FuzzyWuzzy library is used to calculate text similarity. For example, if the input is "hexagonal bolt" vs "M8×20 hexagonal bolt", the string is first segmented into tokens, and then the set similarity is calculated. This method can handle word order changes and missing words, and the final result is 95 points (out of 100). Alternatively, if the input is "hexagonal bolt" vs "external hexagonal bolt M8", the final result is 90 points. For categorical attributes, such as inputting "fasteners" vs "fasteners", an exact match is performed, and the similarity is determined to be 1.0. For material-related attributes, the system uses a material equivalence knowledge base for matching. For example, this knowledge base contains the following mapping relationships:

[0073] If the input is "304 stainless steel" vs "304 stainless steel", the knowledge base query will show that they are completely identical, with a similarity of 1.0; if the input is "304 stainless steel" vs "SUS304", the knowledge base query will show that they are equivalent, with a similarity of 1.0; if the input is "304 stainless steel" vs "stainless steel", the knowledge base query will show that they belong to the same category of "austenitic stainless steel", but the description of "stainless steel" is not specific, with a similarity of 0.7; if the input is "304 stainless steel" vs "316 stainless steel", the knowledge base query will show that they belong to the same category of "austenitic stainless steel", but the grades are different, with a similarity of 0.7; if the input is "304 stainless steel" vs "Q235 steel", the knowledge base query will show that they are different material categories, and the similarity is 0.2 when calculated by reverting to the text editing distance.

[0074] Based on the similarity of the different attributes mentioned above, the system queries the weight configuration table and performs a weighted sum according to the dynamic weight configuration table. For example, for the "Fasteners" category of "Parts", the weight configuration is: name weight is 0.5, category weight is 0.3, and material weight is 0.2. Then, for candidate material BZ00123 (M8×20 hex bolt), its comprehensive similarity = 0.95×0.5 + 1.0×0.3 + 1.0×0.2 = 0.975 (97.5%); for candidate material BZ00125 (external hex bolt M8, SUS304), its comprehensive similarity = 0.90×0.5 + 1.0×0.3 + 1.0×0.2 = 0.95 (95%).

[0075] After the above calculations are completed, the system checks if there are any materials with very close similarity (difference < 0.05). If so, they are sorted by priority. For example, if material A is published, used 50 times, and created in 2024, and material B is under design, used 10 times, and created in 2025, then material A will rank higher (because its status has a higher priority).

[0076] Finally, after fine-tuning and state priority adjustment, the final recommendation list is generated as follows:

[0077] The fine-ranking stage takes about 0.3-0.5 seconds, and the entire two-stage recommendation process takes about 1-1.5 seconds, which meets the requirements for real-time recommendation.

[0078] Furthermore, the system obtains the refined recommendation list. Assuming the user interface displays 10 recommendations per page, the system prepares to return the first page of data. Before returning, the system checks the permission settings for each of the 10 items, confirming whether the currently logged-in user's role and department have permission to view these items by querying the permission control table. If the third item is a confidential project part and the current user does not have permission to view it, that record will be removed from the returned results. The system will then either replenish the list with a permissioned record from subsequent rankings to make 10, or directly return 9 records. Finally, the user-visible list of items, filtered by permissions, is displayed on the front-end interface.

[0079] Based on the displayed results, users can view detailed information about the recommended "hex bolts" materials one by one, including specifications, drawing previews, etc., and perform the following operations based on the results: Scenario 1 (Direct Reference): The user finds that the top-ranked recommended item, "M8*20 hex bolt" (material code: BZ00123), is exactly the same as the item they are about to create. The user can select the material and click the "Direct Reference" button. At this point, the system will terminate the current creation process and apply the material code BZ00123 to the user's BOM design or other higher-level applications.

[0080] Scenario 2 (Reference Creation): The user finds no completely identical material, but the recommended "M825 Hex Bolt" (material code: BZ00125) is very similar, differing only in length. The user can create a new "M820 Hex Bolt" based on BZ00125. The user selects BZ00125, clicks the "Reference Creation" button, and the system pops up an options interface listing various documents associated with BZ00125, such as: "2D Drawing.dwg", "3D Model.prt", "Process Card", "Inspection Standard.pdf", etc. The user can choose the import method for each type of document. For example, the user can choose "Copy" for the 3D model (because the length needs to be modified), "Reference" for the inspection standard (because the inspection standard is universal), and choose not to import the process card (because the process needs to be rewritten). After clicking "Confirm", the system enters a new material creation interface pre-filled with most attributes (inherited from BZ00125), and has already copied or associated the relevant documents to this new material to be created according to the user's selection. The user only needs to modify a few different attributes such as length to quickly complete the creation.

[0081] Scenario 3 (Continue to create): If the user finds that there are no completely identical materials and the recommended similar materials are also very dissimilar, the user can leave the material creation process unchecked and proceed directly to create a new material.

[0082] In summary, the material creation method proposed in this embodiment constructs a complete business loop of user input, two-stage intelligent sorting, permission filtering, and diversified reuse options. It accurately and efficiently recommends existing similar materials to users and provides powerful data reuse capabilities when users need to create new materials, thereby avoiding repeated creation and improving the design efficiency of material creation.

[0083] Example 2 This embodiment provides a material creation system based on intelligent recommendation of similar materials, including: The data acquisition module is used to respond to the user's material creation command and obtain multi-type attribute information of the material to be created; The intelligent recommendation module for similar materials is used to perform a two-stage similarity calculation and sorting of existing materials in the database based on the acquired information, namely, coarse sorting and AI assistant fine sorting, to filter out the recommended results of similar materials. The coarse sorting is: filtering candidate sets from the database through single attribute filtering and similarity calculation based on an improved edit distance algorithm; the AI ​​assistant fine sorting is: performing differential similarity calculations for different types of attributes on each existing material in the candidate set, obtaining a comprehensive similarity through dynamic weight fusion, and then performing a second sorting based on the status priority of existing materials. The display module is used to filter the recommendation results based on user permissions and provide feedback. The material creation module is used to respond to the user's selected operation command, and execute the process of direct reference, reference creation or continued creation based on the selected similar materials to complete the material creation.

[0084] Example 3 This embodiment provides an electronic device, including: a memory for storing executable instructions; and a processor for executing the executable instructions stored in the memory to implement the method provided in this embodiment.

[0085] Example 4 This embodiment also provides a computer-readable storage medium storing executable instructions, which, when executed by a processor, will cause the processor to execute the method described above in this embodiment.

[0086] Example 5 This embodiment provides a computer program product including executable instructions, which are computer instructions; the executable instructions are stored in a computer-readable storage medium. When the processor of an electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the electronic device performs the method described in this embodiment.

[0087] The steps and methods involved in Embodiments 2 to 5 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0088] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0089] The above description is only a preferred embodiment of the present invention. Although the specific implementation of the present invention has been described in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that, based on the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative effort are still within the scope of protection of the present invention.

Claims

1. A material creation method based on intelligent recommendation of similar materials, characterized in that, include: Respond to the user's material creation command and obtain multi-type attribute information of the material to be created; Based on the information obtained, a two-stage similarity calculation and sorting process is performed on the existing materials in the database: coarse sorting and fine sorting by AI assistant, to filter out recommended results for similar materials. The coarse sorting process involves filtering candidate materials from the database through single attribute filtering and similarity calculation based on an improved edit distance algorithm. The fine sorting process involves performing differential similarity calculations on each existing material in the candidate set for different types of attributes, obtaining a comprehensive similarity through dynamic weight fusion, and then performing a second sorting based on the priority of the existing material status. The recommended results are filtered based on user permissions and then fed back. Responding to the user's selected operation command, the system executes the process of direct reference, reference creation, or continued creation based on the selected similar materials to complete the material creation.

2. The material creation method based on intelligent recommendation of similar materials as described in claim 1, characterized in that, The multi-type attributes include: Text-based attributes include name, material description, and function description; Numerical attributes include weight, length, width, height, dimensions, and specifications. Category-based attributes include type, material group, and lifecycle status; Collective attributes include tag sets, applicable scenario sets, and process feature sets; Material attributes include material grade, material category, and material standard number.

3. The material creation method based on intelligent recommendation of similar materials as described in claim 1, characterized in that, The process of coarse-sorting the database is as follows: Based on the type and attribute information of the material to be created input by the user, the existing materials in the database are initially filtered; An improved edit distance algorithm integrated with CLR is used to calculate the multi-attribute similarity between existing materials obtained from the initial filtering and materials to be created. The calculation results are converted into similarity percentages and then weighted and summed. Specifically, for the name, material description, functional description, type, and material attributes, the edit distance algorithm is used to calculate the minimum number of edit operations between two strings to obtain the edit distance. Then, according to the calculation formula: similarity = 100 × (1 - edit distance / max(string 1 length, string 2 length)), the edit distance is converted into a similarity percentage. The top N existing materials are selected as the material candidate set based on their calculated scores in descending order.

4. The material creation method based on intelligent recommendation of similar materials as described in claim 1, characterized in that, The AI ​​assistant's fine-tuning process is as follows: For each existing material in the candidate set, calculate the differential similarity between it and the material to be created for different types of attributes; among them, textual attributes are calculated using a fuzzy matching algorithm, numerical attributes are calculated based on numerical differences, categorical attributes are calculated based on data value matching, set attributes are calculated using the Jaccard similarity coefficient, and material attributes are calculated based on the material equivalence knowledge base. Based on the preset weight configuration table, the similarity of different types of attributes is dynamically weighted and fused to obtain the comprehensive similarity. The existing materials in the candidate set are sorted based on the comprehensive similarity. When the interpolation of the comprehensive similarity is less than a preset threshold, the existing materials are sorted in conjunction with their state priority to obtain the final sorting result. The top M existing materials are then selected as the recommended similar materials.

5. The material creation method based on intelligent recommendation of similar materials as described in claim 4, characterized in that, Dynamic weight configuration includes: Configure attribute weight coefficients according to material type, business characteristics, and project stage; When an attribute is missing, the weights of the acquired attributes of the material to be created are automatically normalized. The priority ranking rules for material status include: Lifecycle status priority: Released > Under Design > Under Review > Obsolete; Frequency priority under the same conditions: high frequency use > low frequency use > no use; Creation time priority under the same conditions: Recently created > Early created.

6. The material creation method based on intelligent recommendation of similar materials as described in claim 1, characterized in that, The recommendation results are filtered based on user permissions and feedback is provided, including: Obtain the current user's role, department, and project group permissions, and simultaneously perform batch verification of the access permissions for each existing material in the recommendation results; After matching user and material permissions, materials without permissions are removed from the recommended results, and materials with permissions are added from subsequent rankings. The system then returns and displays similar materials that the current user has the right to access.

7. A material creation system based on intelligent recommendation of similar materials, characterized in that, include: The data acquisition module is used to respond to the user's material creation command and obtain multi-type attribute information of the material to be created; The intelligent recommendation module for similar materials is used to perform a two-stage similarity calculation and sorting of existing materials in the database based on the acquired information, namely, coarse sorting and AI assistant fine sorting, to filter out the recommended results of similar materials. The coarse sorting is: filtering candidate sets from the database through single attribute filtering and similarity calculation based on an improved edit distance algorithm; the AI ​​assistant fine sorting is: performing differential similarity calculations for different types of attributes on each existing material in the candidate set, obtaining a comprehensive similarity through dynamic weight fusion, and then performing a second sorting based on the status priority of existing materials. The display module is used to filter the recommendation results based on user permissions and provide feedback. The material creation module is used to respond to the user's selected operation command, and execute the process of direct reference, reference creation or continued creation based on the selected similar materials to complete the material creation.

8. An electronic device, characterized in that, include: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the material creation method based on intelligent recommendation of similar materials as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The device stores executable instructions that, when executed by a processor, implement the material creation method based on intelligent recommendation of similar materials as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes executable instructions stored in a computer-readable storage medium; When the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, it implements the material creation method based on intelligent recommendation of similar materials as described in any one of claims 1-6.

Images