Intelligent elevator inspection clause matching method and system based on knowledge graph and medium
Through the elevator inspection method based on knowledge graph, combined with dynamic rule engine and AR interaction, the problems of low efficiency, large error and difficult traceability in traditional elevator inspection are solved, and the intelligent and reliable evidence storage of elevator inspection is realized.
Patent Information
- Application Number
- CN202510719588.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-30
- Publication Date
- 2025-08-15
AI Technical Summary
Traditional elevator inspection relies on manual visual inspection, and there are problems such as low efficiency, large subjective errors, difficulty in inspection and traceability, and unreliable data. The existing system has shortcomings in the face of multi-source heterogeneous data fusion, dynamic knowledge evolution and credible evidence storage throughout the process.
Using a knowledge graph-based method, the knowledge graph is built through digital mapping of components and terms, and dynamic rules engine configuration is combined with machine vision and AR interaction to achieve intelligent matching of elevator components and dynamic feedback and learning of inspection results.
It realizes intelligent closed-loop management of elevator inspection, improves inspection efficiency, reduces manual judgment time, ensures the credibility and traceability of inspection results, and adapts to the rapid update of new elevator components.
Smart Images
Figure CN120494812A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of elevator inspection technology, and in particular to a method, system, and medium for intelligently matching elevator inspection clauses based on a knowledge graph. Background Art
[0002] With the acceleration of urbanization, the safe operation of elevators, as specialized equipment, faces significant challenges. Traditional elevator inspections rely primarily on manual visual inspections, tool measurements, and paper records against technical specifications (such as TSG T7001). This presents three major technical bottlenecks: First, the complex interdependencies of inspection standards, with the mapping of components, clauses, and tools relying on empirical memory, makes manual review inefficient and prone to missed clauses and incorrect tool selection. Second, subjective judgment errors are significant, especially when determining key parameters such as door lock engagement depth and brake clearance, which lack real-time data-driven decision support. Third, the inspection process is difficult to trace, and paper reports are easily tampered with, making it difficult to meet regulatory authorities' mandatory audit requirements for the authenticity and integrity of inspection data.
[0003] In recent years, although database-based inspection assistance systems have emerged, they still have the following defects: (1) Insufficient knowledge relevance. Existing systems mostly use static retrieval models and cannot build a topological reasoning network between component features, inspection procedures, and inspection tools; (2) The level of human-machine collaboration intelligence is low. There is a lack of a real-time guidance mechanism that combines visual recognition and augmented reality (AR), and the training cycle for new inspectors is long; (3) Rule updates are delayed. When facing new elevator components (such as electromagnetic door locks), the code needs to be manually revised, and the system generalization ability is weak; (4) There is a lack of data credibility. Inspection records are mostly stored in centralized servers, which poses a risk of tampering and does not meet the legal effect requirements of the "Special Equipment Safety Law" on electronic evidence.
[0004] Existing technologies attempt to address these issues by introducing single technologies. For example, they employ optical character recognition (OCR) to automatically extract inspection report text, or utilize rule engines for simple logical analysis. However, these approaches fail to address core issues such as multi-source heterogeneous data fusion, dynamic knowledge evolution, and trusted evidence storage throughout the entire process. This is especially true when expanding inspection scenarios to include equipment such as pressure vessels and lifting machinery. Existing systems, due to their rigid architecture, require redevelopment, resulting in redundant resource investment. Summary of the Invention
[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a method, system and medium for intelligent matching of elevator inspection clauses based on knowledge graph.
[0006] The object of the present invention is achieved through the following technical solutions: In a first aspect, the present invention provides: an intelligent matching method for elevator inspection clauses based on a knowledge graph, comprising the following steps: In the digital mapping phase between components and clauses, the elevator components and inspection clauses are disassembled and annotated to establish a mapping relationship between the elevator components and the inspection clauses. During the knowledge graph construction phase, component nodes, clause nodes, and tool nodes are defined, and the elevator components, inspection clauses, and corresponding inspection tools are associated to obtain a knowledge graph. During the dynamic rule engine configuration phase, we dynamically load inspection clauses, convert them into programmable logic rules, query the inspection clauses associated with elevator components through the knowledge graph, and automatically generate Drools rule files. During the automatic matching phase of the inspection process, the elevator components to be identified are detected, and the inspection terms are matched and pushed according to the elevator components to be identified. The AR device is used for three-level intelligent guidance to complete the inspection and obtain the inspection results. During the dynamic feedback and learning phase of test results, the test results are recorded, the knowledge graph is incrementally updated, and the rule engine is optimized.
[0007] Preferably, the component and clause digital mapping stage further includes the following steps: Disassemble the elevator into its components and build a component library. Assign a unique component ID to each component and associate the component ID with the component attributes. Parse the inspection items in national standards, split the inspection clauses according to component ownership, store the clause parameters in a structured manner and establish mapping relationships.
[0008] Preferably, the component attributes include brand, model, historical inspection record and failure mode; and the clause parameters include inspection tools, thresholds and decision logic.
[0009] Preferably, the automatic matching stage of the inspection process further includes the following steps: Detect the elevator components to be identified through the machine vision model and obtain the component ID; Query the knowledge graph based on the component ID and return a list of inspection clauses associated with the component ID. The AR device marks the component location on the AR interface according to the inspection clause list and displays key indicators; when the user triggers the pop-up window through gestures, the original text of the clause, inspection tools and historical data curves are displayed; at the same time, voice assistance is used to complete the inspection.
[0010] Preferably, the dynamic feedback and learning phase of the test results further includes the following steps: Record inspection results based on database and blockchain evidence storage methods; For low-confidence samples, manual review is triggered. After review and confirmation, the component subcategory is automatically expanded and the knowledge graph is updated; The Drool rule weights are trained based on historical test data to improve the accuracy of complex logical judgments.
[0011] Preferably, the machine vision model is a YOLOv8n model.
[0012] The second aspect of the present invention is a knowledge graph-based intelligent matching system for elevator inspection clauses, which is used to implement any of the above-mentioned knowledge graph-based intelligent matching methods for elevator inspection clauses, including: The component and clause digital mapping module is used to disassemble and annotate elevator components and inspection clauses, and establish a mapping relationship between elevator components and inspection clauses; The knowledge graph construction module is used to define component nodes, clause nodes, and tool nodes, and associate elevator components, inspection clauses, and corresponding inspection tools to obtain a knowledge graph. Dynamic rule engine configuration module, used to dynamically load inspection clauses, convert inspection clauses into programmable logic rules, query inspection clauses associated with elevator components through knowledge graph, and automatically generate Drools rule files; The automatic matching module of the inspection process is used to detect the elevator parts to be identified, match and push the inspection terms according to the elevator parts to be identified, and use AR equipment to provide three-level intelligent guidance to complete the inspection and obtain the inspection results; The dynamic feedback and learning module for test results is used to record test results, incrementally update the knowledge graph, and optimize the rule engine.
[0013] The third aspect of the present invention: a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by a processor, any of the above-mentioned knowledge graph-based intelligent matching methods for elevator inspection terms is implemented.
[0014] The beneficial effects of the present invention are: 1) By integrating knowledge graphs (Neo4j), dynamic rule engines (Drools), machine vision (YOLOv8n), AR interaction (MediaPipe+Edge-TTS), and blockchain evidence storage technology, a complete intelligent elevator inspection chain was built. Accurate semantic reasoning was achieved through a topological relationship network of components, clauses, and tools, breaking through the inefficient model of traditional manual review of technical standards.
[0015] 2) After machine vision identifies components, the knowledge graph automatically pushes associated inspection terms and tools, reducing manual judgment time. AR devices mark key inspection areas and voice-announce operating steps, shortening training cycles and assisting non-experienced inspectors in completing high-precision tasks, significantly improving inspection efficiency.
[0016] 3) Dynamic rule engine to prevent misjudgments: This engine converts regulatory clauses (such as TSG T7001) into programmable logic rules, performs real-time threshold comparisons (e.g., triggering an alarm when the door lock engagement depth is less than 7mm), and avoids subjective experience bias. Trusted data storage: Inspection results are stored through hashing and blockchain to ensure that the original data cannot be tampered with, meeting the compliance audit requirements for mandatory inspections of special equipment.
[0017] 4) Incremental self-evolution of the knowledge graph: When a new component is discovered (such as a new brand of door lock), the system triggers a manual review mechanism through low-confidence samples, dynamically expands the component library, and updates the association rules; Generalization and migration potential: The method framework is highly adaptable and can be reused in other special equipment inspection scenarios such as pressure vessels and lifting machinery (only the component library and corresponding regulatory provisions need to be replaced). BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 FIG. 4 is a flow chart of a method in one embodiment of the present invention. DETAILED DESCRIPTION
[0019] The following will clearly and completely describe the technical solutions of the present invention in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.
[0020] See Figure 1 The first aspect of the present invention provides: an intelligent matching method for elevator inspection clauses based on a knowledge graph, comprising the following steps: In the digital mapping phase between components and clauses, the elevator components and inspection clauses are disassembled and annotated to establish a mapping relationship between the elevator components and the inspection clauses. During the knowledge graph construction phase, component nodes, clause nodes, and tool nodes are defined, and the elevator components, inspection clauses, and corresponding inspection tools are associated to obtain a knowledge graph. During the dynamic rule engine configuration phase, we dynamically load inspection clauses, convert them into programmable logic rules, query the inspection clauses associated with elevator components through the knowledge graph, and automatically generate Drools rule files. During the automatic matching phase of the inspection process, the elevator components to be identified are detected, and the inspection terms are matched and pushed according to the elevator components to be identified. The AR device is used for three-level intelligent guidance to complete the inspection and obtain the inspection results. During the dynamic feedback and learning phase of test results, the test results are recorded, the knowledge graph is incrementally updated, and the rule engine is optimized.
[0021] This embodiment achieves an intelligent, closed-loop management system for the entire process, encompassing a "perception-decision-execution-feedback" approach. The perception layer automatically identifies component type and status through machine vision. The decision layer uses a rule engine to compare thresholds and trigger the decision logic (pass / fail). The execution layer uses AR to guide operations step by step, with voice prompts for key steps. The feedback layer uses historical data to optimize rule weights and knowledge graph relationships, creating a self-iterative mechanism.
[0022] Utilizing refined component-clause mapping technology, elevators are broken down into a library of core components (door locks, traction motors, etc.). Unique IDs are then associated with attributes such as brand and failure mode, enabling precise mapping of inspection clauses to physical components. Through structured analysis and attribution analysis of inspection clauses, TSG standard clauses (such as A1.2.7.8) are dynamically assigned to specific components, establishing a scalable digital association system.
[0023] A knowledge-graph-driven intelligent matching mechanism: Based on a graph database (Neo4j), a topological relationship network consisting of components, clauses, and tools is constructed, supporting multi-dimensional queries (e.g., "all inspection clauses related to door locks"). Relationship definitions include REQUIRES_CLAUSE (component → clause), USES_TOOL (clause → tool), and HAS_THRESHOLD (clause threshold logic), enabling semantic reasoning.
[0024] Dynamic rule engine and automated inspection process: Drools rule template programming technology converts regulatory clauses into dynamically loadable logic rules (e.g., "deep engagement < 7mm is considered unqualified"). Rule generation and real-time graph integration: By querying component-related clauses through Neo4j, executable judgment rules are automatically generated, supporting complex logic (and / or conditional) expansion.
[0025] Multimodal interaction and intelligent decision-making support: Machine vision (YOLOv8n) enables real-time component recognition and ID triggering, connecting data between physical inspection and digital systems. AR hierarchical guidance (annotation display → threshold warning → tool invocation) combined with Edge-TTS voice broadcasting creates a closed-loop interactive "visual-audio-touch" integration.
[0026] A continuous optimization system driven by data feedback: Low-confidence samples automatically trigger a manual review mechanism, enabling incremental updates to the knowledge graph (e.g., adding component subcategories and associated terms). Rule weight training and parameter calibration based on historical inspection data improves judgment accuracy (e.g., dynamic optimization of safety gear triggering conditions).
[0027] In some embodiments, the component and item digital mapping stage further includes the following steps: Disassemble the elevator into its components and build a component library. Assign a unique component ID to each component and associate the component ID with the component attributes. Parse the inspection items in national standards, split the inspection clauses according to component ownership, store the clause parameters in a structured manner and establish mapping relationships.
[0028] In this embodiment, the national standard is TSG T7001-2023. Taking a door lock as an example, some codes based on the JSON format are as follows: ; Example of inspection clauses: Clause A1.2.7.8 (Door locking and closing) → Applicable component: Door lock. Clause A1.3.4 (Speed governor and safety gear test) → Applicable component: Speed governor, safety gear.
[0029] The nodes of the knowledge graph are defined as follows: Component node: Component (attributes: ID, type, brand). Clause node: Clause (attributes: number, content, threshold). Tool node: Tool (attributes: name, purpose). Part of the relationship definition code is as follows: ; Part of the code for the knowledge graph based on the Cypher graph database query language is as follows: ; The following is a partial code of the Drools rule using the door lock as an example: ; Part of the code of the Python-based example process for linking Drools rules with knowledge graphs is as follows: ; In some embodiments, the component attributes include brand, model, historical inspection record, and failure mode; and the item parameters include inspection tools, thresholds, and decision logic.
[0030] In this embodiment, elevator components are divided into several categories. Taking elevator supervisory inspection as an example, the correspondence between inspection items and elevator components is shown in Table 1. When the elevator component is identified as "information", the corresponding inspection items are A1.1.1, A1.1.2, and A1.1.3; when the elevator component is identified as "building passage", the corresponding inspection item is A1.2.1.1; and when the elevator component is identified as "lighting facility", the corresponding inspection item is also A1.2.1.1.
[0031] Table 1 Correspondence between inspection items and elevator components Table 1 Correspondence between inspection items and elevator components Table 1 Correspondence between inspection items and elevator components Table 1 Correspondence between inspection items and elevator components In some embodiments, the automatic matching stage of the inspection process further includes the following steps: Detect the elevator components to be identified through the machine vision model and obtain the component ID; Query the knowledge graph based on the component ID and return a list of inspection clauses associated with the component ID. The AR device marks the component location on the AR interface according to the inspection clause list and displays key indicators; when the user triggers the pop-up window through gestures, the original text of the clause, inspection tools and historical data curves are displayed; at the same time, voice assistance is used to complete the inspection.
[0032] In this embodiment, part of the code for querying the knowledge graph based on the component ID is as follows: ; The following code snippet returns a list of inspection terms associated with a part ID: ; The AR device's three levels of intelligent guidance include: Level 1 prompt: The AR interface marks the door lock location and displays key indicators (such as a red warning for "Engagement Depth: 5.2mm"). Level 2 prompt: A gesture-triggered pop-up window displays the original terms and conditions, a testing tool (vernier caliper), and historical data curves. Voice assistance: Edge-TTS announces, "Please use a vernier caliper to measure the door lock engagement depth; the standard value is ≥7mm."
[0033] In some embodiments, the dynamic feedback and learning phase of the test results further includes the following steps: Record inspection results based on database and blockchain evidence storage methods; For low-confidence samples, manual review is triggered. After review and confirmation, the component subcategory is automatically expanded and the knowledge graph is updated; The Drool rule weights are trained based on historical test data to improve the accuracy of complex logical judgments.
[0034] In this example, the inspection results (measurement values, image hashes) are written to a SQLite database. Part of the example code is as follows: ; Low-confidence samples typically include new brands of door locks, for which new nodes are added and associated with the same terms. Complex logic, such as "and / or" conditions, can also be used.
[0035] In some embodiments, the machine vision model is a YOLOv8n model.
[0036] The second aspect of the present invention is a knowledge graph-based intelligent matching system for elevator inspection clauses, which is used to implement any of the above-mentioned knowledge graph-based intelligent matching methods for elevator inspection clauses, including: The component and clause digital mapping module is used to disassemble and annotate elevator components and inspection clauses, and establish a mapping relationship between elevator components and inspection clauses; The knowledge graph construction module is used to define component nodes, clause nodes, and tool nodes, and associate elevator components, inspection clauses, and corresponding inspection tools to obtain a knowledge graph. Dynamic rule engine configuration module, used to dynamically load inspection clauses, convert inspection clauses into programmable logic rules, query inspection clauses associated with elevator components through knowledge graph, and automatically generate Drools rule files; The automatic matching module of the inspection process is used to detect the elevator parts to be identified, match and push the inspection terms according to the elevator parts to be identified, and use AR equipment to provide three-level intelligent guidance to complete the inspection and obtain the inspection results; The dynamic feedback and learning module for test results is used to record test results, incrementally update the knowledge graph, and optimize the rule engine.
[0037] The third aspect of the present invention: a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are loaded and executed by a processor, any of the above-mentioned knowledge graph-based intelligent matching methods for elevator inspection terms is implemented.
[0038] The foregoing description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein and should not be construed as excluding other embodiments. Rather, the present invention can be used in various other combinations, modifications, and environments and can be modified within the scope of the concept described herein through the above teachings or techniques or knowledge in the relevant field. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention are intended to be protected by the appended claims.
Claims
1. An intelligent matching method for elevator inspection clauses based on knowledge graph, characterized by: The following steps are involved: In the digital mapping phase between components and clauses, the elevator components and inspection clauses are disassembled and annotated to establish a mapping relationship between the elevator components and the inspection clauses. During the knowledge graph construction phase, component nodes, clause nodes, and tool nodes are defined, and the elevator components, inspection clauses, and corresponding inspection tools are associated to obtain a knowledge graph. During the dynamic rule engine configuration phase, we dynamically load inspection clauses, convert them into programmable logic rules, query the inspection clauses associated with elevator components through the knowledge graph, and automatically generate Drools rule files. During the automatic matching phase of the inspection process, the elevator components to be identified are detected, and the inspection terms are matched and pushed according to the elevator components to be identified. The AR device is used for three-level intelligent guidance to complete the inspection and obtain the inspection results. During the dynamic feedback and learning phase of test results, the test results are recorded, the knowledge graph is incrementally updated, and the rule engine is optimized.
2. The intelligent matching method for elevator inspection clauses based on knowledge graph according to claim 1 is characterized by: The component and item digital mapping phase also includes the following steps: Disassemble the elevator into its components and build a component library. Assign a unique component ID to each component and associate the component ID with the component attributes. Parse the inspection items in national standards, split the inspection clauses according to component ownership, store the clause parameters in a structured manner and establish mapping relationships.
3. The intelligent matching method for elevator inspection clauses based on knowledge graph according to claim 2 is characterized by: The component attributes include brand, model, historical inspection record and failure mode; the item parameters include inspection tools, thresholds and decision logic.
4. The knowledge graph-based intelligent matching method for elevator inspection clauses according to claim 1 is characterized by: The automatic matching stage of the inspection process also includes the following steps: Detect the elevator components to be identified through the machine vision model and obtain the component ID; Query the knowledge graph based on the component ID and return a list of inspection clauses associated with the component ID. The AR device marks the component location on the AR interface according to the inspection clause list and displays key indicators; when the user triggers the pop-up window through gestures, the original text of the clause, inspection tools and historical data curves are displayed; at the same time, voice assistance is used to complete the inspection.
5. The intelligent matching method for elevator inspection clauses based on knowledge graph according to claim 1 is characterized by: The dynamic feedback and learning phase of the test results also includes the following steps: Record inspection results based on database and blockchain evidence storage methods; For low-confidence samples, manual review is triggered. After review and confirmation, the component subcategory is automatically expanded and the knowledge graph is updated; The Drool rule weights are trained based on historical test data to improve the accuracy of complex logical judgments.
6. The intelligent matching method for elevator inspection clauses based on knowledge graph according to claim 4 is characterized by: The machine vision model is the YOLOv8n model.
7. The intelligent matching system for elevator inspection clauses based on knowledge graph is characterized by: The method for intelligently matching elevator inspection clauses based on a knowledge graph according to any one of claims 1 to 6 comprises: The component and clause digital mapping module is used to disassemble and annotate elevator components and inspection clauses, and establish a mapping relationship between elevator components and inspection clauses; The knowledge graph construction module is used to define component nodes, clause nodes, and tool nodes, and associate elevator components, inspection clauses, and corresponding inspection tools to obtain a knowledge graph. Dynamic rule engine configuration module, used to dynamically load inspection clauses, convert inspection clauses into programmable logic rules, query inspection clauses associated with elevator components through knowledge graph, and automatically generate Drools rule files; The automatic matching module of the inspection process is used to detect the elevator parts to be identified, match and push the inspection terms according to the elevator parts to be identified, and use AR equipment to provide three-level intelligent guidance to complete the inspection and obtain the inspection results; The dynamic feedback and learning module for test results is used to record test results, incrementally update the knowledge graph, and optimize the rule engine.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions. When the computer-executable instructions are loaded and executed by the processor, the intelligent matching method for elevator inspection clauses based on the knowledge graph as described in any one of claims 1 to 6 is implemented.