FMEA data construction method and device based on cognitive digital twinning, medium and product

Abstract

The invention discloses an FMEA data construction method and device based on cognitive digital twinning, a medium and a product, and belongs to the field of data processing. Target information is extracted through intelligent analysis and converted into structured knowledge nodes; performing logic completeness analysis on the nodes to identify missing association, and generating complementation content based on context by adopting a language model; and finally, performing semantic alignment, conflict detection and resolution on the improved node set and the existing FMEA database to realize dynamic updating of the database. According to the method, the external document risk information is automatically identified and positioned to the FMEA structure tree, so that the analysis efficiency is remarkably improved, and the manual workload is greatly reduced; through logic completeness analysis and an automatic complementation mechanism, the completeness, accuracy and logic closed loop of FMEA analysis are ensured; batch import of 8D documents and the like is supported, a continuous iteration updating mechanism is established, and FMEA full-life-cycle intelligent closed loop is achieved.

Description

Technical Field

[0001] This application relates to the field of data processing, and in particular to methods, devices, media and products for constructing FMEA (Failure Mode and Effects Analysis) data based on cognitive digital twins. Background Technology

[0002] As industrial production becomes increasingly complex, Failure Mode and Effects Analysis (FMEA), as a systematic risk assessment and prevention method, plays a crucial role in product design, manufacturing, and quality control. FMEA effectively reduces product and process risks by identifying potential failure modes, assessing their impact, and developing corresponding control measures. However, traditional FMEA data construction methods face numerous challenges, particularly in data acquisition, processing, and knowledge management.

[0003] Current FMEA (Failure Mode and Effects Analysis) data construction methods primarily rely on manual data entry and expert experience, resulting in low efficiency and difficulty in guaranteeing data quality. Although technologies such as knowledge graphs have attempted to integrate multimodal data, the following key shortcomings remain in practical applications:

[0004] 1. Insufficient multimodal document parsing capabilities: It is difficult to effectively process complex unstructured documents containing text, tables, flowcharts, etc., resulting in incomplete or incorrect information extraction.

[0005] 2. Incomplete logical relationships between knowledge nodes: The lack of a complete analysis of the logical relationships between nodes makes it impossible to identify and supplement missing connections, affecting the comprehensiveness of the analysis.

[0006] 3. Poor compatibility with standard versions: It cannot flexibly adapt to the FMEA standard versions adopted by different industries and enterprises, making it difficult to map information into the target analysis framework.

[0007] 4. Insufficient utilization of historical case knowledge: Failure to effectively explore and utilize the experience of linking failures and rectification in historical cases, and lack of ability to learn from past experience.

[0008] 5. Weak semantic alignment and conflict resolution capabilities: When integrating new knowledge nodes, the lack of effective semantic alignment and conflict detection mechanisms can easily lead to data duplication or contradictions.

[0009] In summary, existing methods have significant shortcomings in document parsing, logical completeness, standard adaptation, experience reuse, and semantic integration, which limit the efficiency and quality of FMEA analysis. Summary of the Invention

[0010] To address the significant shortcomings of existing FMEA data construction methods in areas such as document parsing, logical completeness, standard adaptation, experience reuse, and semantic integration, this paper presents a cognitive digital twin-based FMEA data construction method, device, medium, and product. This method aims to efficiently process multimodal unstructured documents, possess logical completeness analysis capabilities, support adaptation to different FMEA standard versions, fully utilize historical case knowledge, and achieve semantic alignment and conflict resolution.

[0011] This application provides a method for constructing FMEA data based on cognitive digital twins, including:

[0012] Obtain the unstructured document in multimodal format to be processed;

[0013] The document to be processed is parsed, the target information is extracted, and the target information is transformed into structured knowledge nodes;

[0014] Logical completeness analysis is performed on the structured knowledge nodes to identify missing related nodes;

[0015] Using a first language model, based on the associated nodes, the structured knowledge nodes, and their context, the missing associated node content is generated;

[0016] The completed structured knowledge node set is semantically aligned, conflict detected, and resolved with the existing structure tree in the FMEA database to update the FMEA database.

[0017] Optionally, parsing the document to be processed, extracting target information, and converting the target information into structured knowledge nodes includes:

[0018] Identify and parse the text, and / or tables, and / or block diagrams, and / or flowcharts in the document to be processed;

[0019] Based on the analysis results, target information corresponding to FMEA elements is identified and extracted. The FMEA elements include at least one of product components, functions, failure modes, failure causes, failure consequences, and control measures.

[0020] The topological relationships in the tables, and / or the block diagrams, and / or the flowcharts are transformed into the association relationships between the structured knowledge nodes.

[0021] Optionally, before performing logical completeness analysis on the structured knowledge nodes and identifying missing related nodes, the method further includes:

[0022] Identify the version type of the document to be processed;

[0023] Based on the version type, the preset version adaptation rules are invoked to map the target information to the analysis framework required by the target FMEA standard version, so as to generate structured knowledge nodes that are consistent with the target version logic.

[0024] Optionally, before performing logical completeness analysis on the structured knowledge nodes and identifying missing related nodes, the method further includes:

[0025] Retrieve historical case documents associated with the document to be processed;

[0026] Based on semantic matching, the correlation between fault items and rectification items is extracted from the historical case documents, and the rectification items include the root cause and corrective measures.

[0027] The aforementioned relationships are added to the current structured knowledge nodes as prior knowledge.

[0028] Optionally, the step of extracting the association between fault items and rectification items from the historical case documents based on semantic matching includes:

[0029] Semantic parsing is performed on the historical case documents to identify the entities of fault items and rectification items;

[0030] Calculate a semantic matching score based on the fault item entity and the rectification item entity;

[0031] Based on the semantic matching scores, the association between fault items and rectification items is established, and a relationship graph is constructed.

[0032] Extract the root cause sub-items and corrective action sub-items from the relationship graph to supplement the relationship.

[0033] Optionally, the step of semantically aligning, detecting, and resolving conflicts between the completed structured knowledge node set and the existing structure tree in the FMEA database to update the FMEA database includes:

[0034] A second language model is used to calculate the semantic similarity between the structured knowledge nodes and the existing structure tree nodes. The second language model is a similarity calculation model based on semantic vectors.

[0035] Based on the semantic similarity, conflicting nodes that are semantically similar but inconsistent in expression are identified, and conflict prompt information is generated;

[0036] In response to the conflict warning information, a third language model is used to analyze the context of the conflict node and generate standardized node terms; the third language model is a generative large language model.

[0037] The standardized node terms are recommended to users for confirmation or automatically stored in the FMEA database.

[0038] Optional, also includes:

[0039] Before or after the structured knowledge node set is updated to the FMEA database, a fourth language model is used to predict the scores of the newly added or updated nodes; wherein, the fourth language model is a generative large language model.

[0040] The score prediction includes:

[0041] The fourth language model is provided with a description of the node to be predicted, historical scoring data in the FMEA database, and preset scoring criteria.

[0042] Obtain the predicted evaluation values ​​of severity, frequency, and detectivity for the node output by the fourth language model, as well as the confidence intervals corresponding to each predicted evaluation value.

[0043] This application also provides an electronic device comprising: one or more processors; and a memory storing computer program instructions, which, when executed, cause the processor to perform the steps of the method described above.

[0044] This application also provides a computer-readable medium having computer program instructions stored thereon, which can be executed by a processor to implement the method described above.

[0045] This application also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method described above.

[0046] The beneficial effects of the above technical solution are as follows:

[0047] This application's FMEA data construction method based on cognitive digital twins can acquire multimodal unstructured documents, extract target information through intelligent parsing and transform it into structured knowledge nodes; then, it performs logical completeness analysis on these nodes to identify missing connections, and uses a language model to generate complete content based on context; finally, it performs semantic alignment, conflict detection and resolution with the existing FMEA database to achieve dynamic database updates. This application significantly improves analysis efficiency and greatly reduces manual workload by automatically identifying risk information from external documents and locating it in the FMEA structure tree; through logical completeness analysis and automatic completion mechanisms, it ensures the completeness, accuracy, and logical closure of FMEA analysis; it supports batch import of documents such as 8D and establishes a continuous iterative update mechanism, strengthening reliability management "problem-driven improvement" and realizing an intelligent closed loop throughout the FMEA lifecycle; with the help of semantic alignment and conflict resolution technologies, it ensures the dynamic updates and practical adaptability of the system. Compared with existing technologies, this application achieves more efficient, accurate, and comprehensive failure mode and impact analysis, providing a stronger guarantee for product quality and reliability. Attached Figure Description

[0048] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0049] Figure 1 This is a flowchart of one embodiment of the FMEA data construction method based on cognitive digital twins described in this application;

[0050] Figure 2 This is a flowchart illustrating a method for converting target information into structured knowledge nodes according to an embodiment of this application.

[0051] Figure 3 This is a flowchart illustrating an embodiment of the method for identifying missing related nodes by performing logical completeness analysis on structured knowledge nodes according to this application.

[0052] Figure 4 A flowchart illustrating a method for generating complete content for missing related nodes in this application;

[0053] Figure 5 This is an exemplary structural diagram of the electronic device of this application. Detailed Implementation

[0054] The advantages of this application are further illustrated below with reference to the accompanying drawings and specific embodiments.

[0055] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0056] The terminology used in this disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. The singular forms “a,” “the,” and “the” as used in this disclosure and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0057] It should be understood that although the terms first, second, third, etc., may be used in this disclosure to describe various information, such information should not be limited to these terms. These terms are used only to distinguish information of the same type from one another. For example, without departing from the scope of this disclosure, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0058] In the description of this application, it should be understood that the numerical labels before the steps do not indicate the order of the steps, but are only used to facilitate the description of this application and to distinguish each step, and therefore should not be construed as a limitation of this application.

[0059] This application addresses the significant shortcomings of existing FMEA data construction methods in areas such as document parsing, logical completeness, standard adaptation, experience reuse, and semantic integration. It proposes a cognitive digital twin-based FMEA data construction method. (See reference...) Figure 1 The FMEA data construction method based on cognitive digital twins provided in this embodiment mainly includes the following steps:

[0060] S1. Obtain the document to be processed in a multimodal unstructured format;

[0061] In step S1, the documents to be processed may include various formats such as PDF files, Word documents, scanned images, and technical manuals. These documents may contain product design information, fault analysis reports, maintenance records, historical FMEA tables, etc., providing raw data sources for subsequent FMEA data construction.

[0062] S2. Parse the document to be processed, extract the target information, and transform the target information into structured knowledge nodes;

[0063] Specifically, see Figure 2 Step S2 shown may include the following steps:

[0064] S21. Identify and parse the text, and / or tables, and / or block diagrams, and / or flowcharts in the document to be processed;

[0065] In this step, optical character recognition technology can be used to process scanned documents and extract text content; for electronic documents, text and table data can be extracted directly; for images and charts, image recognition technology can be used to identify key elements and connections.

[0066] S22. Based on the analysis results, identify and extract the target information corresponding to the FMEA elements;

[0067] The FMEA elements may include at least one of the following: product components, functions, failure modes, failure causes, failure consequences, and control measures. For example, product component information can be extracted from product manuals, failure modes and failure causes can be extracted from maintenance records, and failure consequences can be extracted from quality reports.

[0068] S23. Transform the topological relationships in the table, and / or the block diagram, and / or the flowchart into the association relationships between the structured knowledge nodes.

[0069] In this step, the hierarchical relationships in the product structure diagram can be converted into subordinate relationships between components, and the connecting lines in the flowchart can be converted into causal relationships between functions. This forms the initial set of structured knowledge nodes, where each node represents an FMEA analysis element, and the connections between nodes represent their logical relationships.

[0070] In step S2, the text, tables, and charts in the multimodal document are automatically parsed, and key FMEA elements such as product components, functions, and failure modes are extracted and transformed into structured knowledge nodes. Simultaneously, the topological relationships in the document (such as box hierarchy and flowchart connections) are transformed into logical associations between nodes. This process achieves a breakthrough in automatically constructing knowledge networks from complex unstructured documents, significantly improving the efficiency and coverage of information extraction, and effectively avoiding information omissions and logical breaks caused by traditional manual data entry.

[0071] S3. Perform logical completeness analysis on the structured knowledge nodes to identify missing related nodes;

[0072] In this embodiment, based on the logical framework and completeness requirements of FMEA, the structured knowledge nodes are checked for missing or incomplete information. For example, it checks whether each failure mode has a corresponding failure cause, failure consequence, and control measures; whether each component has completed functional and failure analysis; and whether the risk assessment is complete, etc., to mark missing nodes and relationships, providing a basis for subsequent completion work.

[0073] Further, see Figure 3 Step S3 shown may include the following steps:

[0074] S31. Construct a knowledge graph for the structured knowledge nodes;

[0075] In step S31, a knowledge graph for the FMEA domain is constructed based on the existing structured knowledge nodes and their relationships. This graph uses FMEA elements such as product components, functions, and failure modes as nodes, and the logical relationships between them as edges, forming a complete analysis network.

[0076] S32. Determine the completeness of the paths by traversing the nodes in the knowledge graph;

[0077] Step S32 involves traversing the constructed knowledge graph according to the standard logical path of FMEA analysis. Specifically, this involves checking the completeness of critical paths such as from product components to functions, from functions to failure modes, from failure modes to failure causes, and from failure causes to control measures.

[0078] S33. If the path is interrupted, determine the missing associated node corresponding to the interruption location;

[0079] When a path interruption is detected, the interruption location is precisely pinpointed, and the type of missing associated node is identified. For example, if a failure mode lacks a corresponding failure cause, the node representing that failure cause is marked as a missing node.

[0080] S34. Collect contextual information for the missing associated nodes;

[0081] Step S34 collects the contextual information of each missing associated node in the knowledge graph, including adjacent nodes, associated attributes, and other relevant information in the path.

[0082] S35. Generate a candidate set of missing associated nodes based on the context information;

[0083] Step S35 generates a candidate set of missing related nodes based on the collected context information. This candidate set contains multiple possible supplementary nodes to ensure coverage of various possible analysis directions.

[0084] S36. Verify the validity of nodes in the candidate set using logical reasoning rules;

[0085] Step S36 uses predefined logical reasoning rules to verify the validity of nodes in the candidate set. These rules include judgment criteria across multiple dimensions, such as the rationality of causal relationships, technical feasibility, and compliance with industry standards.

[0086] S37. Based on the verification results, filter the candidate set to obtain the set of missing associated nodes.

[0087] Step S37 filters the candidate set based on the verification results, removes nodes that do not conform to the logical reasoning rules, and obtains the final set of missing related nodes, providing an accurate target for subsequent automatic completion.

[0088] In this embodiment, by constructing a knowledge graph from structured knowledge nodes and performing path traversal analysis, missing related nodes in the failure mode analysis chain can be automatically identified. Subsequently, candidate nodes are generated based on contextual information, and their validity is verified using logical reasoning rules. This automated analysis process achieves a systematic detection of the logical completeness of FMEA, significantly improving the completeness and accuracy of the analysis. It effectively avoids logical breaks and content omissions that are prone to occur in traditional manual analysis, ensuring a tight closed loop in the failure analysis chain and laying a reliable foundation for subsequent automatic completion.

[0089] S4. Using a first language model, based on the associated nodes, the structured knowledge nodes, and their context, generate the complete content for the missing associated nodes;

[0090] In step S4, a pre-trained first language model is invoked. This model has been trained on a large amount of FMEA data and possesses the ability to understand FMEA logic and generate relevant content. Existing structured knowledge nodes and their context are used as input, allowing the language model to generate reasonable content for missing nodes. For example, for a failure mode lacking a cause, the model will generate possible causes based on the description of the failure mode and similar cases; for a failure cause lacking control measures, the model will generate appropriate control measure recommendations based on the nature of the cause and industry best practices.

[0091] Further, see Figure 4 Step S4 may also include the following steps:

[0092] S41. For each node in the set of missing associated nodes, extract the adjacent nodes of the structured knowledge node;

[0093] In step S41, for each target node in the set of missing associated nodes, its direct adjacent nodes in the knowledge graph are automatically extracted. For example, for the missing "failure cause" node, its associated "failure mode" node and the existing "failure consequence" node are extracted as key adjacency information.

[0094] S42. Determine the context sequence based on the adjacent nodes; input the context sequence using the first language model to generate candidate completion content; calculate the semantic consistency score between the candidate completion content and the structured knowledge node;

[0095] In step S42, a structured context sequence is constructed based on the extracted neighboring nodes and input into a first language model pre-trained with FMEA domain data. This model generates multiple candidate completion content based on contextual semantics (e.g., generating 3-5 possible failure cause descriptions for a specific failure mode). Subsequently, a semantic consistency score between each candidate content and existing knowledge nodes is calculated using semantic vectors to ensure the consistency of the generated content with the overall knowledge structure.

[0096] S43. Filter and complete the content based on the semantic consistency score;

[0097] Step S43 allows setting a semantic consistency threshold (e.g., 0.75) to filter the generated completion content candidates. Candidates below the threshold are automatically eliminated, while semantically consistent and technically logically sound candidates are retained to proceed to the next stage.

[0098] S44. Fill the missing related nodes with the filtered and completed content; update the structured knowledge nodes according to the filling results to obtain the completed structured knowledge node set.

[0099] Step S44 formally fills the selected completed content into the corresponding missing node positions and updates the relationships in the knowledge graph. Simultaneously, the generation source and semantic consistency score of each completed node are recorded, forming a complete version tracking record. Finally, a logically complete structured knowledge node set is output, providing standardized data input for subsequent database integration.

[0100] In this embodiment, a pre-trained domain language model (first language model) automatically generates complete content for missing related nodes based on existing knowledge nodes and their contextual relationships. This complete content undergoes semantic consistency scoring to ensure that the generated failure causes, control measures, and other information are highly consistent with the existing knowledge system. This intelligent completion mechanism significantly improves the efficiency and quality of FMEA analysis. It not only effectively solves the problem of broken analysis chains caused by missing information in traditional methods but also ensures the professionalism and practicality of the completed content through deep learning of industry best practices by the model, greatly reducing the subjectivity and workload of manual completion.

[0101] S5. The completed structured knowledge node set is semantically aligned with the existing structure tree in the FMEA database, and conflict detection and resolution are performed to update the FMEA database.

[0102] Furthermore, step S5 may include the following steps:

[0103] S51. Using a second language model, calculate the semantic similarity between the structured knowledge node and the existing structure tree node;

[0104] The second language model is a similarity calculation model based on semantic vectors, which can understand the semantic connotation of terms rather than relying solely on literal matching.

[0105] Specifically, step S51 may also include the following steps:

[0106] S511. Generate semantic vector representations for the structured knowledge nodes and for the target structure tree nodes;

[0107] In step S511, a second language model based on the Transformer architecture is used to convert the structured knowledge nodes to be integrated and the existing structure tree nodes in the FMEA database into high-dimensional semantic vector representations. This model has been pre-trained on a large scale with FMEA domain terminology and is able to deeply understand the semantic connotations of professional terms such as "contact oxidation" and "arc corrosion".

[0108] S512. Calculate the cosine similarity as the semantic similarity based on the semantic vector representation;

[0109] In step S512, based on the generated semantic vector representation, the cosine similarity algorithm is used to calculate the semantic similarity between nodes. Specifically, the cosine value of the angle between two semantic vectors is calculated to obtain a similarity score in the range of 0-1, where 0 represents completely unrelated and 1 represents completely identical semantics.

[0110] S513. For node pairs with semantic similarity higher than a preset threshold, mark them as potential conflict nodes;

[0111] In step S513, a dynamic similarity threshold (preferably 0.85) is set. When the semantic similarity of a node pair exceeds this threshold, the system automatically marks it as a potential conflict node. For example, although "poor circuit board contact" and "unstable PCB connection" are different in expression, they are highly similar in meaning and will be identified as a potential conflict pair.

[0112] S514. Collect representational discrepancy information based on the potential conflict nodes;

[0113] In step S514, for the marked potential conflict nodes, key difference features in their textual expressions are extracted, including but not limited to: differences in terminology usage, differences in sentence structure, differences in modifiers, and differences in technical parameters.

[0114] S515. Construct a conflict feature vector using the aforementioned representation difference information;

[0115] In step S515, the extracted expression difference information is quantified into a multi-dimensional conflict feature vector, including feature dimensions such as term dispersion, syntactic similarity, and semantic overlap, to form a standardized conflict description representation.

[0116] S516. Verify conflicting nodes that are semantically similar but inconsistent in expression based on the conflict feature vectors.

[0117] In step S516, a pre-trained conflict classifier is used for final verification based on the constructed conflict feature vector. This classifier, trained on historical conflict data, can accurately distinguish between "reasonable expression variants" and "substantive content conflicts," ensuring that only genuine semantic conflict nodes are subsequently resolved.

[0118] In step S51, a multi-level semantic analysis mechanism is used to achieve a breakthrough from shallow literal matching to deep semantic understanding, effectively solving the semantic conflict detection problem in FMEA knowledge integration and providing technical support for maintaining the consistency of the knowledge base.

[0119] S52. Based on the semantic similarity, identify conflicting nodes that are semantically similar but inconsistent in expression, and generate conflict prompt information;

[0120] For example, when a new node "bolt loose" is semantically similar to but has a different description than an existing node "fastener loose" in the database, the system will mark this conflict and generate a prompt.

[0121] S53. In response to the conflict warning information, a third language model is used to analyze the context of the conflict node and generate standardized node terms;

[0122] The third language model is a generative large language model that can understand the context and industry norms of terms and propose a unified standard expression.

[0123] Specifically, step S53 may include the following steps:

[0124] S531. Extract the context window for the conflicting node;

[0125] In step S531, for each marked conflict node, its associated nodes and relationships within a three-hop range in both the forward and backward directions can be automatically extracted to form a complete context window. For example, for the conflict node "connector loose", its upstream "vibration environment" and "number of insertions and removals" and its downstream "signal interruption" and "short circuit risk" are also extracted.

[0126] S532. Determine the local semantic environment of the conflicting node based on the context window;

[0127] In step S532, a local semantic environment containing three dimensions—node role, functional attributes, and failure mechanism—can be constructed based on the extracted context window. By analyzing the positional relationships and functional dependencies of nodes in the FMEA structure tree, the semantic positioning of the node in a specific technical scenario can be determined.

[0128] S533. Using the third language model as input to the local semantic environment, standardized term candidates are generated;

[0129] In step S533, the constructed local semantic environment is input into a third language model (generative large language model) trained on the FMEA industry corpus. This model generates 3-5 standardized term candidates that conform to industry norms based on an industry-standard terminology database and contextual semantics. For example, for "connector loose," candidate terms such as "connector poor contact" and "interface connection failure" may be generated.

[0130] S534. Calculate the matching degree between the standardized term candidates and existing terms in the FMEA database;

[0131] In step S534, the semantic matching degree between each standardized term candidate and existing terms in the FMEA database is calculated. An attention-based matching algorithm is used, which considers three indicators: literal similarity, semantic relevance, and frequency of use of terms, to generate a comprehensive matching score in the range of 0-1.

[0132] S535. Filter standardized terms based on the matching degree;

[0133] In step S535, a matching threshold (preferably 0.8) is set, and the candidate terms with the highest matching degree are selected as the preferred standard expressions. For multiple high-scoring candidate terms, the expression with the highest consistency with the current FMEA standard version terminology database is selected first.

[0134] S536. Replace the descriptions in the conflicting nodes with the filtered standardized terms;

[0135] In step S536, the selected standardized terms are automatically used to replace the non-standard expressions in the original conflict nodes, and all associated edge relationships and attribute descriptions are updated synchronously to ensure the integrity and consistency of the knowledge graph.

[0136] S537. Generate updated structured knowledge nodes based on the replacement results.

[0137] In step S537, based on the terminology replacement results, updated structured knowledge nodes are generated, and version change information and decision-making basis in the terminology standardization process are recorded to form a complete terminology tracing record.

[0138] In step S53, a context-aware terminology standardization mechanism effectively resolves the inconsistency in expression during FMEA knowledge integration. This maintains industry-standard terminology while ensuring semantic consistency of the knowledge graph, providing a technical foundation for the long-term maintenance and evolution of FMEA knowledge. Through a semantic vector-based similarity calculation model (third language model), potential conflict nodes between the completed content and the existing FMEA database are intelligently identified, and conflict feature vectors are constructed for precise verification, achieving deep semantic alignment and conflict resolution. This process ensures semantic consistency between new and old knowledge, effectively avoiding duplication and contradictions in the database, significantly improving the data quality and system reliability of the FMEA knowledge base, enabling the FMEA system to maintain dynamic updates and continuous optimization, and enhancing its practicality and long-term maintenance sustainability.

[0139] S54. Recommend the standardized node terms to the user for confirmation or automatically store them in the FMEA database.

[0140] Step S54 can be configured to directly submit the generated standard terms to the user for review, or automatically update them to the database if the confidence level is high.

[0141] In an embodiment, updating the FMEA database may include: generating database insertion records for the standardized node terms; determining a target position in the FMEA database based on the insertion records; performing a term replacement operation using the target position; performing a consistency check on the FMEA database after the replacement operation; confirming the update is complete based on the consistency check result; refreshing the target structure tree nodes using the updated FMEA database; and generating a complete structured knowledge representation based on the target structure tree nodes.

[0142] In this embodiment, the FMEA data construction method based on cognitive digital twins can acquire multimodal unstructured documents, extract target information through intelligent parsing and transform it into structured knowledge nodes; then, logical completeness analysis is performed on these nodes to identify missing connections, and language models are used to generate complete content based on context; finally, the improved node set is semantically aligned with the existing FMEA database, and conflict detection and resolution are performed to achieve dynamic updates of the database. This application significantly improves analysis efficiency and greatly reduces manual workload by automatically identifying risk information in external documents and locating it in the FMEA structure tree; through logical completeness analysis and automatic completion mechanisms, the integrity, accuracy, and logical closed loop of FMEA analysis are ensured; it supports batch import of documents such as 8D and establishes a continuous iterative update mechanism, strengthening reliability management "problem-driven improvement" and realizing an intelligent closed loop throughout the FMEA lifecycle; with the help of semantic alignment and conflict resolution technology, the dynamic updates and practical adaptability of the system are guaranteed; compared with the prior art, this application achieves more efficient, accurate, and comprehensive failure mode and impact analysis, providing a stronger guarantee for product quality and reliability.

[0143] In a preferred embodiment, the procedure may further include the following steps before performing step S3:

[0144] A1. Identify the version type of the document to be processed;

[0145] In this step, by analyzing features such as document format, keywords, and table header structure, the version type (i.e., FMEA standard version) of the document to be processed is identified, such as AIAG Version 4, AIAG-VDA, etc. Different versions of the FMEA standard differ in their analytical framework, scoring criteria, and terminology definitions; accurate identification of the version type is crucial for subsequent data processing.

[0146] A2. Based on the version type, invoke the preset version adaptation rules to map the target information to the analysis framework required by the target FMEA standard version, so as to generate structured knowledge nodes that are consistent with the target version logic.

[0147] In this step, based on the identified version type, the corresponding adaptation rules are retrieved from a pre-built rule base. These rules define the mapping relationships between different versions of the FMEA standard, such as the correspondence between functional analysis and failure analysis between AIAG Version 4 and AIAG-VDA versions. The target information extracted from the document is transformed and reorganized according to these rules to conform to the analytical framework and logical structure of the target version of the FMEA. For example, the "severity," "frequency," and "detectability" scores in the old version of the FMEA are converted into the corresponding scoring standards in the new version of the FMEA.

[0148] In this embodiment, by automatically identifying the version type of the document to be processed and mapping the extracted information to the analysis framework of the target version based on the preset version adaptation rules, automatic conversion and compatibility between different versions of FMEA standards are realized. This effectively solves the problem of inconsistent frameworks caused by the coexistence of multiple standard versions, ensures that the structured knowledge nodes are fully compatible with the logical requirements of the target version, and significantly improves the accuracy of FMEA data construction and system adaptability.

[0149] In a preferred embodiment, the procedure may further include the following steps before performing step S3:

[0150] B1. Obtain historical case documents associated with the document to be processed;

[0151] It should be noted that historical case documents include, but are not limited to: FMEA analysis reports of similar products, quality issue handling records, failure analysis reports, engineering change notices, 8D reports, after-sales data records, test logs, etc. Document association rules are established based on product model, component type, and functional characteristics to ensure the relevance of historical documents. These historical documents contain actual failure cases and their solutions from the past, serving as an important source of knowledge to supplement current FMEA analysis.

[0152] B2. Based on semantic matching, extract the correlation between fault items and rectification items from the historical case documents;

[0153] The rectification items include the root causes and corrective measures.

[0154] Further, step B2 may include the following steps:

[0155] B21. Perform semantic parsing on the aforementioned historical case documents to determine the entities of fault items and rectification items;

[0156] In step B21, natural language processing technology can be used to identify key entities such as fault description, fault cause, and corrective measures taken from historical documents.

[0157] Specifically, step B21 can use the named entity recognition model to identify fault item entities (such as "poor contact" and "overheating failure") and rectification item entities (including root cause entities such as "material aging" and corrective action entities such as "add heat dissipation device") in the document, and classify and label them according to the preset FMEA entity types.

[0158] B22. Calculate the semantic matching score based on the fault item entity and the rectification item entity;

[0159] In step B22, the semantic similarity between entities is calculated to determine the possible causal relationship between them.

[0160] Furthermore, step B22 can generate a causal association score by calculating the similarity between the fault item entity and the rectification item entity in the semantic space. Specifically, an attention-based semantic matching model is used to analyze the potential causal relationships between entities.

[0161] B23. Based on the semantic matching score, establish the association between fault items and rectification items, and construct a relationship graph;

[0162] In step B23, connections are established between entity pairs with matching scores higher than the threshold to form a fault-cause-measure association network.

[0163] In step B23, targeted connections are established for entity pairs with semantic matching scores higher than a preset threshold (preferably 0.75), constructing a three-layer relationship graph of fault-cause-measure. This graph clearly shows the association paths between fault phenomena, root causes, and corrective measures in historical cases.

[0164] B24. Extract the root cause sub-items and corrective action sub-items from the relationship graph to supplement the relationship.

[0165] In step B24, key root cause sub-items and corrective action sub-items are extracted from the constructed relationship graph, especially those solutions that have been repeatedly found and proven effective in multiple historical cases, forming a proven knowledge unit (as reference knowledge for the current FMEA analysis).

[0166] B3. Add the aforementioned relationship as prior knowledge to the current structured knowledge node.

[0167] In step B3, for the failure modes identified in the current FMEA analysis, similar failure items in the historical relationship graph are automatically matched, and the corresponding root causes and corrective measures are attached to the corresponding nodes as recommended knowledge; a historical knowledge credibility scoring mechanism is established, and prior knowledge is weighted based on factors such as the number of historical cases, the effectiveness of solutions, and timeliness; the source traceability information of historical knowledge is recorded in the knowledge nodes to ensure the interpretability of the analysis process.

[0168] This embodiment effectively improves the accuracy and practicality of FMEA analysis by mining and utilizing experiential knowledge from historical cases, avoids the recurrence of the same problems, and realizes the continuous accumulation and intelligent reuse of enterprise reliability knowledge.

[0169] In this embodiment, the fault-cause-response relationships extracted from historical cases are integrated with the currently constructed structured knowledge nodes. For failure modes identified in the current FMEA, if similar failure items exist in historical cases, the corresponding root causes and corrective actions are added to the current node as prior knowledge, enriching the content and depth of the FMEA analysis. This knowledge transfer based on historical experience can avoid repeating mistakes and improve the quality and efficiency of FMEA analysis.

[0170] This embodiment automatically extracts the correlation between fault items and rectification items from historical case documents using semantic analysis technology and integrates it as prior knowledge into the current FMEA analysis. This innovation enables the intelligent reuse of historical experience knowledge, accurately supplementing the current analysis node with root causes and corrective measures that have been verified in practice. This effectively avoids the recurrence of the same faults, significantly improves the depth and accuracy of FMEA analysis, and greatly reduces the workload of manual analysis.

[0171] In a preferred embodiment, the FMEA data construction method based on cognitive digital twins may further include the following steps:

[0172] S6. Before or after the structured knowledge node set is updated to the FMEA database, a fourth language model is used to predict the scores of the newly added or updated nodes.

[0173] The fourth language model is a generative large language model trained with a specific risk assessment. The score prediction includes: providing the fourth language model with a description of the node to be predicted, historical score data in the FMEA database, and preset score criteria; obtaining the predicted evaluation values ​​of severity, frequency, and detectivity of the node output by the fourth language model, as well as the confidence intervals corresponding to each predicted evaluation value.

[0174] In this embodiment, newly added or updated node information, along with relevant historical data and scoring criteria, can be input into the model. The fourth language model not only provides specific predicted scores but also confidence intervals, reflecting the reliability of the predictions. These prediction results can serve as a reference for FMEA analysts, helping them make more accurate risk assessments. This embodiment, combined with the scoring prediction function, further enhances the scientific rigor and forward-looking nature of the analysis.

[0175] This embodiment automates the conversion and construction of multimodal unstructured documents into structured FMEA data, significantly improving the efficiency and quality of FMEA analysis. Utilizing language models and knowledge graph technology, this method can automatically complete missing information, eliminate terminology conflicts, and predict risk scores, providing strong support for product design and quality improvement.

[0176] Furthermore, some embodiments of this application also provide an electronic device. The electronic device can be various forms of digital computer, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, etc. The electronic device can also be various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices.

[0177] The electronic device includes: one or more processors; and a memory storing computer program instructions that, when executed, cause the processor to perform the steps of the methods provided in any one or more of the above embodiments. Figure 5 An exemplary structural diagram of the electronic device is disclosed. For example... Figure 5 As shown, the electronic device includes one or more processors 1101, a memory 1102, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components are interconnected via different buses and can be mounted on a common motherboard or otherwise as required. The processors can process instructions executed within the electronic device, including instructions stored in or on memory to display graphical information of a GUI on an external input / output device (such as a display device coupled to the interface). In some other embodiments, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple electronic devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). The components, their connections and relationships, and their functions shown herein are merely examples and are not intended to limit the implementation of the present application described and / or claimed herein.

[0178] The electronic device may further include an input device 1103 and an output device 1104. The processor 1101, memory 1102, input device 1103, and output device 1104 may be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0179] Input device 1103 can receive input numerical or character information, and generate key signal inputs related to user settings and function control of the electronic device, such as a touch screen, keypad, mouse, trackpad, touchpad, joystick, one or more mouse buttons, trackball, joystick, etc. Output device 1104 may include a display device, auxiliary lighting device (e.g., LED), and haptic feedback device (e.g., vibration motor). The display device may include, but is not limited to, a liquid crystal display (LCD), a light-emitting diode (LED) display, and a plasma display. In some embodiments, the display device may be a touch screen.

[0180] To provide interaction with the user, the electronic device can be a computer. The computer has: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0181] In this embodiment, a computer-readable medium stores a computer program / instructions that, when executed by a processor, implement the steps of the methods provided in any one or more of the above embodiments. This computer-readable medium may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into that device. The aforementioned computer-readable medium carries one or more computer-readable instructions.

[0182] The memory 1102 can serve as a non-transitory computer-readable storage medium, used to store non-transitory software programs, non-transitory computer-executable programs, and modules. The processor 1101 executes various functional applications and data processing of the server by running the non-transitory software programs, instructions, and modules stored in the memory 1102, thereby implementing the program instructions / modules corresponding to the methods provided in any one or more of the embodiments described above in this application.

[0183] The memory 1102 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the electronic device. Furthermore, the memory 1102 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, the memory 1102 may optionally include memory remotely located relative to the processor 1101, and these remote memories can be connected to the electronic device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0184] It should be noted that the computer-readable medium described in this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0185] Computer-readable media include permanent and non-permanent, removable and non-removable media, which can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, read-only optical disc (CD-ROM), digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0186] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0187] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. For example, it can be implemented using an application-specific integrated circuit (ASIC), a general-purpose computer, or any other similar hardware device. In some embodiments, the software program of this application can be executed by a processor to implement the steps or functions described above. Similarly, the software program of this application (including related data structures) can be stored in a computer-readable recording medium, such as RAM memory, magnetic or optical drives, floppy disks, or similar devices. Furthermore, some steps or functions of this application can be implemented in hardware, for example, as circuitry that works with a processor to perform the various steps or functions.

[0188] The computer program product provided in this application includes one or more computer programs / instructions. When executed by a processor, these computer programs / instructions generate, in whole or in part, the processes or functions described in this application. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0189] The flowcharts or block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of devices, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-specific system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0190] The scope of this application is defined by the appended claims rather than the foregoing description, and is therefore intended to encompass all variations falling within the meaning and scope of equivalents of the claims. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a device claim may also be implemented by a single unit or device in software or hardware. Terms such as "first," "second," etc., are used only for distinguishing descriptions and do not indicate any particular order, nor should they be construed as indicating or implying relative importance.

[0191] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily made by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims, and the above embodiments should be regarded as exemplary and non-limiting.

Claims

1. A method for constructing FMEA data based on cognitive digital twins, characterized in that, include: Obtain the unstructured document in multimodal format to be processed; The document to be processed is parsed, the target information is extracted, and the target information is transformed into structured knowledge nodes; Logical completeness analysis is performed on the structured knowledge nodes to identify missing related nodes; Using a first language model, based on the associated nodes, the structured knowledge nodes, and their context, the missing associated node content is generated; The completed structured knowledge node set is semantically aligned, conflict detected, and resolved with the existing structure tree in the FMEA database to update the FMEA database.

2. The FMEA data construction method based on cognitive digital twins according to claim 1, characterized in that, The process of parsing the document to be processed, extracting target information, and converting the target information into structured knowledge nodes includes: Identify and parse the text, and / or tables, and / or block diagrams, and / or flowcharts in the document to be processed; Based on the analysis results, target information corresponding to FMEA elements is identified and extracted. The FMEA elements include at least one of product components, functions, failure modes, failure causes, failure consequences, and control measures. The topological relationships in the tables, and / or the block diagrams, and / or the flowcharts are transformed into the association relationships between the structured knowledge nodes.

3. The FMEA data construction method based on cognitive digital twins according to claim 1, characterized in that, Before performing logical completeness analysis on the structured knowledge nodes and identifying missing related nodes, the method further includes: Identify the version type of the document to be processed; Based on the version type, the preset version adaptation rules are invoked to map the target information to the analysis framework required by the target FMEA standard version, so as to generate structured knowledge nodes that are consistent with the target version logic.

4. The FMEA data construction method based on cognitive digital twins according to claim 1, characterized in that, Before performing logical completeness analysis on the structured knowledge nodes and identifying missing related nodes, the method further includes: Retrieve historical case documents associated with the document to be processed; Based on semantic matching, the correlation between fault items and rectification items is extracted from the historical case documents, and the rectification items include the root cause and corrective measures. The aforementioned relationships are added to the current structured knowledge nodes as prior knowledge.

5. The FMEA data construction method based on cognitive digital twins according to claim 4, characterized in that, The semantic matching-based extraction of the association between fault items and rectification items from the historical case documents includes: Semantic parsing is performed on the historical case documents to identify the entities of fault items and rectification items; Calculate a semantic matching score based on the fault item entity and the rectification item entity; Based on the semantic matching scores, the association between fault items and rectification items is established, and a relationship graph is constructed. Extract the root cause sub-items and corrective action sub-items from the relationship graph to supplement the relationship.

6. The FMEA data construction method based on cognitive digital twins according to claim 1, characterized in that, The step of semantically aligning, detecting, and resolving conflicts between the completed structured knowledge node set and the existing structure tree in the FMEA database to update the FMEA database includes: A second language model is used to calculate the semantic similarity between the structured knowledge nodes and the existing structure tree nodes. The second language model is a similarity calculation model based on semantic vectors. Based on the semantic similarity, conflicting nodes that are semantically similar but inconsistent in expression are identified, and conflict prompt information is generated; In response to the conflict warning information, a third language model is used to analyze the context of the conflict node and generate standardized node terms; the third language model is a generative large language model. The standardized node terms are recommended to users for confirmation or automatically stored in the FMEA database.

7. The FMEA data construction method based on cognitive digital twins according to claim 1, characterized in that, Also includes: Before or after the structured knowledge node set is updated to the FMEA database, a fourth language model is used to predict the scores of the newly added or updated nodes; wherein, the fourth language model is a generative large language model. The score prediction includes: The fourth language model is provided with a description of the node to be predicted, historical scoring data in the FMEA database, and preset scoring criteria. Obtain the predicted evaluation values ​​of severity, frequency, and detectivity for the node output by the fourth language model, as well as the confidence intervals corresponding to each predicted evaluation value.

8. An electronic device, characterized in that, The electronic device includes: One or more processors; and A memory storing computer program instructions, which, when executed, cause the processor to perform the steps of the FMEA data construction method based on cognitive digital twins as described in any one of claims 1 to 7.

9. A computer-readable medium having a computer program / instructions stored thereon, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the FMEA data construction method based on cognitive digital twins as described in any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that, When executed by a processor, the computer program / instruction implements the steps of the FMEA data construction method based on cognitive digital twins as described in any one of claims 1 to 7.

Images