ISO authentication auditing intelligent auxiliary method and system based on AI knowledge enhancement

Abstract

The invention discloses an ISO authentication auditing intelligent auxiliary method and system based on AI knowledge enhancement, and the method comprises the steps: constructing a domain knowledge graph based on the standard structured terms of an ISO management system, and endowing the semantic association between nodes with a learnable weight; performing natural language processing on the input data, identifying and extracting auditing elements, and performing semantic matching on the auditing elements and nodes in the domain knowledge graph to generate a clause candidate set; performing multi-dimensional reasoning on the clause candidate set, and calculating a comprehensive confidence coefficient to determine a risk grading result; collecting correction data output by an auditor to the system, and dynamically adjusting node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism based on errors of the correction data and a system prediction result; automatically generating an ISO authentication auditing report according to the comprehensive confidence coefficient and the risk grading result; manual revision records and supervision feedback of the audit report are collected, and AI audit is driven to execute incremental learning.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology and management system certification audit technology, specifically to an intelligent auxiliary method and system for ISO certification audit based on AI knowledge enhancement. Background Technology

[0002] In the current ISO management system certification industry, auditing activities primarily rely on the auditor's individual professional knowledge and experience. Auditors need to manually read through a large number of documents and records submitted by the company, and combine this with on-site audit evidence collection, comparing and judging the collected information against complex ISO standard clauses. This traditional manual auditing model has the following significant limitations: (1) Limited audit resources and uneven capabilities: The number of high-level audit experts is limited, while the audit workload continues to increase, leading to a shortage of auditor resources. At the same time, differences in knowledge background and experience among different auditors make it difficult to unify the understanding and application of the same standard, and the consistency of audit conclusions is difficult to guarantee.

[0003] (2) Fragmentation of knowledge system and difficulty in inheritance: Audit knowledge is mostly found in the personal experience of auditors and internal documents, which is highly fragmented and difficult to systematically accumulate, manage and reuse. This makes it difficult for the organization to accumulate knowledge assets and the training cycle for new employees is long.

[0004] (3) Low audit efficiency and high subjectivity: The entire process of document review, clause matching and report writing is time-consuming and labor-intensive, resulting in low audit efficiency. At the same time, the audit conclusions are largely influenced by the auditor's subjective judgment, making it difficult to achieve objective and quantifiable risk assessment.

[0005] (4) Insufficient standardization of reports and difficulties in regulatory coordination: Manually written audit reports are prone to inconsistencies in format, wording and clause citations, which do not meet the increasingly high digitalization and standardization requirements of regulatory agencies, and bring obstacles to the automatic collection and analysis of regulatory data.

[0006] While some existing systems utilize artificial intelligence for compliance auditing (as described in Comparative Document 1), these solutions are primarily geared towards general internal corporate compliance reporting or online content review. Their rule-matching and semantic understanding-based models are ill-suited to the multi-layered standard clause system, complex audit logic, and experience-dependent judgment processes unique to ISO certification audits. They lack the ability to perform deep semantic modeling and knowledge network construction for ISO standard clauses, and cannot achieve continuous self-optimization based on audit practice feedback. Therefore, a significant technological gap exists in the existing technology, urgently requiring an intelligent auxiliary solution specifically designed for the ISO certification audit field, possessing knowledge enhancement and learning evolution capabilities. Summary of the Invention

[0007] To achieve the objectives of this invention, this application provides an intelligent assistance method for ISO certification audits based on AI knowledge enhancement, comprising: Step S1: Based on the structured clauses of the ISO management system standard, construct a domain knowledge graph containing standard clause nodes, audit evidence nodes, and industry knowledge nodes, and assign learnable weights to the semantic relationships between nodes; Step S2: Perform natural language processing on the input ISO certification audit materials, identify and extract audit elements, and semantically match the audit elements with nodes in the domain knowledge graph to generate a candidate set of clauses; Step S3: Perform multi-dimensional reasoning on the candidate clause set, calculate the comprehensive confidence level of each audit element and the corresponding standard clause node, and determine the risk classification result based on the numerical range of the comprehensive confidence level; Step S4: Collect the corrected data of the auditors on the system output, and based on the error between the corrected data and the system prediction results, dynamically adjust the node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism; Step S5: Based on the comprehensive confidence level and risk classification results, call the template that conforms to CNAS / CNCA standards to automatically generate a structured ISO certification audit report; Step S6: Collect records of manual revisions to audit reports and regulatory feedback to drive incremental learning in AI audit execution, thereby achieving continuous optimization of the system's audit knowledge.

[0008] In some specific embodiments, step S1 includes: Step S11: Parse the ISO standard text to extract chapter numbers, clause titles, and key requirements; Step S12: Create standard clause nodes, audit evidence nodes, and industry knowledge nodes, and generate semantic vectors for each node; Step S13: Based on the cosine similarity between the semantic vectors of the nodes, the relevance of expert annotations, and the co-occurrence frequency of historical review samples, calculate the comprehensive association weight between nodes to construct the domain knowledge graph.

[0009] In some specific embodiments, in step S3, the overall confidence level is calculated using the following formula: Conf = λ1 * S + λ2 * T + λ3 * H Where Conf is the overall confidence level, S is the semantic similarity between the audit element and the standard clause node, T is the time decay factor based on knowledge timeliness, H is the historical consistency score, and λ1, λ2, and λ3 are adjustable weight parameters, and λ1 + λ2 + λ3 = 1.

[0010] In some specific embodiments, in step S3, the risk classification result is determined based on the following rules: When Conf < 0.6, it is judged as a Class A high-risk level; When 0.6 ≤ Conf < 0.8, it is classified as a medium-risk level (B). When Conf ≥ 0.8, it is judged as a low-risk level (C).

[0011] In some specific embodiments, in step S4, the reinforcement learning mechanism dynamically adjusts the node weights and model parameters in the domain knowledge graph based on the following reward function, which is determined according to the following formula: R = e -|Δ| -λ* TimeCost Where R is the reward value, Δ is the error between the corrected data and the system prediction result, TimeCost is the time penalty term, and λ is the time penalty coefficient.

[0012] In some specific embodiments, step S5 includes: dynamically selecting and filling a preset language template based on the risk classification result using natural language generation technology, wherein: For Level C low risk, generate a concluding statement that "meets standard requirements"; For Level B medium risk, generate a prompt statement such as "Some evidence is insufficient, improvements are recommended"; For Level A high-risk cases, a warning message "Requirements not met, rectification required" will be generated.

[0013] In some specific embodiments, step S6 includes: Step S61: Compare the difference vectors between versions of the audit report to automatically identify the knowledge update areas; Step S62: Determine the weights of the changed nodes in the updated region according to the following formula to perform weight correction, w i' =w i +η*Δi, where w i Let η be the node weight, η be the learning rate, and Δi be the feedback error term; Step S63: Generate a knowledge network update version number based on the revised content and store the change log to achieve traceability.

[0014] To achieve the same inventive objective, this application also provides an AI-based knowledge-enhanced intelligent assistance system for ISO certification audits, comprising: Standard building module: Used to build a domain knowledge graph based on the structured clauses of ISO management system standards, including standard clause nodes, audit evidence nodes, and industry knowledge nodes, and to assign learnable weights to the semantic relationships between nodes; Element extraction module: Used to perform natural language processing on the input ISO certification audit materials, identify and extract audit elements, and perform semantic matching of the audit elements with nodes in the domain knowledge graph to generate a candidate set of clauses; Confidence assessment module: used to perform multi-dimensional reasoning on the candidate clause set, calculate the comprehensive confidence level of each audit element and the corresponding standard clause node, and determine the risk classification result based on the numerical range of the comprehensive confidence level; The reinforcement adjustment module is used to collect the correction data of the auditors to the system output, and dynamically adjust the node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism based on the error between the correction data and the system prediction results. Report generation module: Based on the comprehensive confidence level and risk classification results, it calls a template that conforms to CNAS / CNCA standards to automatically generate a structured ISO certification audit report; Optimize feedback module: Used to collect records of manual revisions to audit reports and regulatory feedback, driving incremental learning of AI audit execution to achieve continuous optimization of the system's audit knowledge.

[0015] In some specific embodiments, the standard building block is used to perform the following steps: Step S11: Parse the ISO standard text to extract chapter numbers, clause titles, and key requirements; Step S12: Create standard clause nodes, audit evidence nodes, and industry knowledge nodes, and generate semantic vectors for each node; Step S13: Based on the cosine similarity between the semantic vectors of the nodes, the relevance of expert annotations, and the co-occurrence frequency of historical review samples, calculate the comprehensive association weight between nodes to construct the domain knowledge graph.

[0016] In some specific embodiments, the optimization feedback module is used to perform the following steps: Step S61: Compare the difference vectors between versions of the audit report to automatically identify the knowledge update areas; Step S62: Determine the weights of the changed nodes in the updated region according to the following formula to perform weight correction, w i' =w i +η*Δi, where w i Let η be the node weight, η be the learning rate, and Δi be the feedback error term; Step S63: Generate a knowledge network update version number based on the revised content and store the change log to achieve traceability.

[0017] The beneficial effects of the above technical solution are as follows: Compared with existing technologies, the AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing provided by this invention has the following significant advantages: (1) Significantly improved audit capabilities and efficiency: Through AI semantic understanding and automated report generation, auditors are freed from the heavy workload of document reading and report writing, resulting in an overall audit efficiency improvement of over 40%. At the same time, the clause matching and risk warnings provided by the system enable ordinary auditors to have near-expert-level judgment capabilities with the assistance of AI.

[0018] (2) Enhanced consistency and objectivity of audit conclusions: Through a quantitative confidence assessment model and a unified risk classification rule, the subjective judgment differences between different auditors were effectively reduced, ensuring the consistency and objectivity of audit conclusions, with a consistency improvement of about 35%.

[0019] (3) Knowledge assetization and continuous evolution: This invention transforms fragmented auditing experience into a structured, iteratively updatable domain knowledge graph. Through reinforcement learning, the system can learn from each auditing practice, dynamically optimize its knowledge network and judgment model, achieve continuous accumulation and autonomous evolution of auditing knowledge, and effectively respond to standard updates and industry changes.

[0020] (4) Report standardization and regulatory compliance: The automatically generated audit reports strictly follow the standard templates of regulatory agencies such as CNAS / CNCA, with unified format, accurate citation of clauses, and standardized language expression, which greatly improves the quality of reports and facilitates digital connection with regulatory platforms.

[0021] (5) Precise risk positioning and decision support: Through comprehensive confidence level calculation and A / B / C three-level risk classification, the system can accurately locate high-risk links and potential non-compliance items in the audit, and provide auditors with clear decision support and rectification suggestions, effectively reducing the enterprise's compliance risk. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating an AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing, provided as an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of an AI-based knowledge-enhanced intelligent assistance system for ISO certification auditing, provided as an embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0025] Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar symbols denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the invention, and should not be construed as limiting the invention.

[0026] Example 1 One embodiment of the present invention provides an intelligent assistance method for ISO certification auditing based on AI knowledge enhancement, referring to... Figure 1 As shown, it includes: Step S1: Based on the structured clauses of the ISO management system standard, construct a domain knowledge graph containing standard clause nodes, audit evidence nodes, and industry knowledge nodes, and assign learnable weights to the semantic relationships between nodes; In a specific embodiment of the present invention, step S1 includes: Step S11: Parse the ISO standard text to extract chapter numbers, clause titles, and key requirements; Step S12: Create standard clause nodes, audit evidence nodes, and industry knowledge nodes, and generate semantic vectors for each node; Step S13: Based on the cosine similarity between the semantic vectors of the nodes, the relevance of expert annotations, and the co-occurrence frequency of historical review samples, calculate the comprehensive association weight between nodes to construct the domain knowledge graph.

[0027] Step S2: Perform natural language processing on the input ISO certification audit materials, identify and extract audit elements, and semantically match the audit elements with nodes in the domain knowledge graph to generate a candidate set of clauses; Its workflow includes: (1) Use text recognition and natural language processing algorithms to perform semantic segmentation and feature extraction on the audit materials; (2) Identify audit elements (object, process, evidence, clause reference) through pre-trained language models (such as Finetuned BERT or ERNIE); (3) The extracted audit elements are stored in the Audit Element Database (Audit Element DB) for subsequent matching and reasoning; (4) Supports speech-to-text and OCR image text recognition, suitable for on-site audit application scenarios.

[0028] Step S3: Perform multi-dimensional reasoning on the candidate clause set, calculate the comprehensive confidence level of each audit element and the corresponding standard clause node, and determine the risk classification result based on the numerical range of the comprehensive confidence level; In a specific embodiment of the present invention, in step S3, the comprehensive confidence level is calculated using the following formula: Conf = λ1 * S + λ2 * T + λ3 * H Where Conf is the overall confidence level, S is the semantic similarity between the audit element and the standard clause node, T is the time decay factor based on knowledge timeliness, H is the historical consistency score, and λ1, λ2, and λ3 are adjustable weight parameters, and λ1 + λ2 + λ3 = 1.

[0029] In a specific embodiment of the present invention, in step S3, the risk classification result is determined based on the following rules: When Conf < 0.6, it is judged as a Class A high-risk level; When 0.6 ≤ Conf < 0.8, it is classified as a medium-risk level (B). When Conf ≥ 0.8, it is judged as a low-risk level (C).

[0030] Step S4: Collect the corrected data of the auditors on the system output, and based on the error between the corrected data and the system prediction results, dynamically adjust the node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism; In a specific embodiment of the present invention, in step S4, the node weights and model parameters in the domain knowledge graph are dynamically adjusted based on the following reward function, wherein the reward function is determined according to the following formula: R = e -|Δ| -λ* TimeCost Where R is the reward value, Δ is the error between the corrected data and the system prediction result, TimeCost is the time penalty term, and λ is the time penalty coefficient.

[0031] Step S5: Based on the comprehensive confidence level and risk classification results, call the template that conforms to CNAS / CNCA standards to automatically generate a structured ISO certification audit report; In a specific embodiment of the present invention, step S5 includes: dynamically selecting and filling a preset language template based on the risk classification result using natural language generation technology, wherein: For Level C low risk, generate a concluding statement that "meets standard requirements"; For Level B medium risk, generate a prompt statement such as "Some evidence is insufficient, improvements are recommended"; For Level A high-risk cases, a warning message "Requirements not met, rectification required" will be generated.

[0032] Step S6: Collect records of manual revisions to audit reports and regulatory feedback to drive incremental learning in AI audit execution, thereby achieving continuous optimization of the system's audit knowledge.

[0033] In a specific embodiment of the present invention, step S6 includes: Step S61: Compare the difference vectors between versions of the audit report to automatically identify the knowledge update areas; Step S62: Determine the weights of the changed nodes in the updated region according to the following formula to perform weight correction, w i' =w i +η*Δi, where w i Let η be the node weight, η be the learning rate, and Δi be the feedback error term; Step S63: Generate a knowledge network update version number based on the revised content and store the change log to achieve traceability.

[0034] Example 2 One embodiment of the present invention provides an intelligent auxiliary system for ISO certification auditing based on AI knowledge enhancement, referring to... Figure 2 As shown, it includes: Standard Construction Module 10: Used to construct a domain knowledge graph based on the structured clauses of ISO management system standards, including standard clause nodes, audit evidence nodes, and industry knowledge nodes, and to assign learnable weights to the semantic relationships between nodes; In one specific embodiment of the present invention, the standard building module 10 is used to perform the following steps: Step S11: Parse the ISO standard text to extract chapter numbers, clause titles, and key requirements; Step S12: Create standard clause nodes, audit evidence nodes, and industry knowledge nodes, and generate semantic vectors for each node; Step S13: Based on the cosine similarity between the semantic vectors of the nodes, the relevance of expert annotations, and the co-occurrence frequency of historical review samples, calculate the comprehensive association weight between nodes to construct the domain knowledge graph.

[0035] Element extraction module 20: Used to perform natural language processing on the input ISO certification audit materials, identify and extract audit elements, and perform semantic matching of the audit elements with nodes in the domain knowledge graph to generate a candidate set of clauses; (1) Use text recognition and natural language processing algorithms to perform semantic segmentation and feature extraction on the audit materials; (2) Identify audit elements (object, process, evidence, clause reference) through pre-trained language models (such as Finetuned BERT or ERNIE); (3) The extracted audit elements are stored in the Audit Element Database (Audit Element DB) for subsequent matching and reasoning; (4) Supports speech-to-text and OCR image text recognition, suitable for on-site audit application scenarios.

[0036] Confidence assessment module 30: used to perform multi-dimensional reasoning on the candidate clause set, calculate the comprehensive confidence level of each audit element and the corresponding standard clause node, and determine the risk classification result based on the numerical range of the comprehensive confidence level; (1) Using knowledge graph reasoning algorithms, perform multi-level semantic comparison between audit elements and standard clause nodes; (2) Calculate the matching score of the review clauses based on sentence vector similarity and semantic distance; (3) Generate a confidence matrix for the terms and classify the risk of the deviation items (A / B / C). (4) Form an intelligent audit result set, including "clause number - matching degree - risk level - remarks"; (5) The output results can be directly entered into the report generation module or be manually reviewed.

[0037] Reinforcement adjustment module 40: used to collect the correction data of the auditor to the system output, and based on the error between the correction data and the system prediction result, dynamically adjust the node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism; (1) The system collects data samples such as the auditor's historical operations and expert review conclusions to form a "human-machine collaborative" learning corpus; (2) Adjust the node weights and semantic connection strength of the knowledge network through a lightweight knowledge enhancement algorithm; (3) Regularly retrain the model to achieve dynamic optimization of knowledge; (4) Support cross-system knowledge transfer (such as transferring experience from the quality system to the environmental system).

[0038] This module enables the system to continuously improve its review and judgment capabilities after multiple rounds of use, achieving the enhanced capability effect of "ordinary reviewers + AI = expert review team".

[0039] Report generation module 50: Based on the comprehensive confidence level and risk classification results, it calls a template that conforms to CNAS / CNCA standards to automatically generate a structured ISO certification audit report; (1) Based on the templates specified in CNAS-CC01 and CNCA-28, the system automatically integrates the audit results to generate a structured report; (2) The report should include: conforming items, non-conforming items, observation items, and recommended measures; (3) Supports bilingual output in Chinese and English, customizable report templates, and electronic signature interface; (4) The report can be directly uploaded to the regulatory platform to achieve data interconnection.

[0040] Optimize Feedback Module 60: Used to collect records of manual revisions to audit reports and regulatory feedback, driving incremental learning in AI audit execution to achieve continuous optimization of the system's audit knowledge.

[0041] In one specific embodiment of the present invention, the optimization feedback module 60 is used to perform the following steps: Step S61: Compare the difference vectors between versions of the audit report to automatically identify the knowledge update areas; Step S62: Determine the weights of the changed nodes in the updated region according to the following formula to perform weight correction, w i' =w i +η*Δi, where w i Let η be the node weight, η be the learning rate, and Δi be the feedback error term; Step S63: Generate a knowledge network update version number based on the revised content and store the change log to achieve traceability.

[0042] (1) The system collects manually corrected data, audit conclusion feedback, and regulatory rectification opinions; (2) Automatically update the attributes and weights of knowledge network nodes; (3) Incremental learning is carried out at the AI ​​model level to continuously optimize the system's knowledge structure; (4) Form a traceable knowledge iteration record to achieve a continuous improvement closed loop.

[0043] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0044] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the invention. Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0045] The methods and apparatus provided by the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

[0046] In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "a specific embodiment" or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0047] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A smart assistance method for ISO certification auditing based on AI knowledge enhancement, characterized in that, include: Step S1: Based on the structured clauses of the ISO management system standard, construct a domain knowledge graph containing standard clause nodes, audit evidence nodes, and industry knowledge nodes, and assign learnable weights to the semantic relationships between nodes; Step S2: Perform natural language processing on the input ISO certification audit materials, identify and extract audit elements, and semantically match the audit elements with nodes in the domain knowledge graph to generate a candidate set of clauses; Step S3: Perform multi-dimensional reasoning on the candidate clause set, calculate the comprehensive confidence level of each audit element and the corresponding standard clause node, and determine the risk classification result based on the numerical range of the comprehensive confidence level; Step S4: Collect the corrected data of the auditors on the system output, and based on the error between the corrected data and the system prediction results, dynamically adjust the node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism; Step S5: Based on the comprehensive confidence level and risk classification results, call the template that conforms to the standard specifications to automatically generate a structured ISO certification audit report; Step S6: Collect records of manual revisions to audit reports and regulatory feedback to drive incremental learning in AI audit execution, thereby achieving continuous optimization of the system's audit knowledge.

2. The AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing as described in claim 1, characterized in that, Step S1 includes: Step S11: Parse the ISO standard text to extract chapter numbers, clause titles, and key requirements; Step S12: Create standard clause nodes, audit evidence nodes, and industry knowledge nodes, and generate semantic vectors for each node; Step S13: Based on the cosine similarity between the semantic vectors of the nodes, the relevance of expert annotations, and the co-occurrence frequency of historical review samples, calculate the comprehensive association weight between nodes to construct the domain knowledge graph.

3. The AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing as described in claim 1, characterized in that, In step S3, the overall confidence level is calculated using the following formula: Conf = λ1 * S + λ2 * T + λ3 * H Where Conf is the overall confidence level, S is the semantic similarity between the audit element and the standard clause node, T is the time decay factor based on knowledge timeliness, H is the historical consistency score, and λ1, λ2, and λ3 are adjustable weight parameters, and λ1 + λ2 + λ3 = 1.

4. The AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing as described in claim 1, characterized in that, In step S3, the risk classification result is determined based on the following rules: When Conf < 0.6, it is judged as a Class A high-risk level; When 0.6 ≤ Conf < 0.8, it is classified as a medium-risk level (B). When Conf ≥ 0.8, it is judged as a low-risk level (C).

5. The AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing according to claim 1, characterized in that, In step S4, the reinforcement learning mechanism dynamically adjusts the node weights and model parameters in the domain knowledge graph based on the following reward function, which is determined according to the following formula: R = e -|Δ| -λ* TimeCost Where R is the reward value, Δ is the error between the corrected data and the system prediction result, TimeCost is the time penalty term, and λ is the time penalty coefficient.

6. The AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing according to claim 4, characterized in that, Step S5 includes: dynamically selecting and filling a preset language template based on the risk classification result using natural language generation technology, wherein: For Level C low risk, generate a concluding statement that "meets standard requirements"; For Level B medium risk, generate a prompt statement such as "Some evidence is insufficient, improvements are recommended"; For Level A high-risk cases, generate a warning statement: "Requirements not met, rectification required." 7. The AI-based knowledge-enhanced intelligent assistance method for ISO certification auditing according to claim 1, characterized in that, Step S6 includes: Step S61: Compare the difference vectors between versions of the audit report to automatically identify the knowledge update areas; Step S62: Determine the weights of the changed nodes in the updated region according to the following formula to perform weight correction, w i ' = w i +η*Δi, where w i Let η be the node weight, η be the learning rate, and Δi be the feedback error term; Step S63: Generate a knowledge network update version number based on the revised content and store the change log to achieve traceability.

8. An intelligent auxiliary system for ISO certification auditing based on AI knowledge enhancement, characterized in that, include: Standard building module: Used to build a domain knowledge graph based on the structured clauses of ISO management system standards, including standard clause nodes, audit evidence nodes, and industry knowledge nodes, and to assign learnable weights to the semantic relationships between nodes; Element extraction module: Used to perform natural language processing on the input ISO certification audit materials, identify and extract audit elements, and perform semantic matching of the audit elements with nodes in the domain knowledge graph to generate a candidate set of clauses; Confidence assessment module: used to perform multi-dimensional reasoning on the candidate clause set, calculate the comprehensive confidence level of each audit element and the corresponding standard clause node, and determine the risk classification result based on the numerical range of the comprehensive confidence level; The reinforcement adjustment module is used to collect the correction data of the auditors to the system output, and dynamically adjust the node weights and model parameters in the domain knowledge graph through a reinforcement learning mechanism based on the error between the correction data and the system prediction results. Report generation module: Based on the comprehensive confidence level and risk classification results, it calls a template that conforms to CNAS / CNCA standards to automatically generate a structured ISO certification audit report; Optimize feedback module: Used to collect records of manual revisions to audit reports and regulatory feedback, driving incremental learning of AI audit execution to achieve continuous optimization of the system's audit knowledge.

9. The AI-based knowledge-enhanced ISO certification audit intelligent assistance system according to claim 8, characterized in that, The standard building block is used to perform the following steps: Step S11: Parse the ISO standard text to extract chapter numbers, clause titles, and key requirements; Step S12: Create standard clause nodes, audit evidence nodes, and industry knowledge nodes, and generate semantic vectors for each node; Step S13: Based on the cosine similarity between the semantic vectors of the nodes, the relevance of expert annotations, and the co-occurrence frequency of historical review samples, calculate the comprehensive association weight between nodes to construct the domain knowledge graph.

10. The AI-based knowledge-enhanced ISO certification audit intelligent assistance system according to claim 8, characterized in that, The optimization feedback module is used to perform the following steps: Step S61: Compare the difference vectors between versions of the audit report to automatically identify the knowledge update areas; Step S62: Determine the weights of the changed nodes in the updated region according to the following formula to perform weight correction, w i ' = w i +η*Δi, where w i Let η be the node weight, η be the learning rate, and Δi be the feedback error term; Step S63: Generate a knowledge network update version number based on the revised content and store the change log to achieve traceability.

Images