A multi-scene-oriented security-oriented digital twin personalization construction method

Abstract

The application discloses a multi-scene-oriented safety-oriented digital twin personalization construction method, relates to the technical field of digital twin, and is used for solving the main technical problems of poor scene adaptability, static lagging rules, high new scene construction cost, lack of dynamic iteration and closed-loop verification in the prior art; the method comprises the following steps: acquiring multi-scene standardized data; constructing a risk feature model, mining real-time risk data and historical feature data, extracting core risk features, and completing model incremental updating; completing cross-scene migration, threshold correction and self-adaptive adaptation of safety rules for medium-high similarity scenes; fusing the standardized data, the iterated risk feature model and the adaptive rules to construct a personalized safety digital twin model; and obtaining the overall matching degree of the model and the physical scene through multi-dimensional weighting, and outputting the model and feeding back iteration if the overall matching degree meets the standard, or reversely reconstructing and optimizing if the overall matching degree does not meet the standard.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and more specifically, to a security-oriented personalized digital twin construction method for multiple scenarios. Background Technology

[0002] Traditional safety protection solutions are mostly based on static rules for single scenarios, relying on fixed equipment failure thresholds and preset personnel density warning lines to achieve protection. However, the safety risk characteristics vary significantly across different scenarios: the core risks in university laboratories are reagent leaks and operational errors; in rail transit stations, the core risks are equipment failures and stampedes; and in industrial workshops, the core risks are high-temperature equipment malfunctions and violations of regulations. General-purpose safety models cannot account for the personalized risk characteristics of different scenarios, and directly applying existing rules can easily lead to unreasonable warning thresholds, resulting in false alarms or missed alarms.

[0003] Currently, digital twin technology is increasingly widely used in the security field. However, most existing technologies focus on building static security models for single, specific scenarios, such as access control security for smart communities or factory early warning systems for forklift operations. These models can only achieve static geometric mapping of the physical scene and lack the ability to dynamically adapt security rules. Furthermore, the construction of existing models largely relies on expert experience and manual work, resulting in low efficiency and insufficient risk coverage. For new application scenarios, models and security rules often need to be built from scratch, lacking cross-scenario rule reuse mechanisms, leading to a significant waste of human and material resources and delayed security response in new scenarios. In addition, existing technologies generally lack closed-loop verification and iteration mechanisms. Once the model is built, it cannot be automatically updated as the environment changes and risks evolve, causing the security protection model to gradually become disconnected from the actual physical scene. This makes it unable to adapt to dynamic changes such as surges in customer traffic, equipment upgrades, and sudden risk mutations. The ability to extract features and update models for sudden risks is insufficient, making it difficult to meet the dynamic and personalized security protection needs of multiple scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a security-oriented personalized digital twin construction method for multiple scenarios, addressing the main technical problems of existing technologies, such as poor scenario adaptability, static and outdated rules, high construction costs for new scenarios, and lack of dynamic iteration and closed-loop verification. In view of this, this invention achieves this through the following solution.

[0005] In a first aspect, the present invention provides a security-oriented personalized digital twin construction method for multiple scenarios, including: The system acquires multi-dimensional data on environmental, equipment, personnel, and scene attributes through a data acquisition terminal. After cleaning, noise reduction, and normalization, the data format and units are unified, and standardized data for multiple scenes are output. A risk feature model is built using a dynamic security feature iteration engine. A density clustering algorithm is used to mine real-time risk data and historical feature data, extract core risk features, and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification, and model optimization. Based on cosine similarity, the feature matching degree between the source scene and the target scene is obtained, and the cross-scene migration, threshold correction and adaptive adaptation of security rules are completed for medium and high similarity scenes. By integrating standardized data, iterated risk characteristic models, and adaptation rules, a personalized security digital twin model can be constructed. The overall matching degree between the model and the physical scene is obtained by multi-dimensional weighting. If the standard is met, the model is output and feedback is provided for iteration; if the standard is not met, the model is reconstructed and optimized in reverse.

[0006] Furthermore, in the safety-oriented digital twin personalized construction method of the present invention for multiple scenarios, the multi-dimensional data includes environmental parameters, equipment status, personnel behavior, and scenario attributes; the environmental parameters include temperature, humidity, and hazardous chemical leakage concentration; the equipment status includes operating parameters and fault information; the personnel behavior includes location and movement trajectory; and the scenario attributes include scenario type, area division, and hazardous chemical storage location.

[0007] Furthermore, in the multi-scenario security-oriented personalized digital twin construction method of the present invention, the dynamic security feature iteration engine has a built-in DBSCAN density clustering risk feature model, and the construction and updating process of this model includes: Define the core parameters: neighborhood radius and minimum density points; The real-time risk dataset is traversed, and core points, boundary points, and noise points are determined using the density calculation formula: [Formula omitted for brevity]. ,in, For sample points neighborhood density, For risk feature datasets, It is a second-order Euclidean distance. For indicator functions; Clustering based on density reachability relationships merges similar risk features, removes abnormal noise data, and outputs an updated set of core risk features for the scenario.

[0008] Furthermore, in the multi-scenario safety-oriented personalized digital twin construction method of the present invention, the value range of the neighborhood radius is 0.5~1.5, and the value range of the minimum density point number is 5~15.

[0009] Furthermore, in the multi-scenario security-oriented personalized digital twin construction method of the present invention, the cosine similarity is expressed as: ; in, Representing the source scene With the target scenario Feature similarity, , , i, represent the standardized risk feature values ​​of the i-th dimension of the source scenario and the target scenario, respectively, where n is the total number of feature dimensions.

[0010] Furthermore, in the security-oriented personalized digital twin construction method of the present invention for multiple scenarios, by setting a similarity threshold range, when the similarity is ≥0.6, it is determined to be a medium-to-high matching scenario, and when the similarity is ≥0.8, it is determined to be a highly homogeneous scenario. Security rule migration and differential adaptation are automatically executed. If the adaptation fails, the rule parameters are automatically backtracked and iterated to adjust.

[0011] Furthermore, in the multi-scenario safety-oriented personalized digital twin construction method of the present invention, the overall matching degree between the model and the physical scene is expressed as follows during the process of obtaining the overall matching degree between the model and the physical scene through multi-dimensional weighted calculation: ; in, Indicates scene feature matching degree, Indicates dynamic response matching degree, Indicates the degree of matching between the effectiveness and compliance of security protection measures; For the corresponding weight coefficients, satisfying , , ,and .

[0012] Furthermore, in the multi-scenario security-oriented personalized digital twin construction method of the present invention, by setting the comprehensive matching degree qualified threshold range to 0.80~0.90 and the verification threshold to 0.85, when the overall matching degree Score is greater than or equal to the verification threshold, the digital twin model is determined to be compatible with the physical scene; if it is lower than the verification threshold, it is determined to be an abnormal fit and needs to be reconstructed and optimized.

[0013] Furthermore, in the multi-scenario security-oriented personalized digital twin construction method of the present invention, the scene feature matching degree in the multi-dimensional weighting... The dynamic response matching degree is obtained by comparing the cosine similarity between the risk distribution output by the digital twin model and the risk distribution collected by sensors in the actual physical scenario. The effectiveness matching degree of security protection is determined by the ratio of the system response time after simulating a sudden risk event to the preset standard response time. It is obtained by the harmonic mean of the actual early warning accuracy rate and recall rate.

[0014] Secondly, this invention provides a security-oriented personalized digital twin construction system for multiple scenarios, including: The preprocessing module is used to acquire multi-dimensional data on environment, equipment, personnel, and scene attributes through the acquisition terminal, and after cleaning, noise reduction, and normalization, the data format and units are unified, and standardized data for multiple scenes are output. The dynamic security feature iteration engine is used to build risk feature models. It uses density clustering algorithm to mine real-time risk data and historical feature data, extract core risk features and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification and model optimization. The rule transfer engine is used to obtain the feature matching degree between the source scene and the target scene based on cosine similarity, and to complete the cross-scene transfer, threshold correction and adaptive adaptation of security rules for medium and high similarity scenes. The scenario building module is used to integrate standardized data, iterated risk feature models, and adaptation rules to build a personalized security digital twin model. The matching degree verification module is used to obtain the overall matching degree between the model and the physical scene through multi-dimensional weighting. If the matching degree meets the standard, the model is output and feedback is fed back for iteration; if the matching degree does not meet the standard, the model is reconstructed and optimized in reverse.

[0015] Compared with existing technologies, the present invention's safety-oriented personalized digital twin construction method for multiple scenarios acquires multi-dimensional data such as environment, equipment, personnel, and scenario attributes through a data acquisition terminal. This data is then cleaned, denoised, and normalized to output standardized data with a unified format and dimensions. This method is compatible with heterogeneous data input from various differentiated scenarios such as industrial workshops, rail transit stations, university laboratories, and chemical industrial parks, providing a standardized data foundation for subsequent feature iteration and rule migration. Based on this, a DBSCAN density clustering risk feature model is constructed using a dynamic safety feature iteration engine. This model performs density clustering on real-time risk data using neighborhood radius and minimum density points, automatically identifying core points, boundary points, and noise points. It extracts core risk features and completes incremental model updates, forming a dynamic closed-loop iterative mechanism for risk acquisition, feature updating, scenario verification, and model optimization. This overcomes the deficiency of static rules not being able to automatically adjust with dynamic changes in scenarios, significantly improving the real-time response capability to sudden risks. Simultaneously, by constructing a cross-scenario safety rule migration engine based on cosine similarity, the risk feature matching degree between the source and target scenarios is quantified, and medium-to-high matching scenarios with similarity ≥ 0.6 are automatically migrated. This invention transfers verified security rules and performs threshold correction and adaptive adaptation based on target scenario characteristics, eliminating the need to build new scenario rules from scratch. This significantly shortens the model building cycle for new scenarios and reduces manpower and time costs. Furthermore, this invention establishes a multi-dimensional weighted matching degree evaluation model, integrating three dimensions: scenario feature matching degree (A), dynamic response matching degree (B), and security protection effectiveness matching degree (C). It quantifies the overall adaptability of the digital twin model to the physical real scenario through weighted quantification, setting a passing threshold of 0.80~0.90. If the threshold is met, the model is output and feedback is iterated; if not, reverse reconstruction and optimization are performed, forming a closed-loop optimization system that ensures the digital twin model and security rules continuously adapt to dynamic changes in the scenario. Through the above technical solutions of this invention, the main technical problems of existing technologies—poor scenario adaptability, static rule lag, high cost of building new scenarios, and lack of dynamic iteration and closed-loop verification—are solved. Attached Figure Description

[0016] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0017] Figure 1 This is a flowchart illustrating the security-oriented personalized digital twin construction method for multiple scenarios according to the present invention. Detailed Implementation

[0018] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0019] It should be noted that when a component is referred to as being "fixed to" or "set on" another component, it can be directly on or indirectly on that other component. When a component is referred to as being "connected to" another component, it can be directly connected to or indirectly connected to that other component.

[0020] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. "Several" means one or more, unless otherwise explicitly specified.

[0021] Currently, digital twin technology is increasingly widely used in the security field. However, most existing technologies focus on building static security models for single, specific scenarios, such as access control security for smart communities or factory early warning systems for forklift operations. These models can only achieve static geometric mapping of the physical scene and lack the ability to dynamically adapt security rules. Furthermore, the construction of existing models largely relies on expert experience and manual work, resulting in low efficiency and insufficient risk coverage. For new application scenarios, models and security rules often need to be built from scratch, lacking cross-scenario rule reuse mechanisms, leading to a significant waste of human and material resources and delayed security response in new scenarios. In addition, existing technologies generally lack closed-loop verification and iteration mechanisms. Once the model is built, it cannot be automatically updated as the environment changes and risks evolve, causing the security protection model to gradually become disconnected from the actual physical scene. This makes it unable to adapt to dynamic changes such as surges in customer traffic, equipment upgrades, and sudden risk mutations. The ability to extract features and update models for sudden risks is insufficient, making it difficult to meet the dynamic and personalized security protection needs of multiple scenarios.

[0022] This invention provides a security-oriented personalized digital twin construction method for multiple scenarios, including: The system acquires multi-dimensional data on environmental, equipment, personnel, and scene attributes through a data acquisition terminal. After cleaning, noise reduction, and normalization, the data format and units are unified, and standardized data for multiple scenes are output. A risk feature model is built using a dynamic security feature iteration engine. A density clustering algorithm is used to mine real-time risk data and historical feature data, extract core risk features, and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification, and model optimization. Based on cosine similarity, the feature matching degree between the source scene and the target scene is obtained, and the cross-scene migration, threshold correction and adaptive adaptation of security rules are completed for medium and high similarity scenes. By integrating standardized data, iterated risk characteristic models, and adaptation rules, a personalized security digital twin model can be constructed. The overall matching degree between the model and the physical scene is obtained by multi-dimensional weighting. If the standard is met, the model is output and feedback is provided for iteration; if the standard is not met, the model is reconstructed and optimized in reverse.

[0023] With the above technical solution, the safety-oriented digital twin personalized construction method of the present invention, which is oriented towards multiple scenarios, acquires multi-dimensional data such as environment, equipment, personnel, and scenario attributes through a data acquisition terminal. After cleaning, denoising, and normalization, it outputs standardized data with a unified format and unit, which can be compatible with heterogeneous data input from various differentiated scenarios such as industrial workshops, rail transit stations, university laboratories, and chemical industrial parks, providing a standardized data foundation for subsequent feature iteration and rule migration. On this basis, a DBSCAN density clustering risk feature model is constructed using a dynamic safety feature iteration engine. The model performs density clustering on real-time risk data through neighborhood radius and minimum density point number, automatically identifies core points, boundary points, and noise points, extracts core risk features, and completes incremental model updates. This forms a dynamic closed-loop iteration mechanism of risk acquisition, feature update, scenario verification, and model optimization, overcoming the defect that static rules cannot be automatically adjusted with dynamic changes in scenarios, and significantly improving the real-time response capability to sudden risks. At the same time, by constructing a cross-scenario safety rule migration engine based on cosine similarity, the risk feature matching degree between the source scenario and the target scenario is quantified, and medium-to-high matching scenarios with similarity ≥ 0.6 are automatically migrated. This invention transfers verified security rules and performs threshold correction and adaptive adaptation based on target scenario characteristics, eliminating the need to build new scenario rules from scratch. This significantly shortens the model building cycle for new scenarios and reduces manpower and time costs. Furthermore, this invention establishes a multi-dimensional weighted matching degree evaluation model, integrating three dimensions: scenario feature matching degree (A), dynamic response matching degree (B), and security protection effectiveness matching degree (C). It quantifies the overall adaptability of the digital twin model to the physical real scenario through weighted quantification, setting a passing threshold of 0.80~0.90. If the threshold is met, the model is output and feedback is iterated; if not, reverse reconstruction and optimization are performed, forming a closed-loop optimization system that ensures the digital twin model and security rules continuously adapt to dynamic changes in the scenario. Through the above technical solutions of this invention, the main technical problems of existing technologies—poor scenario adaptability, static rule lag, high cost of building new scenarios, and lack of dynamic iteration and closed-loop verification—are solved.

[0024] To better understand the present invention, the following specific embodiments further illustrate the content of the present invention, but the content of the present invention is not limited to the following embodiments.

[0025] Example 1 Please see Figure 1 This embodiment provides a security-oriented personalized digital twin construction method for multiple scenarios, including: Step 1: Acquire multi-dimensional data on environment, equipment, personnel, and scene attributes through the data acquisition terminal. After cleaning, noise reduction, and normalization, the data format and units are unified, and multi-scene standardized data is output. Step 2: Utilize the dynamic security feature iteration engine to construct a risk feature model, and employ density clustering algorithm to mine real-time risk data and historical feature data, extract core risk features, and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification, and model optimization. Step 3: Obtain the feature matching degree between the source scene and the target scene based on cosine similarity, and complete the cross-scene migration, threshold correction and adaptive adaptation of security rules for medium and high similarity scenes; Step 4: Integrate standardized data, iterated risk characteristic models, and adaptation rules to construct a personalized security digital twin model; Step 5: Obtain the overall matching degree between the model and the physical scene through multi-dimensional weighting. If the standard is met, output the model and provide feedback for iteration; if the standard is not met, reverse the reconstruction and optimization.

[0026] Example 2 Firstly, this embodiment provides a security-oriented personalized digital twin construction method for multiple scenarios, including: Step 1: Acquire multi-dimensional data on environment, equipment, personnel, and scene attributes through the data acquisition terminal. After cleaning, noise reduction, and normalization, the data format and units are unified, and multi-scene standardized data is output. Furthermore, the multidimensional data includes environmental parameters, equipment status, personnel behavior, and scene attributes; environmental parameters include temperature, humidity, and hazardous chemical leakage concentration; equipment status includes operating parameters and fault information; personnel behavior includes location and movement trajectory; and scene attributes include scene type, area division, and hazardous chemical storage location.

[0027] Step 2: Utilize the dynamic security feature iteration engine to construct a risk feature model, and employ density clustering algorithm to mine real-time risk data and historical feature data, extract core risk features, and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification, and model optimization. Furthermore, the dynamic security feature iteration engine incorporates the DBSCAN density clustering risk feature model, the construction and updating process of which includes: Define the core parameters: neighborhood radius and minimum density points; the neighborhood radius ranges from 0.5 to 1.5, and the minimum density points range from 5 to 15; the neighborhood radius and minimum density points can be determined using the k-distance graph method or grid search method according to the risk data distribution of the specific scenario. The parameter values ​​given in this embodiment are only examples. The real-time risk dataset is traversed, and core points, boundary points, and noise points are determined using the density calculation formula: [Formula omitted for brevity]. ,in, For sample points neighborhood density, For risk feature datasets, It is a second-order Euclidean distance. For indicator functions; Clustering based on density reachability relationships merges similar risk features, removes abnormal noise data, and outputs an updated set of core risk features for the scenario.

[0028] Furthermore, the process can be as follows: maintain the core point set and boundary point set for each existing cluster; when new data points... Upon arrival, calculate the minimum Euclidean distance between it and the core points of each cluster. ;like Then If a node is assigned to the corresponding cluster, its core and boundary points are updated; otherwise, the node is reassigned. Marked as temporary points, when the number of temporary points accumulates to the minimum density point number and the distance between them is uniform. When the number of noise points exceeds a threshold, a new cluster is formed; periodically or when the number of noise points exceeds a threshold, global re-clustering is triggered to optimize the model.

[0029] Step 3: Obtain the feature matching degree between the source scene and the target scene based on cosine similarity, and complete the cross-scene migration, threshold correction and adaptive adaptation of security rules for medium and high similarity scenes; Furthermore, cosine similarity is expressed as: ; in, Representing the source scene With the target scenario Feature similarity, , , i, represent the standardized risk feature values ​​of the i-th dimension of the source scenario and the target scenario, respectively, where n is the total number of feature dimensions; By setting a similarity threshold range, a similarity of ≥0.6 is determined to be a medium-to-high matching scenario, and a similarity of ≥0.8 is determined to be a highly homogeneous scenario. The system automatically performs security rule migration and differential adaptation. If the adaptation fails, the system automatically backtracks and iterates to adjust the rule parameters.

[0030] Step 4: Integrate standardized data, the iterated risk feature model, and adaptation rules to construct a personalized security digital twin model. Specifically, this can be achieved by: establishing a digital twin model of the physical scene based on the scene attributes in the standardized data, including physical objects such as devices, areas, and channels, and their spatial layout; binding the core risk feature set and risk thresholds from the iterated risk feature model to the corresponding device or area attributes as parameters; and embedding the warning triggering conditions and linkage actions from the adapted security rules into the model as logical rules to generate a personalized security digital twin model that can respond in real time.

[0031] Step 5: Obtain the overall matching degree between the model and the physical scene through multi-dimensional weighting. If the standard is met, output the model and provide feedback for iteration; if the standard is not met, reverse the reconstruction and optimization. Furthermore, in the process of obtaining the overall matching degree between the model and the physical scene through multi-dimensional weighting, the overall matching degree is expressed as: ; in, Indicates scene feature matching degree, Indicates dynamic response matching degree, Indicates the degree of matching between the effectiveness and compliance of security protection measures; For the corresponding weight coefficients, satisfying , , ,and ; By setting a comprehensive matching score threshold range of 0.80~0.90 and a verification threshold of 0.85, when the overall matching score is greater than or equal to the verification threshold, the digital twin model is deemed to be adapted to the physical scene; if it is lower than the verification threshold, it is deemed to be an adaptation anomaly, requiring reconstruction and optimization; the scene feature matching score in the multi-dimensional weighted calculation... The dynamic response matching degree is obtained by comparing the cosine similarity between the risk distribution output by the digital twin model and the risk distribution collected by sensors in the actual physical scenario. The effectiveness matching degree of security protection is determined by the ratio of the system response time after simulating a sudden risk event to the preset standard response time. It is obtained by the harmonic mean of the actual warning accuracy and recall rate. Furthermore, the accuracy... Recall rate Where TP is the number of true risk events that were correctly identified and alerted, FP is the number of false alarms, and FN is the number of missed alarms; the harmonic mean is... .

[0032] Secondly, this embodiment provides a security-oriented personalized digital twin construction system for multiple scenarios, including: The preprocessing module is used to acquire multi-dimensional data on environment, equipment, personnel, and scene attributes through the acquisition terminal, and after cleaning, noise reduction, and normalization, the data format and units are unified, and standardized data for multiple scenes are output. The dynamic security feature iteration engine is used to build risk feature models. It uses density clustering algorithm to mine real-time risk data and historical feature data, extract core risk features and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification and model optimization. The rule transfer engine is used to obtain the feature matching degree between the source scene and the target scene based on cosine similarity, and to complete the cross-scene transfer, threshold correction and adaptive adaptation of security rules for medium and high similarity scenes. The scenario building module is used to integrate standardized data, iterated risk feature models, and adaptation rules to build a personalized security digital twin model. The matching degree verification module is used to obtain the overall matching degree between the model and the physical scene through multi-dimensional weighting. If the matching degree meets the standard, the model is output and feedback is fed back for iteration; if the matching degree does not meet the standard, the model is reconstructed and optimized in reverse.

[0033] The output of the preprocessing module is connected to the input of the dynamic security feature iteration engine, the output of the dynamic security feature iteration engine is connected to the input of the rule migration engine, the output of the rule migration engine is connected to the input of the scene construction module, the output of the scene construction module is connected to the input of the matching degree verification module, and the output of the matching degree verification module is fed back to the input of the dynamic security feature iteration engine.

[0034] Example 3 Using the technical solution of Embodiment 2 above, this embodiment addresses the differences in equipment safety protection scenarios between industrial production workshops and urban rail transit stations, completing the cross-scenario migration and adaptation of high-temperature anomaly early warning rules to fit the actual operating conditions of both scenarios. The specific implementation steps are as follows: Step 1: Select a conventional machining industrial workshop and a regular subway station hall as the source and target scenarios, respectively. Data collected from the industrial workshop includes: shell temperature of production equipment such as machine tools and air compressors; ambient temperature and humidity in the workshop; continuous operating time of equipment; distribution trajectory of fixed workers in the workshop; and data on the workshop's closed ventilation conditions. Each data collection sample has 28 dimensions, accumulating 120,000 data points collected throughout the day. Data collected from the subway station includes: shell temperature of electromechanical cabinets, air conditioning units, and exhaust equipment in the station hall; passenger density in the station hall; ventilation system wind speed; passenger distribution during peak hours; and data on semi-open ventilation conditions. The data collection dimensions are consistent with the source scenario. The raw data from both scenarios are input into the preprocessing module. A range standardization algorithm is used to normalize the data, eliminating differences in equipment range, detection accuracy, and dimensions, and outputting a standardized feature dataset with unified dimensions and noise removal.

[0035] Step 2: Considering the industry characteristics of high-temperature failures in electromechanical equipment, fix the optimal parameters of the DBSCAN model: neighborhood radius of 0.5 and minimum density points of 8. Import the historical dataset of high-temperature failures of equipment throughout the year in the industrial workshop, and calculate the density using the formula... The sample density was calculated point by point, and the high-temperature anomaly characteristics of the equipment were analyzed by clustering. Invalid noise data such as diurnal temperature fluctuations and instantaneous temperature fluctuations during equipment start-up and shutdown were removed. The incremental iteration update of the risk feature model was completed in 8 seconds. The core feature of the model is: when the industrial workshop is sealed and has poor heat dissipation, the equipment is running under continuous high load, and the equipment shell temperature is ≥80℃, the probability of equipment bearing wear and circuit aging failures increases significantly, which are the core high-risk features.

[0036] Step 3: Extract six core risk features from the two scenarios: equipment attributes, ventilation conditions, personnel density, fault risk type, equipment operation mode, and environmental interference. Substitute these features into the cosine similarity formula to calculate the scenario similarity Sim=0.82, indicating a scenario with highly similar equipment. Adapt the mature 80℃ high-temperature warning rules from industrial workshops, the core protection logic of triggering warnings after 30 seconds of overheating, and triggering equipment load reduction after 60 seconds of overheating. Adaptively optimize the system based on the actual operating conditions of rail transit stations: while the semi-open ventilation and heat dissipation conditions in the station hall are better than in the workshop, the high personnel density and equipment overheating can easily cause passenger panic and pose a higher risk of electrical fire spread. Therefore, the first-level warning threshold is revised to 75℃, the second-level alarm threshold is retained at 80℃, the equipment shutdown rule is canceled, and it is replaced with a rule that links the ventilation system to full-load startup and the evacuation guidance of personnel in parts of the station hall.

[0037] Step 4: Based on the standardized dataset of rail transit stations, the iterative equipment high-temperature risk characteristic model, and the adapted hierarchical temperature control early warning rules, build a full-element digital twin model of rail transit stations, complete equipment location binding, temperature threshold parameter input, early warning linkage logic configuration, and high-risk area labeling, and generate a personalized safety protection twin model adapted to rail transit scenarios.

[0038] Step 5: Optimal weight configuration By simulating high-temperature overload conditions to test the model's performance, the measured results showed a scenario feature matching degree A=0.88, a dynamic response matching degree B=0.91, and a safety protection effectiveness C=0.89; substituting these values ​​into the comprehensive matching degree formula, the calculated results were... The score was higher than the preset optimal threshold of 0.85, indicating that the model was successfully adapted. Actual testing showed that after adaptation, the false alarm rate for high-temperature warnings at rail transit station equipment decreased from 18.2% in the original general rule to 3.1%, and the dynamic response time was shortened by 42%. Simultaneously, the adaptation data and operational verification results were fed back to the dynamic safety feature iteration engine to complete the closed-loop iterative optimization of the model.

[0039] Example 4 Using the technical solution of Embodiment 2 above, this embodiment addresses the scenarios of small-scale hazardous chemical storage in university teaching laboratories and large-scale hazardous chemical storage in fine chemical industrial parks, completing the cross-scenario migration and adaptation of leakage emergency protection rules to conform to the real industry standards for the protection of two types of hazardous chemicals. The specific implementation steps are as follows: Step 1: Select university chemistry teaching laboratories and hazardous chemical storage areas in small and medium-sized fine chemical industrial parks as the source and target scenarios, respectively. Data collected from university laboratories includes: storage quantities of common hazardous chemicals such as ethanol and acid / base reagents; concentration monitoring data from sealed storage cabinets; laboratory ventilation frequency; number of personnel per experiment; and environmental parameters of local work areas. Data collected from chemical industrial parks includes: storage concentrations of drummed organic solvents and corrosive reagents; operating parameters of the storage area ventilation system; personnel inspection routes in the storage area; zoning of the storage area; and environmental temperature and humidity diffusion parameters, unifying the eight core risk feature collection dimensions. The 350,000 historical monitoring data points were cleaned, outlier removed, and normalized preprocessed to eliminate data discrepancies between small-scale laboratory testing and large-scale industrial park monitoring, outputting a standardized feature dataset.

[0040] Step 2: Configure DBSCAN model parameters based on the characteristics of hazardous chemical leakage and diffusion: neighborhood radius of 0.6 and minimum density points of 10. Import historical data on hazardous chemical leakage and volatilization anomalies from university laboratories over the past three years. Use density clustering formulas to iterate through all samples and automatically extract core risk characteristics: small inventory of hazardous chemicals in the laboratory and leakage diffusion range ≤ 5m. 2 Local ventilation can quickly dilute the concentration. The core threshold for exceeding the standard is 200 ppm. The risk is concentrated on the experimental workbench and surrounding area. Noise data such as instantaneous sensor drift and environmental airflow disturbance are eliminated. The risk feature model incremental iterative update is completed in 12 seconds.

[0041] Step 3: Select eight features: hazardous chemical type, leakage and diffusion characteristics, concentration risk threshold, ventilation conditions, personnel operation mode, risk and hazard level, emergency response requirements, and area enclosure degree. Substitute these features into the cosine similarity formula to calculate the scene similarity Sim=0.78, which meets the medium-to-high source matching condition. Transfer the mature basic emergency rules from the laboratory, such as 200ppm concentration exceedance warning, local ventilation activation, and close-range personnel evacuation. Adapt and optimize the rules to the actual safety specifications of the chemical industrial park: given the large amount of hazardous chemicals in the park, the easy diffusion of leaks throughout the area, and the risk of overlapping flammable and explosive substances, supplement the original rules with 300ppm level two full-area alarm, explosion-proof fan linkage in the storage area, delineation of surrounding isolation zones, park emergency broadcasting, and pre-activation adaptation rules for the fire protection system. Fine-tune the basic warning threshold to 180ppm to achieve early warning and early response.

[0042] Step 4: Integrate standardized monitoring data of the chemical industrial park, the iterated hazardous chemical leakage and diffusion risk model, and hierarchical emergency protection rules to complete the reconstruction of the chemical industrial park twin model. Accurately identify high-risk leakage points, leakage diffusion simulation paths, and emergency evacuation routes in each storage area, embed hierarchical early warning and linkage response logic, and generate a personalized safety protection model adapted to large-scale hazardous chemical storage scenarios.

[0043] Step 5, Multi-dimensional matching degree verification and closed-loop optimization: Adopt the optimal weight configuration uniformly. Tests were conducted under simulated real-world conditions of minute and moderate leaks. The measured results showed a scenario feature matching degree A = 0.86, a dynamic response matching degree B = 0.87, and a safety protection effectiveness C = 0.90. Substituting these values ​​into the formula, the results were calculated... The result was higher than the preset optimal threshold of 0.85, indicating that the model verification was successful. Actual measurements showed that the emergency response time in the industrial park was reduced by 58% and the leakage risk diffusion control range was reduced by 65% ​​after migration and adaptation, fully meeting the safety production standards for chemical industrial parks. Test data and operational results were fed back to the iterative engine for continuous optimization and iteration of the risk characteristic model.

[0044] In the description of the above embodiments, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0045] 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.

Claims

1. A security-oriented personalized digital twin construction method for multiple scenarios, characterized in that, include: The system acquires multi-dimensional data on environmental, equipment, personnel, and scene attributes through a data acquisition terminal. After cleaning, noise reduction, and normalization, the data format and units are unified, and standardized data for multiple scenes are output. A risk feature model is built using a dynamic security feature iteration engine. A density clustering algorithm is used to mine real-time risk data and historical feature data, extract core risk features, and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification, and model optimization. Based on cosine similarity, the feature matching degree between the source scene and the target scene is obtained, and the cross-scene migration, threshold correction and adaptive adaptation of security rules are completed for medium and high similarity scenes. By integrating standardized data, iterated risk characteristic models, and adaptation rules, a personalized security digital twin model can be constructed. The overall matching degree between the model and the physical scene is obtained by multi-dimensional weighting. If the standard is met, the model is output and feedback is provided for iteration; if the standard is not met, the model is reconstructed and optimized in reverse.

2. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 1, characterized in that, The multidimensional data includes environmental parameters, equipment status, personnel behavior, and scene attributes; the environmental parameters include temperature, humidity, and hazardous chemical leakage concentration; the equipment status includes operating parameters and fault information; the personnel behavior includes location and movement trajectory; and the scene attributes include scene type, area division, and hazardous chemical storage location.

3. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 1, characterized in that, The dynamic security feature iteration engine incorporates the DBSCAN density clustering risk feature model. The construction and updating process of this model includes: Define the core parameters: neighborhood radius and minimum density points; The real-time risk dataset is traversed, and core points, boundary points, and noise points are determined using the density calculation formula: [Formula omitted for brevity]. ,in, For sample points neighborhood density, For risk feature datasets, It is a second-order Euclidean distance. For indicator functions; Clustering based on density reachability relationships merges similar risk features, removes abnormal noise data, and outputs an updated set of core risk features for the scenario.

4. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 3, characterized in that, The neighborhood radius ranges from 0.5 to 1.5, and the minimum density number of points ranges from 5 to 15.

5. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 1, characterized in that, The cosine similarity is expressed as: ； in, Representing the source scene With the target scenario Feature similarity, , , i, represent the standardized risk feature values ​​of the i-th dimension of the source scenario and the target scenario, respectively, where n is the total number of feature dimensions.

6. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 5, characterized in that, By setting a similarity threshold range, a similarity of ≥0.6 is determined to be a medium-to-high matching scenario, and a similarity of ≥0.8 is determined to be a highly homogeneous scenario. The system automatically performs security rule migration and differential adaptation. If the adaptation fails, the system automatically backtracks and iterates to adjust the rule parameters.

7. The method for constructing personalized, security-oriented digital twins for multiple scenarios according to claim 1, characterized in that, In the process of obtaining the overall matching degree between the model and the physical scene through multi-dimensional weighting, the overall matching degree is expressed as: ； in, Indicates scene feature matching degree, Indicates dynamic response matching degree, Indicates the degree of matching between the effectiveness and compliance of security protection measures; For the corresponding weight coefficients, satisfying , , ,and .

8. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 7, characterized in that, By setting the overall matching score threshold range to 0.80~0.90 and the verification threshold to 0.85, when the overall matching score is greater than or equal to the verification threshold, the digital twin model is deemed to be adapted to the physical scene; if it is lower than the verification threshold, it is deemed to be an adaptation anomaly and needs to be reconstructed and optimized.

9. The method for constructing a security-oriented personalized digital twin for multiple scenarios according to claim 8, characterized in that, The scene feature matching degree in the multi-dimensional weighting The dynamic response matching degree is obtained by comparing the cosine similarity between the risk distribution output by the digital twin model and the risk distribution collected by sensors in the actual physical scenario. The effectiveness matching degree of security protection is determined by the ratio of the system response time after simulating a sudden risk event to the preset standard response time. It is obtained by the harmonic mean of the actual early warning accuracy rate and recall rate.

10. A security-oriented personalized digital twin construction system for multiple scenarios, characterized in that, include: The preprocessing module is used to acquire multi-dimensional data on environment, equipment, personnel, and scene attributes through the acquisition terminal, and after cleaning, noise reduction, and normalization, the data format and units are unified, and standardized data for multiple scenes are output. The dynamic security feature iteration engine is used to build risk feature models. It uses density clustering algorithm to mine real-time risk data and historical feature data, extract core risk features and complete incremental model updates, forming a dynamic closed loop of risk collection, feature update, scenario verification and model optimization. The rule transfer engine is used to obtain the feature matching degree between the source scene and the target scene based on cosine similarity, and to complete the cross-scene transfer, threshold correction and adaptive adaptation of security rules for medium and high similarity scenes. The scenario building module is used to integrate standardized data, iterated risk feature models, and adaptation rules to build a personalized security digital twin model. The matching degree verification module is used to obtain the overall matching degree between the model and the physical scene through multi-dimensional weighting. If the matching degree is met, the model is output and feedback is provided for iteration; if the matching degree is not met, the model is reconstructed and optimized in reverse.

Images