An unmanned aerial vehicle inspection knowledge graph reasoning safety monitoring method

Abstract

The application discloses an unmanned aerial vehicle (UAV) inspection knowledge graph reasoning safety monitoring method, and belongs to the technical field of UAV intelligent inspection and knowledge graph reasoning. In view of the problems of lack of dynamic correlation modeling, insufficient abnormal propagation analysis and limited safety risk assessment capacity of complex operation environment in the prior art, the application constructs an inspection environment knowledge graph by analyzing inspection data, extracts real-time topological data and performs space-time alignment; correlation reasoning is performed by applying a graph attention network and a path reasoning algorithm to obtain an abnormal propagation path; multi-factor linear weighted risk assessment is performed based on the path to obtain a risk index, a risk grade and a risk range; finally, a graded early warning is triggered according to the risk index, and a disposal scheme is pushed. The application realizes dynamic reasoning and quantitative evaluation of operation environment risks, improves the real-time performance of safety monitoring, and is suitable for fields of UAV intelligent inspection, safety monitoring and knowledge graph reasoning.

Description

Technical Field

[0001] This invention belongs to the field of UAV intelligent inspection and knowledge graph reasoning technology, specifically involving a UAV inspection knowledge graph reasoning safety monitoring technology. Background Technology

[0002] With the development of drone inspection, image recognition, and knowledge graph technologies, inspection systems are gradually evolving from single image acquisition and defect identification towards multi-source data fusion, knowledge representation, and intelligent early warning. Existing technologies can already utilize drones to acquire inspection images, environmental parameters, and location information, and combine them with intelligent analysis models to identify and issue early warnings for faults, defects, or abnormal states.

[0003] However, most existing technologies still focus on fault detection, defect identification, or matching of remediation solutions, lacking systematic modeling of the relationship between equipment entities and environmental elements, making it difficult to reveal the risk propagation process in complex operating environments. Furthermore, while some solutions incorporate knowledge graphs, they are primarily used for defect information retrieval and remediation assistance, failing to achieve continuous reasoning about dynamic risks and overall situational analysis.

[0004] Furthermore, existing technologies still have shortcomings in processing the spatiotemporal correspondence between inspection data and the environment, identifying abnormal propagation paths, and jointly outputting risk levels and impact ranges, resulting in limited real-time safety risk assessment capabilities for complex operating environments. Therefore, there is an urgent need for an integrated safety monitoring method that can construct knowledge graphs, perform correlation reasoning, risk assessment, and safety early warning for inspection data in UAV operating environments. Summary of the Invention

[0005] To address the shortcomings of existing technologies, such as the lack of dynamic correlation modeling for complex operating environments, insufficient anomaly propagation analysis, and limited real-time safety risk assessment capabilities, this invention proposes the following solutions: A method for safety monitoring of unmanned aerial vehicle (UAV) inspections using knowledge graph reasoning, the method comprising: S1. Obtain inspection data of the drone's operating environment; S2. Parse the inspection data to obtain equipment entities and environmental element entities, and construct an inspection environment knowledge graph based on the equipment entities and environmental element entities. S3. Extract real-time topology data based on the inspection environment knowledge graph, and perform spatiotemporal alignment between the inspection data and the real-time topology data to obtain spatiotemporally aligned topology data. S4. Based on the spatiotemporally aligned topological data, perform correlation reasoning to obtain the anomaly propagation path; S5. Based on the abnormal propagation path, conduct a risk assessment to obtain the risk index, risk level, and scope of impact. S6. Trigger a safety warning based on the risk index, risk level, and scope of impact.

[0006] Furthermore, the inspection data mentioned in S1 is real-time multimodal data collected by the UAV inspection module. The real-time multimodal data includes visible light image data, infrared thermal imaging data, temperature and humidity data, air pressure data, gas concentration data, and positioning and navigation data.

[0007] Furthermore, S1, acquiring inspection data of the UAV operating environment includes collecting real-time multimodal data of the operating environment through the UAV inspection module, and performing sensor data fusion and noise filtering on the real-time multimodal data to obtain preprocessed inspection data; wherein, the sensor data fusion and noise filtering includes using a Kalman filter algorithm to perform state estimation and update on the real-time multimodal data.

[0008] Furthermore, the step S2, which involves constructing an inspection environment knowledge graph based on the equipment entities and environmental element entities, includes extracting entities from the inspection data to obtain the equipment entities and environmental element entities, constructing an entity topology structure between the equipment entities and environmental element entities based on the spatiotemporal dimension, and storing triplet data of equipment status, environmental parameters, and safety rules based on the entity topology structure.

[0009] Furthermore, the entity topology structure between the device entity and the environmental element entity constructed based on the spatiotemporal dimension as described in S2 includes defining the device entity, the environmental entity and their relationships through the ontology modeling unit, and completing the entity relationship construction through the dynamic relationship extraction unit and the semantic fusion unit.

[0010] Furthermore, the real-time topology data mentioned in S3 is data extracted based on the entity topology structure between equipment entities and environmental element entities in the inspection environment knowledge graph.

[0011] Furthermore, S4 describes performing correlation reasoning based on the spatiotemporally aligned topological data to obtain anomaly propagation paths, including applying graph attention networks to calculate the correlation degree of environmental elements; and identifying anomaly propagation paths through path reasoning algorithms.

[0012] Furthermore, S5 describes conducting a risk assessment based on the abnormal propagation path to obtain a risk index, risk level, and impact range. This includes inputting the correlation between environmental elements and the abnormal propagation path into a dynamic risk assessment model, and obtaining the risk index through multi-factor linear weighted assessment.

[0013] Furthermore, S6 triggers a security alert based on the aforementioned risk index, risk level, and scope of impact, including: The risk index is compared with a dynamically adjusted safety threshold matrix; When the risk index meets the preset threshold conditions, a tiered early warning is triggered and an emergency response plan is pushed out. When the risk index satisfy When a Level 1 warning is triggered, A level-two warning is triggered at this time, among which and This is a preset threshold parameter.

[0014] Based on the same inventive concept, the present invention also proposes a computer storage medium on which a computer program is stored, which, when executed by a processor, implements the above-described method.

[0015] Compared with the prior art, the present invention has the following beneficial effects: This invention parses inspection data to obtain equipment entities and environmental element entities, and constructs an inspection environment knowledge graph based on these entities. This allows for the association between equipment information and environmental information in the inspection data, thereby addressing the lack of systematic modeling of the relationships between equipment entities and environmental elements in existing technologies. Compared to existing technologies, this invention establishes a knowledge representation foundation oriented towards entity relationships within the inspection environment, thus providing support for subsequent association reasoning and risk analysis.

[0016] This invention extracts real-time topology data based on an inspection environment knowledge graph and aligns the inspection data with the real-time topology data spatiotemporally, thus establishing a correspondence between the inspection data and the entity relationship states in the inspection environment. This addresses the problem of insufficient spatiotemporal correspondence processing between inspection data and environmental relationships in existing technologies. Compared to existing technologies, this invention improves the matching degree between inspection data and environmental relationship information, thereby providing a more accurate data foundation for identifying abnormal propagation paths.

[0017] This invention utilizes spatiotemporally aligned topological data for correlation reasoning, applies graph attention networks to calculate the correlation degree of environmental elements, and identifies abnormal propagation paths through path reasoning algorithms. This allows the correlation relationships during inspections to be transformed into abnormal propagation analysis results, thereby improving upon existing technologies that struggle to reveal risk propagation processes in complex operational environments. Compared to existing technologies, it enables the identification of abnormal propagation paths, thus enhancing the ability to analyze the overall situation of complex operational environments.

[0018] This invention conducts risk assessment based on anomaly propagation paths and inputs the correlation between environmental elements and the anomaly propagation paths into a dynamic risk assessment model. This further transforms the anomaly propagation analysis results into risk indices, risk levels, and impact ranges, thereby improving upon the lack of continuous reasoning and quantitative assessment of dynamic risks in existing technologies. Compared to existing technologies, it can generate risk assessment results for complex operating environments, thus enhancing the real-time assessment capability of safety risks.

[0019] This invention compares the risk index with a dynamically adjusted safety threshold matrix and triggers tiered early warnings and pushes emergency response plans when the risk index meets preset threshold conditions. This allows risk assessment results to be further used for safety early warning processing, thereby improving the problems of untimely and insufficient safety risk early warning response in existing technologies. Compared with existing technologies, it can trigger corresponding safety early warnings based on risk index, risk level, and scope of impact, thereby improving the real-time nature and targeting of safety monitoring of UAV operating environments.

[0020] This invention has the technical effects of modeling the association between equipment entities and environmental elements in the inspection environment, identifying abnormal propagation paths, assessing risks, and providing safety early warnings. It can improve the real-time monitoring capability of safety risks in complex operating environments and is applicable to the fields of intelligent inspection of drones, safety monitoring, and knowledge graph reasoning. Attached Figure Description

[0021] Figure 1 This is a flowchart of the UAV inspection knowledge graph reasoning safety monitoring method described in the implementation method; Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Implementation Method 1 like Figure 1 As shown, this embodiment proposes a knowledge graph reasoning-based security monitoring method for unmanned aerial vehicle (UAV) inspections, the method comprising: S1. Obtain inspection data of the drone's operating environment; S2. Parse the inspection data to obtain equipment entities and environmental element entities, and construct an inspection environment knowledge graph based on the equipment entities and environmental element entities. S3. Extract real-time topology data based on the inspection environment knowledge graph, and perform spatiotemporal alignment between the inspection data and the real-time topology data to obtain spatiotemporally aligned topology data. S4. Based on the spatiotemporally aligned topological data, perform correlation reasoning to obtain the anomaly propagation path; S5. Based on the abnormal propagation path, conduct a risk assessment to obtain the risk index, risk level, and scope of impact. S6. Trigger a safety warning based on the risk index, risk level, and scope of impact.

[0023] By acquiring inspection data of the UAV operating environment and further completing the construction of an inspection environment knowledge graph, real-time topology data extraction and spatiotemporal alignment, anomaly propagation path identification, risk assessment and safety early warning processing, a continuous technical processing link for the UAV operating environment can be formed, thereby realizing the monitoring of safety risks in complex operating environments.

[0024] Furthermore, the inspection data mentioned in S1 is real-time multimodal data collected by the UAV inspection module. The real-time multimodal data includes visible light image data, infrared thermal imaging data, temperature and humidity data, air pressure data, gas concentration data, and positioning and navigation data.

[0025] By limiting the inspection data to real-time multimodal data collected by the UAV inspection module, image information, environmental parameter information, and positioning and navigation information in the working environment can be acquired in a unified manner, thereby providing a data foundation for subsequent inspection environment knowledge graph construction and association reasoning.

[0026] Preferably, the UAV inspection module can be configured with a multispectral image acquisition unit, an environmental sensor array, and a positioning and navigation unit. The multispectral image acquisition unit is used to acquire visible light image data and infrared thermal imaging data, the environmental sensor array is used to acquire temperature and humidity data, air pressure data, and gas concentration data, and the positioning and navigation unit is used to acquire positioning and navigation data.

[0027] Furthermore, S1, acquiring inspection data of the UAV operating environment includes collecting real-time multimodal data of the operating environment through the UAV inspection module, and performing sensor data fusion and noise filtering on the real-time multimodal data to obtain preprocessed inspection data; wherein, the sensor data fusion and noise filtering includes using a Kalman filter algorithm to perform state estimation and update on the real-time multimodal data.

[0028] By performing sensor data fusion and noise filtering on real-time multimodal data and using the Kalman filter algorithm for state estimation updates, the stability and consistency of inspection data can be improved, thereby providing more reliable data input for subsequent entity extraction, entity topology construction, and spatiotemporal alignment.

[0029] Preferably, the sensor data fusion and noise filtering can perform corresponding data fusion processing and noise suppression processing on image acquisition data, environmental parameter data and positioning and navigation data respectively, so as to improve the coordination between different types of inspection data.

[0030] Furthermore, the step S2, which involves constructing an inspection environment knowledge graph based on the equipment entities and environmental element entities, includes extracting entities from the inspection data to obtain the equipment entities and environmental element entities, constructing an entity topology structure between the equipment entities and environmental element entities based on the spatiotemporal dimension, and storing triplet data of equipment status, environmental parameters, and safety rules based on the entity topology structure.

[0031] By extracting entities from inspection data, we obtain equipment entities and environmental element entities. Furthermore, based on the spatiotemporal dimension, we construct an entity topology structure and store triple data of device status, environmental parameters, and safety rules. This enables the association and expression of equipment information and environmental information in the inspection environment, thereby improving the organization and expression capabilities of entity relationships in the work environment.

[0032] Preferably, the equipment entity can be used to represent the equipment status information in the inspection object, the environmental element entity can be used to represent the environmental parameter information around the inspection object, and the entity topology is used to represent the association relationship between the equipment entity and the environmental element entity.

[0033] Furthermore, S2 describes constructing the entity topology structure between the device entity and the environmental element entity based on the spatiotemporal dimension, including defining the device entity, environmental entity, and their relationships through the ontology modeling unit, and completing the entity relationship construction through the dynamic relationship extraction unit and the semantic fusion unit.

[0034] By defining equipment entities, environmental entities, and their relationships through ontology modeling units, and combining them with dynamic relationship extraction units and semantic fusion units to complete entity relationship construction, the entity relationship expression in the inspection environment can be made clearer, thereby improving the structuring degree of the inspection environment knowledge graph and its applicability for subsequent reasoning.

[0035] Preferably, the ontology modeling unit, dynamic relation extraction unit, and semantic fusion unit can serve as functional components in the knowledge graph construction process, used to respectively complete entity definition, relation extraction, and relation fusion processing.

[0036] Furthermore, the real-time topology data mentioned in S3 is data extracted based on the entity topology structure between equipment entities and environmental element entities in the inspection environment knowledge graph.

[0037] By limiting real-time topology data to data extracted from the entity topology structure between equipment entities and environmental element entities in the inspection environment knowledge graph, subsequent spatiotemporal alignment processing can be based on entity relationship data, thereby improving the correspondence between inspection data and the relationship status of the inspection environment.

[0038] Preferably, the real-time topology data can be used to characterize the relationship between equipment entities and environmental element entities during the inspection.

[0039] Furthermore, S4 describes performing correlation reasoning based on the spatiotemporally aligned topological data to obtain anomaly propagation paths, including applying graph attention network to calculate the correlation degree of environmental elements; and identifying anomaly propagation paths through path reasoning algorithms.

[0040] By performing correlation reasoning on spatiotemporally aligned topological data, applying graph attention networks to calculate the correlation degree of environmental elements, and identifying abnormal propagation paths through path reasoning algorithms, the correlation relationships in the inspection environment can be transformed into abnormal propagation analysis results, thereby improving the ability to identify risk propagation processes in complex operational environments.

[0041] Preferably, the graph attention network is used to calculate the weights of the relationships in the spatiotemporally aligned topological data, and the path reasoning algorithm is used to identify abnormal propagation paths based on the relationships.

[0042] Furthermore, S5 describes conducting a risk assessment based on the abnormal propagation path to obtain a risk index, risk level, and impact range. This includes inputting the correlation between environmental elements and the abnormal propagation path into a dynamic risk assessment model, and obtaining the risk index through multi-factor linear weighted assessment.

[0043] By inputting the correlation between environmental elements and the abnormal propagation path into the dynamic risk assessment model, and obtaining the risk index through multi-factor linear weighted assessment, the risk level and scope of impact can be further determined, thereby improving the ability to quantitatively assess the environmental safety risks of inspections.

[0044] Preferably, the dynamic risk assessment model is used to comprehensively analyze the risk status corresponding to the abnormal propagation path, and the multi-factor linear weighted assessment is used to jointly assess the correlation between environmental elements and the abnormal propagation path.

[0045] Furthermore, S6 triggers a security alert based on the aforementioned risk index, risk level, and scope of impact, including: The risk index is compared with a dynamically adjusted safety threshold matrix; When the risk index meets the preset threshold conditions, a tiered early warning is triggered and an emergency response plan is pushed out. When the risk index satisfy When a Level 1 warning is triggered, A level-two warning is triggered at this time, among which and This is a preset threshold parameter.

[0046] By comparing the risk index with a dynamically adjusted safety threshold matrix, and triggering tiered early warnings and pushing emergency response plans when preset threshold conditions are met, risk assessment results can be further transformed into safety early warning results, thereby improving the real-time nature and pertinence of safety monitoring of the UAV operating environment.

[0047] Preferably, the dynamically adjusted safety threshold matrix can be used to adapt to the risk assessment needs under different inspection environments, and the graded early warning is used to distinguish the early warning level under different risk conditions.

[0048] The method described in this embodiment can be executed by a processor calling a computer program, which can be stored in a computer storage medium. When the computer program is executed by the processor, the above-mentioned UAV inspection knowledge graph reasoning security monitoring method can be implemented.

[0049] Implementation Method 2 The UAV inspection knowledge graph reasoning safety monitoring method of this embodiment can be implemented based on a UAV inspection unit, a knowledge graph reasoning engine, and a safety monitoring and control platform. Data interaction between the modules is achieved through a wireless communication network. The system architecture includes a UAV inspection data acquisition module, a knowledge graph construction module, a safety monitoring and reasoning module, a control center module, and a user interaction module to realize inspection data acquisition, knowledge graph construction, correlation reasoning, risk assessment, and safety early warning in the UAV operating environment.

[0050] Furthermore, the UAV inspection data acquisition module includes a multispectral image acquisition unit, an environmental sensor array, a positioning and navigation unit, and a data preprocessing unit. The multispectral image acquisition unit is equipped with a visible light camera and an infrared thermal imager; the environmental sensor array integrates temperature, humidity, air pressure, and gas concentration sensors; the positioning and navigation unit is equipped with an RTK-GNSS positioning system and obstacle avoidance radar; and the data preprocessing unit performs sensor data fusion and noise filtering. The data preprocessing unit uses a Kalman filter algorithm.

[0051] In the formula, The state estimate at time k. Here is the state transition matrix. For Kalman gain, For the observed values, This is the observation matrix.

[0052] The above structure allows for the acquisition of real-time multimodal data of the UAV's operating environment, forming inspection data for subsequent analysis and processing.

[0053] Furthermore, the knowledge graph construction module includes an ontology modeling unit, a relation extraction unit, and a graph storage submodule. The ontology modeling unit defines device entities, environment entities, and their relationships; the relation extraction unit uses the TransE algorithm to establish entity relationship embeddings.

[0054] In the formula, , , These represent the vector representations of the head entity, relation, and tail entity, respectively; the graph storage submodule adopts a dual-storage architecture of Neo4j graph database and Elasticsearch.

[0055] Furthermore, the process of constructing the inspection environment knowledge graph includes defining equipment entities.

[0056] Environmental element entities

[0057] and its relation set

[0058] The above methods can be used to express the entity relationships between equipment entities and environmental element entities, and provide a foundation for subsequent real-time topology data extraction and correlation reasoning.

[0059] Furthermore, the reasoning and analysis module integrates a graph neural network reasoning engine and a temporal evolution analysis unit. The knowledge graph reasoning method adopts a dynamic graph embedding model, and iterative updates of node features are achieved through the following formula:

[0060] in Indicates the first Nodes in a layered network eigenvectors, For nodes The neighborhood group, The edge weights are calculated for the attention mechanism. For a trainable parameter matrix, This represents a non-linear activation function.

[0061] Furthermore, the reasoning and analysis process includes constructing a knowledge graph of the inspection environment and defining equipment entities.

[0062] Environmental element entities

[0063] and its relation set

[0064] Design a dual-channel feature fusion mechanism to integrate real-time data collected by the UAV.

[0065] With knowledge graph features

[0066] Spatiotemporal alignment is performed; and an improved graph attention network is applied to calculate risk propagation weights:

[0067] in For the attention mechanism parameter vector, The characteristic transformation matrix, This indicates a vector concatenation operation.

[0068] Using the above methods, the correlation degree of environmental elements can be calculated, and abnormal propagation paths can be identified through path reasoning algorithms, outputting an analysis report that includes risk level and impact range.

[0069] Furthermore, the safety monitoring inference module includes: a rule-based inference unit with built-in IF-THEN type multi-factor expert rules; and a dynamic risk assessment unit that performs multi-factor linear weighted assessment.

[0070] In the formula, For the comprehensive risk value, For the first i The dynamic weight of each indicator Standardized scores for indicators; real-time early warning unit, triggering a three-level alarm threshold mechanism; adaptive optimization unit, using gradient descent to update inference parameters:

[0071] In the formula, For model parameters, For learning rate, This is the loss function.

[0072] Furthermore, a dynamic risk assessment model can be established to output a risk index:

[0073] in This is the set of model parameters.

[0074] The above methods can be used to further transform abnormal propagation paths and the correlation between environmental factors into risk assessment results, and provide a basis for subsequent safety early warning.

[0075] Furthermore, the control center module includes a task scheduling unit, a communication management unit, and an anomaly handling unit. The task scheduling unit dynamically allocates drone inspection paths, the communication management unit supports 5G and LoRa dual-mode transmission, and the anomaly handling unit provides a fault diagnosis code library. The user interaction module includes a 3D visualization unit, a motion alarm unit, a parameter configuration unit, and a historical query unit. The 3D visualization unit renders the device status topology map, the motion alarm unit pushes tiered early warning information, the parameter configuration unit sets risk assessment thresholds, and the historical query unit stores inspection data from the past three years. Through this structure, inspection path scheduling, communication management, fault handling, early warning display, and historical data querying can be achieved.

[0076] Furthermore, during system implementation, the UAV inspection module uploads environmental monitoring data to the knowledge graph construction module every 30 seconds. The knowledge graph reasoning module performs a full graph traversal every 5 minutes, detecting abnormal equipment status and environmental risks. When the risk value R exceeds the threshold, the real-time early warning unit triggers an audible and visual alarm and generates a response plan. The control center module automatically adjusts the UAV inspection frequency according to the early warning level, up to a maximum of once per second. Through this operational mode, continuous execution of data collection, knowledge graph construction, reasoning analysis, risk assessment, and safety early warning in the UAV operating environment can be achieved.

[0077] The above detailed description of the technical solution provided by the present invention is intended to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above detailed embodiments are not intended to limit the scope of protection of the present invention. Any reasonable modifications and improvements to the present invention, recombination of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0078] Those skilled in the art will understand that the above description is merely a preferred embodiment of the present invention, and the features described in the various embodiments and / or claims disclosed in the present invention can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle scope of the present invention should be considered to fall within the protection scope of the present invention.

[0079] 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 both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A method for safety monitoring using knowledge graph reasoning in unmanned aerial vehicle (UAV) inspections, characterized in that, The method includes: S1. Obtain inspection data of the drone's operating environment; S2. Parse the inspection data to obtain equipment entities and environmental element entities, and construct an inspection environment knowledge graph based on the equipment entities and environmental element entities. S3. Extract real-time topology data based on the inspection environment knowledge graph, and perform spatiotemporal alignment between the inspection data and the real-time topology data to obtain spatiotemporally aligned topology data. S4. Based on the spatiotemporally aligned topological data, perform correlation reasoning to obtain the anomaly propagation path; S5. Based on the abnormal propagation path, conduct a risk assessment to obtain the risk index, risk level, and scope of impact. S6. Trigger a safety warning based on the risk index, risk level, and scope of impact.

2. The method according to claim 1, characterized in that, The inspection data mentioned in S1 is real-time multimodal data collected by the UAV inspection module. The real-time multimodal data includes visible light image data, infrared thermal imaging data, temperature and humidity data, air pressure data, gas concentration data, and positioning and navigation data.

3. The method according to claim 1, characterized in that, S1 describes acquiring inspection data of the working environment, which includes collecting real-time multimodal data of the working environment through a drone inspection module, and performing sensor data fusion and noise filtering on the real-time multimodal data to obtain preprocessed inspection data. The sensor data fusion and noise filtering includes using a Kalman filter algorithm to perform state estimation and update on the real-time multimodal data.

4. The method according to claim 1, characterized in that, S2 constructs an inspection environment knowledge graph based on the equipment entities and environmental element entities, including extracting entities from the inspection data to obtain the equipment entities and environmental element entities, constructing an entity topology structure between the equipment entities and environmental element entities based on the spatiotemporal dimension, and storing triplet data of equipment status, environmental parameters and safety rules based on the entity topology structure.

5. The method according to claim 4, characterized in that, The construction of the entity topology structure between the device entity and the environmental element entity based on the spatiotemporal dimension includes defining the device entity, the environmental entity and their relationships through the ontology modeling unit, and completing the entity relationship construction through the dynamic relationship extraction unit and the semantic fusion unit.

6. The method according to claim 1, characterized in that, The real-time topology data mentioned in S3 is data extracted based on the entity topology structure between equipment entities and environmental element entities in the inspection environment knowledge graph.

7. The method according to claim 1, characterized in that, S4 describes performing correlation reasoning based on the spatiotemporally aligned topological data to obtain anomaly propagation paths, including applying graph attention networks to calculate the correlation degree of environmental elements; and identifying anomaly propagation paths through path reasoning algorithms.

8. The method according to claim 1, characterized in that, S5 describes risk assessment based on the abnormal propagation path to obtain a risk index, risk level, and impact range. This includes inputting the correlation between environmental elements and the abnormal propagation path into a dynamic risk assessment model, and obtaining the risk index through multi-factor linear weighted assessment.

9. The method according to claim 1, characterized in that, S6 triggers a security alert based on the aforementioned risk index, risk level, and scope of impact, including: The risk index is compared with a dynamically adjusted safety threshold matrix; When the risk index meets the preset threshold conditions, a tiered early warning is triggered and an emergency response plan is pushed out. When the risk index satisfy When a Level 1 warning is triggered, A level-two warning is triggered at this time, among which and This is a preset threshold parameter.

10. A computer storage medium having a computer program stored thereon, characterized in that, The computer program, when run on a processor, performs the method described in any one of claims 1-9.

Images