Operation environment safety analysis and evaluation method and device, computer equipment and medium

Abstract

The invention relates to the technical field of knowledge maps, and discloses a working environment safety analysis and evaluation method and device, computer equipment and a medium, and the method comprises the steps: building a hierarchical knowledge base of working environment elements, building entity association between hierarchies, and generating an initial knowledge map; comparing pre-acquired index data of the working environment elements with the hierarchical knowledge base to identify risk indexes, calculating risk values of the risk indexes at the same time, and constructing a safety risk analysis and evaluation set based on the risk indexes and the corresponding risk values; and constructing a risk knowledge graph for analyzing and evaluating the safety of the working environment based on the safety risk analysis and evaluation set and the initial knowledge graph, and performing safety analysis and evaluation on the working environment based on the risk knowledge graph. The method can be effectively applied to safety analysis and evaluation of working environments in different industries, and a scientific method and an effective means are provided for guaranteeing the safety of the working environments.

Description

Technical Field

[0001] The present invention relates to the technical field of knowledge graphs, and particularly to a method, device, computer device and medium for safety analysis and evaluation of an operating environment. Background Art

[0002] The operating environment is an important factor affecting work safety. Operating environment elements include air, lighting, temperature, humidity, noise, vibration, dust, radiation, toxic and harmful substances, etc. Currently, the analysis and evaluation of operating environment factors are incomplete and inaccurate, the control standards are scattered or even missing, and the safety control measures lack systematicness and pertinence. Production accidents caused by unsafe operating environments occur frequently. Therefore, adopting knowledge graph technology and establishing scientific methods to comprehensively and scientifically evaluate the safety risks of the operating environment and ensure that the production operating environment complies with laws and regulations are of great significance for improving work safety levels.

[0003] When using knowledge graph technology to carry out safety analysis and evaluation of the operating environment, technical problems such as rigid knowledge graph architecture and inaccurate risk index judgment are faced. Summary of the Invention

[0004] In view of this, the present invention provides a method, device, computer device and medium for safety analysis and evaluation of an operating environment to solve the technical problems of rigid knowledge graph architecture and inaccurate risk index judgment in the operating environment scenario, and provides effective methods and means for carrying out safety analysis and evaluation of the operating environment.

[0005] In a first aspect, the present invention provides a method for safety analysis and evaluation of an operating environment, the method comprising: Establish a hierarchical knowledge base of operating environment elements, establish entity associations between its levels, and generate an initial knowledge graph; Compare the index data of the operating environment elements obtained in advance with the hierarchical knowledge base of the operating environment elements to identify risk indicators, calculate the risk values of the risk indicators at the same time, and construct a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values; Construct a risk knowledge graph for analyzing and evaluating the safety of the operating environment based on the safety risk analysis and evaluation set and the initial knowledge graph, and perform safety analysis and evaluation of the operating environment based on the risk knowledge graph.

[0006] A method for safety analysis and evaluation of an operating environment provided by the present invention constructs a multi-level knowledge base, integrates multi-source data, compares the index data of the operating environment elements obtained in advance with the hierarchical knowledge base to accurately identify risk indicators, calculates the risk values of the risk indicators, and finally forms a risk knowledge graph that can integrate information from multiple aspects. It can be effectively applied to the safety evaluation of operating environments in different industries, improving the accuracy and universality of safety evaluation, and solving the problems of rigid knowledge graph architecture and inaccurate judgment of risk indicators in the operating environment scenario.

[0007] In an alternative embodiment, a hierarchical knowledge base of operating environment elements is established, and entity associations are established between its levels to generate an initial knowledge graph, including: Obtain the index data, operating standards, influencing factors on the operating environment elements, and operating accident cases of the operating environment elements; Establish an index data layer based on the index data of the operating environment elements, an operating standard layer based on the operating standards of the operating environment elements, an impact analysis layer based on the influencing factors on the operating environment elements, and an accident case layer based on the operating accident cases; Screen the data and entity relationship networks in the index data layer, operating standard layer, impact analysis layer, and accident case layer, store the data in a relational database, store the entity relationship network in a graph database, and generate a hierarchical knowledge base based on the stored relational database and the stored graph database; Calculate the semantic similarity of cross-level entities in the hierarchical knowledge base through a pre-trained model, and perform cross-level association on entities with semantic similarity exceeding a preset threshold to generate an initial knowledge graph.

[0008] A method for safety analysis and evaluation of an operating environment provided by the present invention systematically classifies and structurally stores multi-source heterogeneous data by clarifying the functions and stored contents of different levels. Calculate the semantic similarity between entities at each level through a pre-trained model, associate entities with similarity greater than the similarity threshold, optimize the logical level of knowledge representation, significantly improve the accuracy and response speed of data retrieval, and can more comprehensively and accurately evaluate the safety of the operating environment, effectively solving the problem of weak semantic association between levels. Each level is interconnected to form a comprehensive and systematic hierarchical knowledge base, which can provide rich and accurate data support for the safety analysis and evaluation of the operating environment and improve the utilization rate of data.

[0009] In an alternative embodiment, the index data of the operating environment elements includes real-time data of the operating environment elements, and the real-time data of the operating environment elements is obtained through the following method: Distribute a sensor network in a job environment monitoring scenario, and collect real-time data of job environment elements from multiple sensors in the sensor network. The real-time data of job environment elements includes a timestamp, a sensor measurement value, and a measurement location; Taking a preset reference time starting point or the initial acquisition time of the first sensor as a reference, perform offset compensation on the timestamps of the remaining sensors to generate real-time data of job environment elements under a unified time reference; If there are missing values in the real-time data of job environment elements, based on the data points before and after the missing position, calculate the supplementary value through bidirectional linear interpolation to generate complete real-time data of job environment elements.

[0010] A job environment safety analysis and evaluation method provided by the present invention realizes accurate alignment and fusion of implementation job data in time, and supplements missing values in the real-time data of job environment elements through a bidirectional linear interpolation method. The job environment data can more truly reflect the actual situation of the job environment, effectively solves the problems of incomplete data and inconsistent time, and improves the quality and usability of the data.

[0011] In an optional implementation manner, the job standard layer includes a safety range; the job environment index data further includes historical data of job environment elements; Compare the index data of job environment elements obtained in advance with a hierarchical knowledge base to identify risk indicators, calculate the risk values of the risk indicators at the same time, and construct a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values, including: Compare the real-time data of job environment elements with the safety range of the job standard layer, and identify the real-time data of job environment elements not within the safety range as risk indicators; Based on the historical data of job environment elements, use an attention mechanism to assign two-dimensional weights to risk indicators, and calculate the risk values corresponding to the risk indicators by using a weighted calculation method; Establish an entity association between the risk indicators and the corresponding risk values, construct a safety risk analysis and evaluation set based on the entity association, and store the safety risk analysis and evaluation set in the index data layer.

[0012] A job environment safety analysis and evaluation method provided by the present invention can more accurately identify job environment risk indicators, effectively solves the problems of difficult dynamic weight assignment and accurate identification of job environment risk indicators, and improves the accuracy and reliability of safety evaluation.

[0013] In an optional implementation manner, the two-dimensional weights include the historical data attention weight of job environment elements and the real-time data attention weight of job environment elements; Based on the historical data of job environment elements, the attention mechanism is used to assign two-dimensional weights to risk indicators, and based on the two-dimensional weights, a weighted calculation method is used to calculate the risk value corresponding to the risk indicator, including: Calculating the attention weights of the historical data of job environment elements based on a pre-trained gated recurrent unit model; Calculating the abnormality degree and change rate of the risk indicator, calculating the real-time emergency score based on the abnormality degree and change rate of the risk indicator, and calculating the attention weights of the real-time data of job environment elements based on the real-time emergency score; Calculating the comprehensive attention weights of risk indicators based on the attention weights of the historical data of job environment elements and the attention weights of the real-time data of job environment elements; Calculating the risk value corresponding to the risk indicator by using a weighted calculation method based on the comprehensive attention weights of risk indicators.

[0014] A job environment safety analysis and evaluation method provided by the present invention mines the temporal dependence relationship of historical data through a recurrent neural network, captures the evolution law of risk indicators in the long-term dimension, avoids the one-sidedness of single reliance on real-time data, and the real-time emergency scoring mechanism combines the abnormality degree and change rate of risk indicators to dynamically reflect the immediate risk urgency of the job environment. The two-dimensional weight fusion enables the risk value calculation to include both "long-term risk accumulation" and cover "short-term sudden risks", and is closer to the complexity of the actual job environment. The attention mechanism can automatically identify the more critical features in historical data that affect risks, avoid the neglect of key information by traditional equal-weight average or fixed-weight models, and provide decision-making support for enterprise safety production management.

[0015] In an alternative embodiment, a risk knowledge graph for analyzing and evaluating job environment safety is constructed based on a safety risk analysis and evaluation set and an initial knowledge graph, including: Semantically matching each risk indicator in the safety risk analysis and evaluation set with the entity nodes in the indicator data layer, and establishing an association relationship between the risk indicator and the entity nodes at the corresponding level; Adjusting the visualization identification attributes of risk indicators according to the size of the risk value, and the visualization identification attributes include color and size; Calculating the semantic similarity between risk indicators and entity nodes in the job standard layer, impact analysis layer, and accident case layer respectively based on a pre-trained model, and establishing a cross-level association relationship based on the relationship between the semantic similarity and a preset threshold; Constructing a risk knowledge graph for evaluating job environment safety based on the cross-level association relationship.

[0016] The present invention provides a method for analyzing and evaluating the safety of working environments. The real-time risk indicators in the safety risk analysis and evaluation set are associated with different types of data such as the indicator data layer, the operating standard layer, the impact analysis layer, and the accident case layer in the initial knowledge graph. It is not only associated with the historical monitoring data of the indicator data layer, but also with the occupational exposure limit in the operating standard layer, the poisoning hazard in the impact analysis layer, and the similar accident cases in the accident case layer to form a complete knowledge chain. By establishing semantic associations between entities, the originally fragmented safety data is converted into structured knowledge. Presented in the form of a graph structure, the source, impact, and countermeasures of the risk indicators are clearly displayed, which is easy to understand and analyze. The visual identification attributes of the risk indicators are adjusted according to the size of the risk value. The intuitive visualization method allows safety managers to quickly locate high-risk areas and key risk indicators and take timely measures. By using a pre-trained model to calculate semantic similarity and establish cross-level associations, the implicit connection between risk indicators and other knowledge entities can be mined.

[0017] In an optional implementation, the semantic similarity between the risk indicator and the entity nodes in the operation standard layer, the impact analysis layer, and the accident case layer is calculated based on the pre-trained model, and a cross-level association relationship is established based on the relationship between the semantic similarity and the preset threshold, including: The semantic similarity between risk indicators and entities at the operation standard layer is calculated through the pre-trained model to establish compliance or violation relationships; Based on the semantic matching of risk indicators, the potential consequence description in the impact analysis layer is formed to form the risk propagation path and construct the causal relationship between risk and consequence; Multi-dimensional matching is performed between current risk indicators and historical accident cases at the accident case layer. When the matching degree exceeds the preset threshold, a similarity relationship is established and historical processing measures are associated to form risk prediction and disposal recommendations. Multi-dimensional matching includes indicator type similarity matching, numerical range similarity matching and time pattern similarity matching.

[0018] The present invention provides a method for analyzing and evaluating the safety of a working environment. By calculating the semantic similarity between risk indicators and entities at the working standard layer through a pre-trained model, the method can accurately identify the compliance or violation relationship between risk indicators and regulatory standards. Semantic matching is used to mine the potential connections between knowledge at different levels, and to organically integrate scattered knowledge of regulations, impact analysis, and accident cases. Based on semantic matching, a risk propagation path is formed, and a causal relationship between risk and consequence is constructed to achieve a systematic analysis of the impact of risk. The current risk indicators are matched with historical cases at the accident case layer in multiple dimensions such as indicator type, numerical range, and time pattern, so that the risk development trend can be predicted based on historical experience.

[0019] In a second aspect, the present invention provides a device for analyzing and evaluating the safety of a working environment, the device comprising: An initialization module for establishing a hierarchical knowledge base of the operation environment, establishing entity associations between its hierarchies, and generating an initial knowledge graph; An attention module for comparing the pre-acquired operation environment index data with the risk index identified by the hierarchical knowledge base, calculating the risk value of the risk index, and constructing a safety risk analysis and evaluation set based on the risk index and the corresponding risk value; A safety evaluation module for constructing a risk knowledge graph for analyzing and evaluating the safety of the operation environment based on the safety risk analysis and evaluation set and the initial knowledge graph, and performing safety analysis and evaluation on the operation environment based on the risk knowledge graph.

[0020] In a third aspect, the present invention provides a computer device, including: a memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to perform the operation environment safety analysis and evaluation method of the first aspect or any corresponding embodiment thereof.

[0021] In a fourth aspect, the present invention provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to cause a computer to perform the operation environment safety analysis and evaluation method of the first aspect or any corresponding embodiment thereof.

[0022] In a fifth aspect, the present invention provides a computer program product, including computer instructions, and the computer instructions are used to cause a computer to perform the operation environment safety analysis and evaluation method of the first aspect or any corresponding embodiment thereof. Description of the Drawings

[0023] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0024] Figure 1 is a flowchart of the operation environment safety analysis and evaluation method according to an embodiment of the present invention; Figure 2 is a flowchart of another operation environment safety analysis and evaluation method according to an embodiment of the present invention; Figure 3 is a flowchart of yet another operation environment safety analysis and evaluation method according to an embodiment of the present invention; Figure 4 is a flowchart of still another operation environment safety analysis and evaluation method according to an embodiment of the present invention; Figure 5 It is a schematic flowchart of yet another working environment safety analysis and evaluation method according to an embodiment of the present invention; Figure 6 It is a schematic diagram of a risk knowledge graph for evaluating the safety of a working environment according to an embodiment of the present invention; Figure 7 It is a structural block diagram of a working environment safety analysis and evaluation device according to an embodiment of the present invention; Figure 8 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed implementation manners

[0025] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] The field of work safety involves multi-source heterogeneous data such as monitoring data, hidden danger rectification materials, laws and regulations, and accident cases, and its types cover structured, semi-structured, and unstructured data. Due to the lack of an effective data integration mechanism, there is a common problem of low data utilization rate in the actual production process.

[0027] Currently, in response to the problem of low data utilization rate, some scholars have proposed and established knowledge graph construction methods and related invention patents in related fields. In some technical solutions, by processing historical construction data and establishing an initial knowledge graph, and processing the data at the power construction site, the risks existing at the site are determined, and then corresponding control plans are formulated, and the initial knowledge graph is dynamically updated; in some technical solutions, by preprocessing real-time collected data and historical data, a knowledge graph is constructed, and anomaly detection, accident prediction, and risk assessment are performed on the real-time collected data to generate warning information and formulate emergency measures. However, when dealing with multi-source heterogeneous data, the knowledge graph will face the problem of semantic confusion caused by "data islands" and "flat" architectures, which seriously restricts the dynamic retrieval efficiency. Using knowledge graph technology to carry out working environment safety analysis and evaluation faces technical problems such as rigid knowledge graph architectures and inaccurate risk index judgments. The embodiments of the present invention provide a working environment safety analysis and evaluation method, which integrates multi-source data through a multi-level knowledge base and uses an attention mechanism for weight assignment to construct a risk knowledge graph that can synthesize information from multiple aspects.

[0028] According to an embodiment of the present invention, an embodiment of a method for analyzing and evaluating the safety of an operating environment is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. And although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0029] In this embodiment, a method for analyzing and evaluating the safety of an operating environment is provided, which can be used for the above computer device. Figure 1 It is a flowchart of a method for analyzing and evaluating the safety of an operating environment according to an embodiment of the present invention, as Figure 1 shown, the process includes the following steps: Step S101, establish a hierarchical knowledge base for operating environment elements, and establish entity associations between its levels to generate an initial knowledge graph.

[0030] Specifically, a hierarchical knowledge base is a management method that organizes a knowledge base according to different hierarchical structures, aiming to improve the efficiency of knowledge management and the scalability of the system. A hierarchical knowledge base is usually divided according to factors such as the degree of abstraction, data source, application scenario, etc., and can be divided into multiple levels. A Knowledge Graph is a semantic network used to describe the relationships between entities. It is a representation method of semi-structured data used to describe entities, attributes, and the relationships between entities.

[0031] The hierarchical knowledge base in this embodiment includes an index data layer that stores real-time data of operating environment elements and historical data of operating environment elements; an operating standard layer that stores the names of regulations and standards, the effective time, and the corresponding safety ranges; an impact analysis layer that stores the impacts of operating environment indicators on operating personnel, on production, and the factors affecting operating environment indicators; an accident case layer that stores accident causes, data characteristics of operating environment indicators, and improvement measures.

[0032] Among them, operating environment elements include air, lighting, temperature, humidity, noise, vibration, dust, radiation, toxic and harmful substances, etc.

[0033] Store the above-mentioned hierarchical data in a relational database, and store the entity relationship network in a graph database, forming cross-level entities, and perform cross-level association of the entities of the cross-level entities through a pre-trained model to generate an initial knowledge graph for each operating environment element.

[0034] Step S102, compare the index data of the operating environment elements obtained in advance with the hierarchical knowledge base of the operating environment elements to identify risk indicators, calculate the risk values of the risk indicators at the same time, and construct a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values.

[0035] Specifically, the operation environment index data includes the real-time data of the operation environment elements described above and the historical data of the operation environment elements.

[0036] Compare the real-time data of the operation environment elements with the operation standard layer of the hierarchical knowledge base to identify risk indicators. Use the attention mechanism to assign two-dimensional weights to the risk indicators and obtain the risk value based on weighted calculation, where the two-dimensional weights include the historical data attention weight of the operation environment elements and the real-time data attention weight of the operation environment elements. Establish an entity association between the risk indicators and their risk values based on the two-dimensional weights, and combine them into a safety risk analysis and evaluation set.

[0037] Step S103, construct a risk knowledge graph for analyzing and evaluating the safety of the operation environment based on the safety risk analysis and evaluation set and the initial knowledge graph, and perform safety analysis and evaluation on the operation environment based on the risk knowledge graph.

[0038] Specifically, match each risk indicator in the safety risk analysis and evaluation set with the entity nodes in the indicator data layer, establish the association relationship between the risk indicator and the entity nodes at the corresponding level, calculate the semantic similarity between the risk indicator and the entity nodes in the operation standard layer, impact analysis layer, and accident case layer. If the similarity exceeds the preset threshold, establish a cross-level association relationship to generate a risk knowledge graph for evaluating the safety of the operation environment.

[0039] The operation environment safety analysis and evaluation method provided in this embodiment, by constructing a multi-level knowledge base, integrating multi-source data, comparing the pre-acquired operation environment index data with the hierarchical knowledge base to accurately identify risk indicators, and calculating the risk value of the risk indicators at the same time, finally forms a risk knowledge graph that can comprehensively integrate various aspects of information, which can be effectively applied to the safety evaluation of operation environments in different industries, improving the accuracy and universality of safety evaluation, and solving the problems of rigid knowledge graph architecture and inaccurate risk indicator judgment in the operation environment scenario.

[0040] In this embodiment, an operation environment safety analysis and evaluation method is provided, which can be used for the above-mentioned computer device. Figure 2 It is a flowchart of the operation environment safety analysis and evaluation method according to an embodiment of the present invention, as Figure 2 shown, and this process includes the following steps: Step S201, establish a hierarchical knowledge base of operation environment elements, and establish entity associations between its levels to generate an initial knowledge graph.

[0041] Specifically, the above step S201 includes: Step S2011, obtain the index data of operation environment elements, operation standards, influencing factors on operation environment elements, and operation accident cases.

[0042] Specifically, the operation environment index data includes the real-time data of operation environment elements and the historical data of operation environment elements. The operation standards include the name of the regulatory standard, the effective time, and the corresponding safety range. The safety range refers to the allowable fluctuation range of the monitoring indicators stipulated in the regulatory standard, which is determined by the threshold requirements in the regulatory clauses and is denoted as the interval [L i , H i . Each rule node embeds the name of the regulatory standard, the safety range, etc. The operation influencing factors include the impact of the operation environment index data on the operation personnel, the impact on production, and the factors affecting the operation environment index data. The operation accident cases include the accident causes, the characteristics of the operation environment index data, and the improvement measures.

[0043] Step S2012: Establish an index data layer based on the index data of operation environment elements, establish an operation standard layer based on the operation standards of operation environment elements, establish an impact analysis layer based on the influencing factors of operation environment elements, and establish an accident case layer based on operation accident cases.

[0044] Specifically, the index data layer is used to store the real-time data of operation environment elements and the historical data of operation environment elements. The operation standard layer is used to store the name of the regulatory standard, the effective time, and the corresponding safety range. The impact analysis layer is used to store the impact of the operation environment index data on the operation personnel, the impact on production, and the factors affecting the operation environment index data. The accident case layer is used to store the relevant operation accident causes, the characteristics of the operation environment index data, and the improvement measures. Among them, through entity extraction technology, the accident causes, the characteristics of the operation environment index data, and the improvement measures are extracted from the accident cases.

[0045] Step S2013: Screen the data and entity relationship networks in the index data layer, the operation standard layer, the impact analysis layer, and the accident case layer, store the data in a relational database, store the entity relationship network in a graph database, and generate a hierarchical knowledge base based on the stored relational database and the stored graph database.

[0046] Specifically, relational databases store data in tabular form, with strict data structures and relational models, and are suitable for storing data with strong structuring and clear row-column relationships. In this method for analyzing and evaluating job environment safety, the data in the index data layer, job standard layer, impact analysis layer, and accident case layer, such as real-time data and historical data of job environment elements, names of regulations and standards, effective times, and the impacts of job environment indicators on job personnel, all have clear field attributes and fixed formats. Storing this data in a relational database can utilize its powerful transaction processing capabilities and data integrity constraint mechanisms to ensure data consistency and accuracy. For example, when updating data in the job standard layer, transaction operations can ensure the synchronous update of related fields such as effective time and safety range, avoiding data inconsistency issues.

[0047] Graph databases, on the other hand, store data in graph structures, are good at processing complex relational networks, and can efficiently represent and query the association relationships between entities. The entity relationship network in job environment safety analysis and evaluation, including the relationships between entities within each level and the associations between cross-level entities, has a high degree of complexity and dynamism. Storing it in a graph database can facilitate relationship-based queries and analyses. For example, when querying the relationship between a certain job environment indicator and related accident cases, a graph database can quickly traverse nodes and edges to obtain association information, without the need for complex multi-table join operations like relational databases, greatly improving the efficiency of relationship queries. Graph databases store (entity, relationship, entity) triples.

[0048] Generate a hierarchical knowledge base based on the stored relational database and the stored graph database.

[0049] Step S2014, calculate the semantic similarity of cross-level entities in the hierarchical knowledge base through a pre-trained model, and perform cross-level association on entities with semantic similarity exceeding a preset threshold to generate an initial knowledge graph.

[0050] Specifically, the hierarchical knowledge base covers multi-source heterogeneous data such as the index data layer, job standard layer, impact analysis layer, and accident case layer. Before calculating semantic similarity, it is necessary to clean and standardize this data. For example, unify the units and formats of real-time monitoring data in the index data layer; extract key entities from the text data in the job standard layer, such as standard names, applicable scopes, specific values, etc.; perform preprocessing such as word segmentation and part-of-speech tagging on the unstructured text in the accident case layer to convert it into a structured form that can be processed by a computer.

[0051] Input the preprocessed data into a pre-trained model (such as natural language processing models like BERT, GPT series, etc.). Utilize the powerful feature extraction ability of the model to map each entity into a high-dimensional semantic vector. These vectors contain the semantic information, context information of the entity, as well as the potential association information with other entities. For example, for the entity "hydrogen sulfide concentration", its vector not only contains the literal meaning of this indicator, but also contains semantic features such as related attributes and impacts in the field of workplace safety.

[0052] Based on the generated entity semantic vectors, adopt a suitable similarity measurement method (such as cosine similarity, Euclidean distance, etc.) to calculate the semantic similarity between entities at different levels. Taking cosine similarity as an example, by calculating the cosine value of the angle between two vectors, measure their similarity degree in the semantic space. The closer the value is to 1, the more similar the semantics of the two entities are.

[0053] The system traverses the entities in the hierarchical knowledge base across levels. For each entity from the indicator data layer, calculate the semantic similarity with all entities in the operation standard layer, impact analysis layer, and accident case layer in turn. For example, for the entity "excessive dust concentration" in the indicator data layer, calculate its semantic similarity with the relevant standard clauses of dust concentration in the operation standard layer, the description of dust hazards in the impact analysis layer, and the accident cases of dust explosions in the accident case layer respectively, to obtain a series of similarity scores.

[0054] Set a preset threshold (such as 0.7) to screen entity pairs with strong semantic associations. Compare the calculated semantic similarity scores with the preset threshold. Only entity pairs with similarity scores exceeding the threshold are considered to have sufficient semantic associations, and then cross-level association relationships are established. For example, if the semantic similarity between "excessive dust concentration" and a certain regulatory standard clause is 0.8, which is greater than the preset threshold of 0.7, it is determined that the two are related.

[0055] According to the characteristics of the level where the entity is located, define different types for the established cross-level association relationships. For example, the association relationship between the indicator data layer and the operation standard layer may be "compliance", "violation"; between the indicator data layer and the impact analysis layer may be "may cause", "trigger"; between the indicator data layer and the accident case layer may be "similar to", "related to", etc. At the same time, add relevant attributes to the association relationship, such as the confidence level of the association (which can be transformed according to the semantic similarity score), the basis of the association, and other information.

[0056] Store the entities and their associations that have been screened and defined relationships in the form of nodes and edges into the graph database to construct an initial knowledge graph. Each entity serves as a node, and the association relationship between entities serves as an edge, thus forming a structured and semantic knowledge graph to achieve the systematic organization and efficient management of knowledge related to workplace safety.

[0057] Exemplarily, taking a chemical plant as an example, the index data layer stores the real-time data and historical data of the working environment elements such as illumination intensity, temperature, and concentration of toxic and harmful gases. The working standard layer analyzes relevant regulations and standards to obtain the names of regulations and standards, effective times, safety ranges, etc. The impact analysis layer stores the impacts of working environment indicators on workers, on production, and the factors affecting the working environment indicators. The accident case layer is used to extract accident causes, data characteristics of working environment indicators, and improvement measures. After establishing the hierarchical knowledge base, the data of each layer is stored in a relational database, and the entity relationship network is stored in a graph database. The semantic similarity of cross-layer entities is calculated through a pre-trained model, and the entities with similarity exceeding the threshold are cross-layer associated to generate an initial knowledge graph.

[0058] Exemplarily, taking a chemical plant as an example, at the index data layer, an entity node "Illuminance Monitoring Point A01" is created to store the real-time illuminance data and the historical illuminance data of the past 30 days. At the working standard layer, by analyzing HG / T 20586 "Technical Regulations for Lighting Design of Chemical Enterprises", the illuminance threshold [40, 300] for chemical plant buildings is extracted, and a rule node "Rule_Light_lx" is created. At the impact analysis layer, a node "Health Risk of Hydrogen Sulfide Concentration" and its attribute "May cause respiratory diseases and nervous system damage to workers, affecting work efficiency" are created. In the accident case layer, a report on a hydrogen sulfide poisoning accident in a certain place in August 2021 is extracted, and nodes such as "Incident_2021-08", "H2S_Conc_High", "H2S_15mg m / m³", and "Arrange special personnel to regularly check, calibrate, and maintain gas detectors" are obtained, and a "has_risk_relation" relationship is established with the monitoring point. The semantic similarity between different nodes is calculated through a semantic alignment engine, and if the threshold is met, a "comply_with" relationship is established to generate an initial cross-layer knowledge graph.

[0059] Step S202, the working environment indicator data includes the real-time data of the working environment elements. The real-time data of the working environment elements is obtained through the following methods: A sensor network is distributed in the working environment monitoring scenario, and the real-time data of the working environment elements is collected from multiple sensors of the sensor network. The real-time data of the working environment elements includes a timestamp, a sensor measurement value, and a measurement location. Based on a preset reference time start point or the initial acquisition time of the first sensor as a reference, the timestamps of the remaining sensors are offset compensated to generate the real-time data of the working environment elements under a unified time reference. If there are missing values in the real-time data of the working environment elements, based on the data points before and after the missing position, the missing values are calculated and supplemented through bidirectional linear interpolation to generate complete real-time data of the working environment elements.

[0060] Exemplarily, in a chemical plant, real-time data of working environment elements such as illuminance, temperature, and hydrogen sulfide concentration at monitoring points are collected through a sensor network. Taking the initial collection time of the first sensor as a reference, the timestamps of the remaining sensors are offset and compensated to generate a data sequence under a unified time reference. When the data of the hydrogen sulfide sensor is lost between 14:05 and 14:07, a supplementary value is calculated through bidirectional linear interpolation. For example, the weight of the data point on the left side of the missing position is 0.7, and the weight of the data point on the right side is 0.3. In this way, the missing value is estimated based on the information of adjacent data points. After filling in the missing value, the hydrogen sulfide concentration data remains at 12.0 mg / m³. The technical solution of this embodiment realizes the accurate alignment and fusion of operation data in time, and supplements the missing values in the real-time data of working environment elements through the bidirectional linear interpolation method. The working environment data can more truly reflect the actual situation of the working environment, effectively solves the problems of incomplete data and inconsistent time, and improves the quality and usability of the data.

[0061] Step S203: Compare the pre-obtained working environment index data with the hierarchical knowledge base to identify risk indicators, calculate the risk values of the risk indicators at the same time, and construct a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values. For details, please refer to Figure 1 Step S102 of the embodiment shown, which will not be elaborated here.

[0062] Step S204: Based on the safety risk analysis and evaluation set and the initial knowledge graph, construct a risk knowledge graph for analyzing and evaluating the safety of the working environment, and perform safety analysis and evaluation on the working environment based on the risk knowledge graph. For details, please refer to Figure 1 Step S103 of the embodiment shown, which will not be elaborated here.

[0063] The working environment safety analysis and evaluation method provided in this embodiment systematically classifies and structurally stores multi-source heterogeneous data by clarifying the functions and storage contents of different levels. By calculating the semantic similarity between entities at each level through a pre-trained model, entities with a similarity greater than the similarity threshold are associated, optimizing the logical level of knowledge representation, significantly improving the accuracy and response speed of data retrieval, being able to more comprehensively and accurately evaluate the safety of the working environment, and effectively solving the problem of weak semantic association between levels. Each level is interconnected to form a comprehensive and systematic hierarchical knowledge base, which can provide rich and accurate data support for the safety analysis and evaluation of the working environment and improve the utilization rate of data.

[0064] In this embodiment, a working environment safety analysis and evaluation method is provided, which can be used for the above computer device. Figure 3 It is a flowchart of the working environment safety analysis and evaluation method according to an embodiment of the present invention, as shown in Figure 3As shown in the figure, the process includes the following steps: Step S301, establish a hierarchical knowledge base of job environment elements, establish entity associations between its levels, and generate an initial knowledge graph. For details, please refer to Figure 2 Step S201 of the embodiment shown, which will not be elaborated here.

[0065] Step S302, compare the index data of the job environment elements obtained in advance with the hierarchical knowledge base of the job environment elements to identify risk indicators, calculate the risk values of the risk indicators at the same time, and construct a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values.

[0066] Specifically, the job standard layer includes a safety range; the job environment index data also includes historical data of the job environment elements; the above step S302 includes: Step S3021, compare the real-time data of the job environment elements with the safety range of the job standard layer, and identify the real-time data of the job environment elements that are not within the safety range as risk indicators.

[0067] Specifically, in the risk indicator identification stage, the system first carefully compares the real-time data of the job environment elements with the safety range of the job standard layer. The real-time data of the job environment elements is collected in real time by the sensor network, covering various environmental parameters, such as gas concentration, temperature and humidity, light intensity, etc. Each data point contains accurate timestamp, measurement value and measurement location information.

[0068] The job standard layer stores the safety ranges determined by national regulations, industry specifications, etc. For example, the short-term exposure limit standard value of hydrogen sulfide is 0 - 10.0 mg / m³. The system checks each parameter measurement value in the real-time data one by one through a preset comparison algorithm. If a certain parameter (such as the measured value of hydrogen sulfide concentration is 12.0 mg / m³) exceeds the corresponding safety range, this real-time data point will be initially marked as abnormal.

[0069] For the marked abnormal real-time data, the system further traces back its historical environmental data. If in the historical data, this parameter also frequently shows similar abnormal situations, or there is an abnormal fluctuation trend, then this parameter is identified as a risk indicator. For example, if the hydrogen sulfide concentration has been close to or exceeded the standard value many times recently, and the historical data shows that the concentration is likely to increase under certain working conditions, then the hydrogen sulfide concentration will be officially identified as a risk indicator and included in the subsequent evaluation process.

[0070] Step S3022, assign two-dimensional weights to the risk indicators based on the historical data of the job environment elements by using the attention mechanism, and calculate the risk values corresponding to the risk indicators by using the weighted calculation method.

[0071] In some alternative embodiments, the two-dimensional weight includes the historical data attention weight of the operation environment elements and the real-time data attention weight of the operation environment elements; the above step S3022 includes: Step a1, calculating the historical data attention weight of the operation environment elements based on a pre-trained gated recurrent unit model.

[0072] Specifically, the gated recurrent unit model is abbreviated as the GRU model. GRU (Gate Recurrent Unit) is an improved recurrent neural network (RNN), which can effectively capture long sequence dependencies through a gating mechanism and solve the gradient vanishing problem of traditional RNNs.

[0073] Collect the historical data of the operation environment elements of the target risk indicator. These data usually exist in the form of time series and contain the measured values of this indicator over a period of time in the past. For example, for the risk indicator of "hydrogen sulfide concentration", the concentration measurement data corresponding to each time point in the past few hours, days or even months will be obtained. After sorting these historical data into a sequence in chronological order, they are input into the pre-trained GRU model. During the training process of the model, through learning a large amount of historical data, it has mastered the potential patterns and laws of various operation environment indicator data. When processing the input data, the hidden layer of the GRU will continuously update its state, combining the information of the previous time step with the input of the current time step, and gradually extracting the key features in the data. The model outputs a value representing the historical data attention weight of the operation environment elements. This weight reflects the importance of historical data for the current risk assessment. The higher the weight, the closer the information contained in the historical data is related to the current risk situation. For example, if in a certain operation area, "accidents have frequently occurred after the abnormal increase of hydrogen sulfide concentration" in history, then when evaluating the current hydrogen sulfide concentration risk in this area, the attention weight of the historical data will be relatively high, indicating that historical experience has important reference value for the current risk judgment.

[0074] The calculation formula for the historical data attention weight of the operation environment elements is: α hist,i = softma x( Attn ( h i {(t)} , h i {(t-k)} ,…, h i {(t-1)} ))), where is the feature at the current moment, k is the length of the historical window, and Attn(·) is the additive attention function.

[0075] Step a2, calculate the abnormality degree and change rate of the risk index, calculate the real-time emergency score based on the abnormality degree and change rate of the risk index, and calculate the real-time data attention weight of the operation environment elements based on the real-time emergency score.

[0076] Specifically, calculate the abnormality degree of the risk index, that is, the deviation degree between the current measured value and the normal range or standard value. The calculation formula is: , represents the normal range, represents the safety range.

[0077] Taking "hydrogen sulfide concentration" as an example, if the short-term exposure limit standard value is 10.0mg / m³, and the current real-time measured value is 12.0mg / m³, then the abnormality degree can be calculated by (12.0 - 10.0) / 10.0, and the abnormality degree is 0.2, indicating that the current concentration exceeds the standard value by 20%. At the same time, calculate the change rate of the risk index, which is determined by analyzing the change of the measured values at adjacent time points. The calculation formula is: .

[0078] For example, comparing the hydrogen sulfide concentration value at the current moment with that at the previous moment, if the concentration rises rapidly in a short period of time, it indicates a large change rate and an increase in potential risk. Based on the abnormality degree and change rate calculated above, further calculate the real-time emergency score. Weighted summation or other methods can be used to combine the abnormality degree and change rate according to a certain weight ratio to obtain a value that can reflect the current risk emergency level. The emergency score calculation formula is: , where, is the Sigmoid function, z 1, z 2 is the industry weight coefficient, which is determined by logistic regression analysis of a large number of accident cases in the accident case layer, and the default value z 1 = 0.6, z 2 = 0.4.

[0079] For example, set the abnormality degree weight to 0.6 and the change rate weight to 0.4. Through the formula: Calculate the specific score. Finally, calculate the real-time data attention weight of the job environment elements according to the real-time emergency score. Generally speaking, the higher the real-time emergency score, the greater the real-time data attention weight of the corresponding job environment elements. The real-time emergency score can be mapped to a value within the range of [0, 1] as the real-time data attention weight of the job environment elements through a preset mapping rule. For example, when the real-time emergency score is 0.8, according to the mapping rule, its real-time data attention weight for the job environment elements is converted to 0.9, indicating that the current real-time data is of extremely high importance in risk assessment.

[0080] Step a3: Calculate the comprehensive attention weight of the risk index based on the historical data attention weight of the job environment elements and the real-time data attention weight of the job environment elements.

[0081] Specifically, after obtaining the historical data attention weight of the job environment elements and the real-time data attention weight of the job environment elements respectively, fuse the two to calculate the comprehensive attention weight of the risk index. Usually, the weighted average method is adopted, and different weight coefficients are assigned to the historical weight and the real-time weight according to actual needs. For example, if it is considered that in the current job environment, the real-time data is more critical for risk assessment, the historical data attention weight coefficient of the job environment elements can be set to 0.4, and the real-time data attention weight coefficient of the job environment elements can be set to 0.6. The comprehensive attention weight is calculated through the formula: Comprehensive attention weight = Historical data attention weight of job environment elements × 0.4 + Real-time data attention weight of job environment elements × 0.6.

[0082] Step a4: Calculate the risk value corresponding to the risk index by using a weighted calculation method based on the comprehensive attention weight of the risk index.

[0083] Specifically, combine some relevant features of the risk index (such as the degree of abnormality, the matching degree with the historical risk pattern, etc.) with the comprehensive attention weight, and obtain the final risk value through calculation methods such as weighted summation. The risk value calculation formula is , is a regulation coefficient used to balance the influence of the historical data of the job environment elements and the real-time data of the job environment elements. Among them, is the historical data attention weight of the job environment elements, is the real-time data attention weight of the job environment elements.

[0084] For example, set the anomaly degree weight to 0.5, the matching degree weight with the historical risk pattern to 0.3, and the comprehensive attention weight to 0.2. Calculate the specific risk value through the formula: risk value = anomaly degree × 0.5 + matching degree with the historical risk pattern × 0.3 + comprehensive attention weight × 0.2. This risk value can comprehensively reflect the impact of historical data experience and real-time data status on the current risk indicator, provide a quantitative basis for the safety analysis and evaluation of the working environment, and facilitate safety managers to more accurately judge the risk level and take corresponding preventive measures.

[0085] Step S3023, establish an entity association between the risk indicator and the corresponding risk value, and construct a safety risk analysis and evaluation set based on the entity association, and store the safety risk analysis and evaluation set in the indicator data layer.

[0086] Specifically, after obtaining the risk indicator and its corresponding risk value, the system establishes an entity association between the two. Based on the nodes and edges of the graph database, the risk indicator (such as the excessive concentration of hydrogen sulfide) and the risk value (such as the calculated 0.85) are used as different nodes respectively, and are connected by a specific type of edge (such as "has a risk value") to clarify their corresponding relationship. At the same time, add attributes to the associated edge, such as weight allocation details, calculation basis and other information, to enhance the interpretability of the association. Based on the above entity association, the system constructs a safety risk analysis and evaluation set. The evaluation set is presented in the form of structured data, including detailed information such as the risk indicator name, measurement value, measurement location, timestamp, risk value, two-dimensional weight value, etc. Finally, store the constructed safety risk analysis and evaluation set in the indicator data layer. Exemplarily, in a chemical plant, compare the working environment data [50 lx, 25 °C, 12.0 mg / m³] with the safety range [0, 10.0 mg / m³] of the operation standard layer, and find that the short-term exposure limit concentration of hydrogen sulfide is not within the safety range interval, and identify it as a risk indicator.

[0087] Call the pre-trained GRU model, input the historical data of the working environment elements of the short-term exposure limit concentration of hydrogen sulfide in the past 1 hour, and calculate the attention weight of the historical data of the working environment elements. At the same time, calculate the real-time emergency score E1 = 0.785 according to the indicator anomaly degree and change rate, and obtain the attention weight of the real-time data of the working environment elements. And calculate the comprehensive attention weight of the risk indicator. It is 0.888.

[0088] Establish an entity association between the risk indicator node and its risk value node, combine them into a safety risk analysis and evaluation set, and store it in the indicator data layer. Establish a risk indicator W i and its risk value w iThe entity associations among them are combined into a safety risk analysis and evaluation set, W = { (W1, w1), (W2, w2), …, (W n , w n )}, where i is the i-th risk indicator and is stored in the indicator data layer.

[0089] The technical solution of this embodiment can more accurately identify the risk indicators of the operating environment, effectively solve the problems of difficult dynamic weight allocation and accurate identification of the risk indicators of the operating environment, and improve the accuracy and reliability of safety evaluation.

[0090] Step S303: Based on the safety risk analysis and evaluation set and the initial knowledge graph, construct a risk knowledge graph for analyzing and evaluating the safety of the operating environment, and perform safety analysis and evaluation on the operating environment based on the risk knowledge graph.

[0091] Specifically, the above step S303 includes: Step S3031: Semantically match each risk indicator in the safety risk analysis and evaluation set with the entity nodes in the indicator data layer, and establish an association relationship between the risk indicator and the entity nodes at the corresponding level.

[0092] Specifically, when matching the risk indicators in the safety risk analysis and evaluation set with the entity nodes in the indicator data layer, the system first parses the text description of the risk indicator to extract key information. For example, for the risk indicator of "excessive hydrogen sulfide concentration", "hydrogen sulfide concentration" is extracted as the core keyword. Then, in the indicator data layer, the system searches for the matching entity nodes by combining keyword search and semantic understanding. The indicator data layer stores a large amount of historical and real-time operating environment indicator data, and the system will use a pre-trained language model (such as the BERT model, Bidirectional Encoder Representation from Transformers) to calculate the semantic similarity between the risk indicator and the text of each entity node. For example, calculate the semantic similarity between "excessive hydrogen sulfide concentration" and the entity nodes such as "hydrogen sulfide concentration monitoring data" and "historical change trend of hydrogen sulfide concentration" recorded in the indicator data layer.

[0093] When the semantic similarity between a certain entity node and the risk indicator exceeds a preset threshold (such as 0.7), the system determines that the two match successfully and establishes an association relationship. This association relationship not only includes a simple correspondence relationship, but also records metadata such as the confidence level and association time of the association. For example, it is clear that the risk indicator of "excessive hydrogen sulfide concentration" has a strong association with the entity node of "hydrogen sulfide concentration monitoring data" within a specific time period in the indicator data layer, providing a data traceability path for subsequent risk analysis.

[0094] Step S3032: Adjust the visualization identification attributes of the risk indicators according to the magnitude of the risk value. The visualization identification attributes include color and size.

[0095] Specifically, to more intuitively display the risk level, the system adjusts the visualization identification attributes of the risk indicators according to the magnitude of the risk value. For the color attribute, a gradient mapping method is adopted, and different risk value intervals are set to correspond to different colors. For example, the risk value interval of 0 - 0.3 is mapped to green, indicating low risk; the interval of 0.3 - 0.7 is mapped to yellow, indicating medium risk; the interval of 0.7 - 1 is mapped to red, indicating high risk. For instance, if the risk value of the risk indicator "exceedance of hydrogen sulfide concentration" is 0.85, the system sets the color of its visualization node to red to warn of the severity of this risk. Regarding the adjustment of the size attribute, linear or non - linear scaling is also performed according to the magnitude of the risk value. When the risk value is set to 0, the node size is the minimum reference value (such as a diameter of 10 pixels); as the risk value increases, the node size increases accordingly. For example, when the risk value is 1, the node diameter is enlarged to 30 pixels. Through the change in size, safety managers can quickly identify high - risk indicators from the visualization interface and focus on key risk points.

[0096] Step S3033: Calculate the semantic similarity between the risk indicators and the entity nodes in the operation standard layer, impact analysis layer, and accident case layer based on the pre - trained model, and establish cross - level association relationships based on the relationship between the semantic similarity and the preset threshold.

[0097] In some optional implementation manners, the two - dimensional weight includes the historical data attention weight of the operation environment elements and the real - time data attention weight of the operation environment elements; the above - mentioned step S3033 includes: Step b1: Calculate the semantic similarity between the risk indicator and the entities in the operation standard layer through the pre - trained model, and establish a compliance or violation relationship.

[0098] Specifically, using the pre - trained model, calculate the semantic similarity between the risk indicator and the entity nodes in the operation standard layer. The operation standard layer stores text contents such as various safety regulations and industry standards. For each risk indicator, the system extracts relevant clauses, limits, etc. in the operation standard layer as entity nodes.

[0099] Taking "exceedance of hydrogen sulfide concentration" as an example, the system retrieves the standard clauses related to hydrogen sulfide concentration from the operation standard layer, such as the regulations on the short - term exposure limit value of hydrogen sulfide in "Occupational Exposure Limits for Hazardous Agents in the Workplace". Calculate the semantic similarity between this risk indicator and these standard clauses through the pre - trained model. If the similarity exceeds the preset threshold (such as 0.6), then establish a "violation" or "non - compliance" association relationship, and mark the specific violated standard content and limit value, clearly presenting the deviation of the risk indicator from the regulatory requirements.

[0100] Step b2: Based on the semantic matching of risk indicators with the potential consequence descriptions in the impact analysis layer, form a risk propagation path and construct the causal relationship between risks and consequences.

[0101] Specifically, the impact analysis layer includes the analysis of possible consequences, impact paths, etc. of operation environment indicators. The system performs semantic matching between risk indicators and the entity nodes in the impact analysis layer to mine the potential consequences that risks may trigger. For example, for the risk indicator of "excessive hydrogen sulfide concentration", entity nodes with relevant descriptions such as "may cause personnel poisoning" and "may corrode equipment" in the impact analysis layer are found through semantic matching. When the semantic similarity meets the conditions, causal association relationships such as "may cause" and "trigger" are established to form a propagation path from the risk indicator to the potential consequence. At the same time, attributes such as weights or confidence levels are added to the association relationships to reflect the likelihood of risks triggering corresponding consequences, helping safety managers comprehensively understand the scope of influence and harm degree of risks.

[0102] Step b3: Perform multi-dimensional matching between the current risk indicator and the historical accident cases in the accident case layer. When the matching degree exceeds the preset threshold, establish a similarity relationship and associate the historical handling measures to form risk prediction and disposal suggestions; the multi-dimensional matching includes index type similarity matching, numerical range similarity matching, and time pattern similarity matching.

[0103] Specifically, the accident case layer stores the detailed information of historical operation environment accidents. The system performs multi-dimensional matching between the current risk indicator and the historical accident cases in the accident case layer. In addition to calculating semantic similarity, dimensions such as index type, numerical range, and time pattern are also considered. For example, for the risk indicator of "excessive hydrogen sulfide concentration", not only the semantic similarity with the descriptions of hydrogen sulfide accidents in historical cases is compared, but also features such as the numerical range of hydrogen sulfide concentration and the concentration change trend at the time of the accident are compared. When the multi-dimensional matching degree exceeds the preset threshold (such as 0.7), an association relationship of "similar to" is established, and information such as the handling measures and lessons learned from historical accident cases is associated. Through this association, a reference basis is provided for the response to the current risk, the reuse of historical experience is realized, and the effectiveness of risk disposal is improved.

[0104] Step S3034: Construct a risk knowledge graph for analyzing and evaluating the safety of the operation environment based on the cross-level association relationships.

[0105] Specifically, after the above cross-level association relationships are established, the system constructs a risk knowledge graph in the form of a graph based on these association relationships. Risk indicators, entity nodes in the operation standard layer, entity nodes in the impact analysis layer, entity nodes in the accident case layer, etc. are used as nodes, and various established association relationships are used as edges to form a structured and semantic knowledge network. The schematic diagram of the constructed risk knowledge graph is asFigure 6 as shown

[0106] Each node and edge is attached with rich attribute information, such as the name, type, risk value, visualization attributes of the node, etc., and the association type, weight, confidence level of the edge, etc. In this way, the risk knowledge graph not only presents the basic information of the risk indicators, but also comprehensively shows the complex associations between them and regulatory standards, potential impacts, and historical cases, providing an intuitive, comprehensive, and in-depth analysis-capable knowledge framework for the safety analysis and evaluation of the operating environment, facilitating risk judgment, formulation of preventive measures, and decision-making support for safety management personnel.

[0107] Exemplarily, taking the hydrogen sulfide concentration monitoring in a chemical plant as an example, the implementation process of S400 is specifically described. The risk indicator W1 in the safety risk analysis and evaluation set W is matched with the entity node "hydrogen sulfide monitoring point C03" in the indicator data layer to establish a "has_risk" association relationship, and the visualization attributes of this node are adjusted according to the risk value w1 = 0.888: the node color changes from black to red, and the diameter expands from 20px to 28px. After calculating the semantic similarity between the risk indicator W1 and the cross-level nodes based on the BERT pre-trained model, for the operation standard layer, the node "risk indicator W1" is semantically matched with the regulatory text of the node "Rule_H2S_Conc" to generate a similarity score of 0.92 (threshold 0.85), and a "violates_standard" association relationship is established; in the impact analysis layer, the "hydrogen sulfide concentration health risk" node is associated with the "factors affecting hydrogen sulfide concentration" node, and "triggers_health_impact" relationship and "requires_intervention" relationship are respectively constructed; in the accident case layer, the semantic similarity between the node "risk indicator W1" and the accident cause "H2S_Conc_High" of the historical accident node "Incident_2021-08" reaches 0.89, and a "similar_to_incident" relationship is established. After completing the semantic association, triples such as (hydrogen sulfide monitoring point C03, has_risk, risk indicator W1), (risk indicator W1, violates_standard, Rule_H2S_Conc), (risk indicator W1, similar_to_incident, Incident_2021-08) are newly added to the graph database. In the finally formed risk knowledge graph, the nodes "risk indicator W1" and "hydrogen sulfide monitoring point C03" are highlighted with large red icons. Provide decision-making support for enterprise safety production management.

[0108] The operation environment safety analysis and evaluation method provided in this embodiment associates the real-time risk indicators in the safety risk analysis and evaluation set with different types of data such as the indicator data layer, operation standard layer, impact analysis layer, and accident case layer in the initial knowledge graph. It not only associates with the historical monitoring data in the indicator data layer but also can be linked to the occupational exposure limits in the operation standard layer, the poisoning hazards in the impact analysis layer, and the similar accident cases in the accident case layer, forming a complete knowledge chain. By establishing semantic association relationships between entities, the originally fragmented safety data is transformed into structured knowledge. Presented in the form of a graph structure, it clearly shows information such as the source, impact, and countermeasures of risk indicators, facilitating understanding and analysis. Adjust the visualization identification attributes of risk indicators according to the risk value size. With an intuitive visualization method, safety managers can quickly locate high-risk areas and key risk indicators and take measures in a timely manner. Using a pre-trained model to calculate semantic similarity and establish cross-level association relationships can uncover the implicit connections between risk indicators and other knowledge entities. As one or more specific application embodiments of the present invention, in combination with Figures 4 to 6 A further detailed description of the operation environment safety analysis and evaluation method provided by the present invention is as follows: As Figure 4 shown, this embodiment provides an operation environment safety analysis and evaluation method, which includes: S100, establish a hierarchical knowledge base of operation environment elements, establish entity associations between its levels, and generate an initial knowledge graph.

[0109] S200, the sensor network collects real-time data of operation environment elements, and performs timestamp alignment and missing value supplementation.

[0110] S300, compare the real-time data of operation environment elements with the operation standard layer of the hierarchical knowledge base, identify risk indicators, use the attention mechanism to assign two-dimensional weights to risk indicators and calculate risk values, and combine the risk indicators and their risk values into a safety risk analysis and evaluation set.

[0111] S400, associate the safety risk analysis and evaluation set with the initial knowledge graph to form a risk knowledge graph for analyzing and evaluating operation environment safety.

[0112] Furthermore, as Figure 5 shown, this embodiment provides a knowledge graph-based operation environment safety analysis and evaluation method, which includes: Step S1, establish a hierarchical knowledge base D of the operation environment, establish entity associations between its levels, and generate an initial knowledge graph. The hierarchical knowledge base D includes an indicator data layer, an operation standard layer, an impact analysis layer, and an accident case layer.

[0113] Specifically, a hierarchical knowledge base D is established, where: The index data layer includes, but is not limited to, real-time data of operation environment elements corresponding to operation environment indicators such as storage lighting, temperature, humidity, noise, vibration, dust, radiation, concentration of toxic and harmful substances, and ventilation volume, as well as historical data of operation environment elements; The operation standard layer extracts the safety range in the standard specifications through structured parsing technology. The safety range refers to the allowable fluctuation range of the monitoring indicators stipulated in the regulations and standards, which is determined by the threshold requirements in the regulation clauses and is denoted as the interval [L i , H i , and each rule node embeds the regulation name, safety range, etc.; The impact analysis layer stores the impacts of operation environment indicators on operators, on production, and the factors affecting operation environment indicators; The accident case layer stores a large number of accident cases. Through entity extraction technology, accident causes, data characteristics of operation environment indicators, and improvement measures are extracted from the accident cases.

[0114] When generating the initial knowledge graph based on the hierarchical knowledge base D, a hybrid storage architecture and a semantic alignment algorithm are used to achieve multi-source data fusion. The relational database stores hierarchical metadata and index information, and the graph database stores (entity, relationship, entity) triples. The semantic alignment engine calculates the text similarity through a pre-trained model, identifies nodes with consistent semantics in different hierarchies, and establishes cross-hierarchy association relationships.

[0115] Step S2: Collect real-time data U of operation environment elements from multiple sensors in the sensor network, and perform timestamp alignment and missing value supplementation.

[0116] Specifically, collecting real-time data U of operation environment elements from multiple sensors in the sensor network and performing timestamp alignment and missing value supplementation includes: Collecting real-time data of operation environment elements through a distributed sensor network deployed in the operation environment, where the sensor nodes adopt a redundant check protocol to ensure data transmission integrity. The real-time data of the operation environment elements includes timestamps, sensor measurement values, and measurement locations. Taking the preset reference time starting point or the initial acquisition moment of the first sensor as the benchmark, offset compensation is performed on the timestamps of the remaining sensors to generate a data sequence under a unified time benchmark. If there are missing values in the real-time data of the operation environment elements, based on the left and right data points at the missing position, the supplementary value is calculated through bidirectional linear interpolation to generate a complete data sequence. For example, the weight of the left data point at the missing position is 0.7, and the weight of the right data point is 0.3.

[0117] Step S3: Compare the real-time data U of the operation environment elements with the operation standard layer of the hierarchical knowledge base of the operation environment elements, identify the risk indicators Wi, assign weights to the risk indicators using the attention mechanism and calculate the risk values wi, and combine the risk indicators and their risk values into the safety risk analysis and evaluation set W = {(W1, w1), (W2, w2), …, (Wn, wn)}, where i is the i-th risk indicator.

[0118] Specifically, let the real-time data of the operation environment elements be an n-dimensional vector U = u 1, u 2, …, u n , and the safety range corresponding to the operation standard layer is [L i , H i , then the risk indicator is . Assign two-dimensional weights to the risk indicators using the attention mechanism and obtain the risk values based on weighted calculation. Among them, the two-dimensional weights include the historical data attention weight and the real-time data attention weight of the operation environment elements.

[0119] First, extract the time series features of the indicators based on the deep learning model (LSTM, GRU, etc.) and calculate the historical data attention weight of the operation environment elements: = softma x( Attn ( h i {(t)} , h i {(t-k)} , …, h i {(t-1)} )), where is the feature at the current moment, k is the historical window length, and Attn(·) is the additive attention function. Second, according to the anomaly degree and the change rate of the risk indicator, calculate the real-time emergency score: , where is the Sigmoid function, z 1, z 2 are the industry weight coefficients determined by logistic regression analysis of a large number of accident cases in the accident case layer, and the default values z 1 = 0.6, z 2 = 0.4. Furthermore, obtain the real-time data attention weight of the operation environment elements: . Based on the two-dimensional weights of the risk indicators, calculate the risk values of the risk indicators by weighted calculation. The risk value calculation formula is , is an adjustment coefficient used to balance the influence of historical data and real-time data of operation environment elements. Establish a risk index W i and its risk value w i to establish an entity association between them, and combine them into a safety risk analysis and evaluation set W = { (W1, w1), (W2, w2), …, (W n , w n ), where i is the i-th risk index, and store it in the index data layer, is the attention weight of historical data of operation environment elements, is the attention weight of real-time data of operation environment elements, represents the real-time emergency score of the j-th risk index, is the total number of currently identified risk indexes.

[0120] Step S4: Associate the safety risk analysis and evaluation set with the initial knowledge graph to form a risk knowledge graph for analyzing and evaluating the safety of the operation environment.

[0121] Specifically, associating the safety risk analysis and evaluation set with the initial knowledge graph to form a risk knowledge graph for evaluating the safety of the operation environment includes: matching each risk index in the safety risk analysis and evaluation set with the entity nodes in the index data layer, establishing an association relationship between the risk index and the entity nodes at the corresponding level, and adjusting the visualization identification attributes of the risk index according to the magnitude of the risk value, including color and size; calculating the semantic similarity between the risk index and the entity nodes in the operation standard layer, impact analysis layer, and accident case layer based on a pre-trained model, and establishing a cross-level association relationship if the similarity exceeds a preset threshold.

[0122] Taking the hazardous chemical production workshop of a chemical plant as an example, the implementation process of the present invention is specifically described below. This area is equipped with a light intensity sensor for monitoring the area illuminance (Area_lx); a temperature sensor for monitoring the area temperature (Area_temp); a hydrogen sulfide sensor for monitoring the hydrogen sulfide concentration (H2S_Conc), as Figure 5 shown, the specific implementation steps are as follows: S1: Establish a hierarchical knowledge base D for specific operation environment elements, establish entity associations between its levels, and generate an initial knowledge graph. The hierarchical knowledge base D includes an index data layer, an operation standard layer, an accident case layer, and an impact analysis layer.

[0123] Specifically, the index data layer creates entity nodes such as "Illuminance Monitoring Point A01", "Temperature Monitoring Point B02", and "Hydrogen Sulfide Monitoring Point C03", and the monitoring data of these nodes is stored and associated with the corresponding relational database; the operation standard layer parses HG / T 20586 "Technical Regulations for Lighting Design in Chemical Enterprises", extracts the illuminance threshold [40, 300] for chemical plant workshops, and creates nodes "Rule_Light_lx", "HG / T 20586 'Technical Regulations for Lighting Design in Chemical Enterprises'", and "[40, 300]". Similarly, it parses GB50016-2014 "Code for Design of Industrial Enterprises", creates nodes "Rule_Area_Temp", "GB 50016-2014 'Code for Design of Industrial Enterprises'", and "[20°C, 28°C]", and parses GBZ 2.1-2019 "Occupational Exposure Limits for Hazardous Agents in the Workplace - Part 1: Chemical Hazardous Agents", extracts the short-term exposure limit concentration of hydrogen sulfide as 10.0 mg / m³, and creates nodes "Rule_H2S_Conc", "GBZ 2.1-2019 'Occupational Exposure Limits for Hazardous Agents in the Workplace - Part 1: Chemical Hazardous Agents'", and "10.0 mg / m³"; the impact analysis layer, based on the detailed analysis of the operation environment report, creates nodes such as "Health Risk of Hydrogen Sulfide Concentration" and its attributes "May cause respiratory diseases and nervous system damage to operators, affecting operation efficiency", "Production Impact of Hydrogen Sulfide Concentration" and its attributes "Cause equipment corrosion, shorten equipment life, and increase maintenance costs", and "Factors Affecting Hydrogen Sulfide Concentration" and its attributes "Low efficiency of ventilation system, gas leakage source not repaired in time, unreasonable workshop layout"; the accident case layer extracts the hydrogen sulfide poisoning accident report in a certain place in August 2021, associates the environmental state characteristics (illuminance 50 lx, temperature 30°C, peak hydrogen sulfide concentration 15 mg / m³), constructs event nodes "Incident_2021-08", accident cause nodes "H2S_Conc_High", operation environment index data characteristics "Light_50lx", "Temp_30°C", "H2S_15mg / m³", and improvement measure nodes "Arrange special personnel to regularly inspect, calibrate, and maintain gas detectors". At the same time, it establishes a "has_risk_relation" relationship with "Illuminance Monitoring Point A01", "Temperature Monitoring Point B02", and "Hydrogen Sulfide Monitoring Point C03". These nodes establish a "has_influence_on" relationship with the accident inducement node "H2S_Conc_High". A semantic alignment engine based on the BERT model establishes cross-level associated triples. For example, the "comply_with" relationship between the node "Illuminance Monitoring Point A01" and "Rule_Light_lx". Similarly, relationship triples for other operation environment indicators are established. The dataset corresponding to the knowledge graph is shown in Table 1 below: Table 1 Dataset corresponding to the knowledge graph

[0124] S2: Collect real-time data U of operation environment elements from multiple sensors in the sensor network, and perform timestamp alignment and missing value supplementation. The steps include: Collect real-time data U of operation environment elements from multiple sensors in the sensor network, including the illuminance of the monitoring point u 1 = 50 lx , the temperature of the monitoring point is 25°C, and the hydrogen sulfide concentration u 3 = 12.0 mg / m³ (600 sampling points within 10 minutes, the data is only an example); Based on the initial collection time of the first sensor, perform offset compensation on the timestamps of the remaining sensors to generate a data sequence under a unified time reference; When the hydrogen sulfide sensor data is lost between 14:05 and 14:07, calculate the supplementary value through bilinear interpolation. For example, when the distance is 6m, the weight is 0.7. In this way, estimate the missing value according to the information of adjacent data points, and keep the hydrogen sulfide concentration data at 12.0 mg / m³ after filling in the missing value.

[0125] S3: Compare the real-time data U of operation environment elements with the operation standard layer of the hierarchical knowledge base to identify risk indicators W i , use the attention mechanism to assign weights to the risk indicators and calculate the risk value w i , combine the risk indicators and their risk values into a safety risk analysis and evaluation set W = {(W1, w1), (W2, w2), …, (W n , w n ), where i is the i-th risk indicator. It includes: Compare the operation environment data [50 lx, 25°C, 12.0 mg / m³] with the safety range [0, 10.0 mg / m³] of the operation standard layer, and find that the short-term exposure limit concentration of hydrogen sulfide is not within the safety range interval, and identify it as the risk indicator W1. Use the attention mechanism to assign two-dimensional weights to the risk indicator, call the pre-trained GRU model, input the historical data of operation environment elements of the short-term exposure limit concentration of hydrogen sulfide in the past 1 hour, and calculate the attention weight α hist = 0.72 of the historical data of operation environment elements. At the same time, according to the abnormal degree and the change rate of the short-term exposure limit concentration of hydrogen sulfide, substitute them into the Sigmoid function to calculate the real-time emergency score E 1 = 0.785, and then obtain the real-time attention weight . Assume the adjustment coefficient = 0.2, the calculated risk value w1 is 0.888. Combine the risk indicators and their risk values into the safety risk analysis and evaluation set W = {(W1, w1)}.

[0126] S4: Associate the safety risk analysis and evaluation set W with the initial knowledge graph to form a risk knowledge graph for analyzing and evaluating the safety of the working environment, including: Match the risk indicator W1 in the safety risk analysis and evaluation set W with the entity node "Hydrogen Sulfide Monitoring Point C03" in the indicator data layer, establish a "has_risk" association relationship, and adjust the visualization attributes of this node according to the risk value w1 = 0.888: the node color changes from black to red, and the diameter expands from 20px to 28px. After calculating the semantic similarity between the risk indicator W1 and the cross-level nodes based on the BERT pre-trained model, for the operation standard layer, semantically match the node "Risk Indicator W1" with the regulatory text of the node "Rule_H2S_Conc", generate a similarity score of 0.92 (threshold 0.85), and establish a "violates_standard" association relationship; in the accident case layer, the semantic similarity between the node "Risk Indicator W1" and the accident cause "H2S_Conc_High" of the historical accident node "Incident_2021-08" reaches 0.89, and establish a "similar_to_incident" relationship. After completing the semantic association, add triples such as (Hydrogen Sulfide Monitoring Point C03, has_risk, Risk Indicator W1), (Risk Indicator W1, violates_standard, Rule_H2S_Conc), (Risk Indicator W1, similar_to_incident, Incident_2021-08) to the graph database. The finally formed risk knowledge graph is as Figure 6 shown. Among them, the nodes "Risk Indicator W1" and "Hydrogen Sulfide Monitoring Point C03" are highlighted with large red icons, providing decision-making support for the enterprise's work safety management.

[0127] The hierarchical structure can not only optimize the logical level of knowledge representation but also significantly improve the accuracy and response speed of data retrieval by establishing multi-dimensional classification paths and semantic association networks. It should be noted that although there are significant differences in process standards, operation specifications, etc. among different industries, the working environment, as a common core element of work safety, is always a key area for risk control in enterprises of all industries. Therefore, constructing a working environment safety analysis and evaluation system based on a hierarchical knowledge graph has significant technical innovation value and practical guiding significance.

[0128] In this embodiment, a device for analyzing and evaluating the safety of the working environment is further provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be elaborated here. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0129] This embodiment provides a device for analyzing and evaluating the safety of the working environment, as Figure 7 shown, including: An initialization module 701, configured to establish a hierarchical knowledge base of working environment elements, establish entity associations between its levels, and generate an initial knowledge graph.

[0130] An attention module 702, configured to compare the index data of the working environment elements obtained in advance with the hierarchical knowledge base to identify risk indicators, calculate the risk values of the risk indicators at the same time, and construct a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values.

[0131] A safety evaluation module 703, configured to construct a risk knowledge graph for analyzing and evaluating the safety of the working environment based on the safety risk analysis and evaluation set and the initial knowledge graph, and perform safety analysis and evaluation on the working environment based on the risk knowledge graph.

[0132] In some alternative implementation manners, the initialization module 701 includes: A data acquisition unit, configured to acquire the index data of the working environment elements, working standards, influencing factors on the working environment elements, and working accident cases.

[0133] A level establishment unit, configured to establish an index data layer based on the index data of the working environment elements, a working standard layer based on the working standards of the working environment elements, an impact analysis layer based on the influencing factors on the working environment elements, and an accident case layer based on the working accident cases.

[0134] A hierarchical knowledge base generation unit, configured to screen the data and entity relationship networks in the index data layer, working standard layer, impact analysis layer, and accident case layer, store the data in a relational database, store the entity relationship network in a graph database, and generate a hierarchical knowledge base based on the stored relational database and the stored graph database.

[0135] An initial knowledge graph construction unit, configured to calculate the semantic similarity of cross-level entities in the hierarchical knowledge base through a pre-trained model, and perform cross-level association on the entities whose semantic similarity exceeds a preset threshold to generate an initial knowledge graph.

[0136] In some alternative embodiments, the indicator data of the operating environment elements includes real-time data of the operating environment elements. The operating environment safety analysis and evaluation device further includes: A real-time data acquisition module for the operating environment elements, which is used to deploy a sensor network distributively in the operating environment monitoring scenario, and collect real-time data of the operating environment elements from multiple sensors of the sensor network. The real-time data of the operating environment elements includes a timestamp, a sensor measurement value, and a measurement location; taking a preset reference time starting point or the initial acquisition moment of the first sensor as a reference, perform offset compensation on the timestamps of the remaining sensors to generate real-time data of the operating environment elements under a unified time reference; if there are missing values in the real-time data of the operating environment elements, then based on the data points before and after the missing position, calculate the supplementary value through bidirectional linear interpolation to generate complete real-time data of the operating environment elements.

[0137] In some alternative embodiments, the operation standard layer includes a safety range; the operating environment indicator data further includes historical data of the operating environment elements; the attention module 702 includes: A risk indicator identification unit, which is used to compare the real-time data of the operating environment elements with the safety range of the operation standard layer, and identify the real-time data and historical environment data of the operating environment elements that are not within the safety range as risk indicators.

[0138] An attention unit, which is used to assign two-dimensional weights to the risk indicators by using an attention mechanism based on the historical data of the operating environment elements, and calculate the risk value corresponding to the risk indicators by using a weighted calculation method based on the two-dimensional weights. An entity association establishment and safety risk analysis and evaluation set construction unit, which is used to establish an entity association between the risk indicators and the corresponding risk values, construct a safety risk analysis and evaluation set based on the entity association, and store the safety risk analysis and evaluation set in the indicator data layer. In some alternative embodiments, the two-dimensional weights include the historical data attention weight of the operating environment elements and the real-time data attention weight of the operating environment elements; the attention unit includes: A historical data attention weight calculation subunit for the operating environment elements, which is used to calculate the historical data attention weight of the operating environment elements based on a pre-trained gated recurrent unit model.

[0139] A real-time data attention weight calculation subunit for the operating environment elements, which is used to calculate the anomaly degree and change rate of the risk indicators, calculate a real-time emergency score based on the anomaly degree and change rate of the risk indicators, and calculate the real-time data attention weight of the operating environment elements based on the real-time emergency score.

[0140] A comprehensive attention weight calculation subunit, which is used to calculate the comprehensive attention weight of the risk indicators based on the historical data attention weight of the operating environment elements and the real-time data attention weight of the operating environment elements.

[0141] A risk value calculation subunit, configured to calculate the risk value corresponding to a risk indicator by using a weighted calculation method based on the comprehensive attention weight of the risk indicators.

[0142] In some optional implementation manners, the safety evaluation module 703 includes: An association relationship establishment unit, configured to perform semantic matching between each risk indicator in the safety risk analysis and evaluation set and the entity nodes in the indicator data layer, and establish an association relationship between the risk indicator and the entity nodes at the corresponding level.

[0143] A visualization unit, configured to adjust the visualization identification attributes of the risk indicators according to the magnitude of the risk value, where the visualization identification attributes include color and size.

[0144] A cross-level association relationship establishment unit, configured to calculate the semantic similarity between the risk indicator and the entity nodes in the operation standard layer, impact analysis layer, and accident case layer respectively based on a pre-trained model, and establish a cross-level association relationship based on the relationship between the semantic similarity and a preset threshold.

[0145] A risk knowledge graph construction unit, configured to construct a risk knowledge graph for analyzing and evaluating the safety of the operation environment based on the cross-level association relationship.

[0146] In some optional implementation manners, the cross-level association relationship establishment unit includes: A first relationship establishment subunit, configured to calculate the semantic similarity between the risk indicator and the entity in the operation standard layer through a pre-trained model, and establish a compliance or violation relationship.

[0147] A second relationship establishment subunit, configured to form a risk propagation path based on the semantic matching of the potential consequences description in the impact analysis layer by the risk indicator, and construct a causal relationship between the risk and the consequences.

[0148] A third relationship establishment subunit, configured to perform multi-dimensional matching between the current risk indicator and the historical accident cases in the accident case layer. When the matching degree exceeds the preset threshold, establish a similarity relationship and associate the historical treatment measures to form risk prediction and disposal suggestions; the multi-dimensional matching includes index type similarity matching, numerical range similarity matching, and time pattern similarity matching.

[0149] The further function descriptions of the above-mentioned various modules and units are the same as those in the corresponding embodiments above, and will not be elaborated here.

[0150] The working environment safety analysis and evaluation device in this embodiment is presented in the form of functional units. Here, the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and a memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0151] An embodiment of the present invention further provides a computer device having the above Figure 7 working environment safety analysis and evaluation device shown.

[0152] Please refer to Figure 8 , Figure 8 which is a schematic structural diagram of a computer device provided by an optional embodiment of the present invention. As Figure 8 shown, the computer device includes: one or more processors 10, a memory 20, and an interface for connecting each component, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor system). Figure 8 In

[0153] FIG. 14, one processor 10 is taken as an example.

[0154] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.

[0155] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 20 may include high-speed random access memory, and may also include non-transitory memory, such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage devices. In some alternative embodiments, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories can be connected to the computer device through a network. Examples of the above-mentioned network include but are not limited to the Internet, intranet, local area network, mobile communication network, and combinations thereof.

[0156] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk, or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0157] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30, and the output device 40 may be connected through a bus or other means. Figure 8 Taking connection through the bus as an example.

[0158] The input device 30 can receive input digital or character information, and generate key signal inputs related to the user settings and function control of the computer device, such as a touch screen, keypad, mouse, trackpad, touchpad, pointing stick, one or more mouse buttons, trackball, joystick, etc. The output device 40 may include a display device, auxiliary lighting device (such as an LED), and tactile feedback device (such as a vibration motor), etc. The above-mentioned display device includes but is not limited to liquid crystal display, light-emitting diode, display, and plasma display. In some alternative embodiments, the display device may be a touch screen.

[0159] Embodiments of the present invention also provide a computer-readable storage medium. The method according to the embodiments of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code that is originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and will be stored in a local storage medium, so that the methods described herein can be stored as such software processes on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods shown in the above embodiments are implemented.

[0160] A part of the present invention can be applied as a computer program product, for example, computer program instructions, which when executed by a computer, can call or provide the methods and / or technical solutions according to the present invention through the operations of the computer. Those skilled in the art should be able to understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible by the computer.

[0161] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for safety analysis and evaluation of an operating environment, characterized in that, The method includes: Establishing a hierarchical knowledge base of job environment elements, establishing entity associations between its levels, and generating an initial knowledge graph; Comparing the index data of job environment elements obtained in advance with the hierarchical knowledge base of job environment elements to identify risk indicators, calculating the risk values of the risk indicators at the same time, and constructing a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values; Constructing a risk knowledge graph for analyzing and evaluating job environment safety based on the safety risk analysis and evaluation set and the initial knowledge graph, and performing safety analysis and evaluation on the job environment based on the risk knowledge graph.

2. The method according to claim 1, characterized in that The establishing a hierarchical knowledge base of job environment elements, establishing entity associations between its levels, and generating an initial knowledge graph includes: Obtaining the index data of job environment elements, job standards, influencing factors on job environment elements, and job accident cases; Establishing an index data layer based on the index data of job environment elements, a job standard layer based on the job standards of job environment elements, an impact analysis layer based on the influencing factors on job environment elements, and an accident case layer based on the job accident cases; Screening the data and entity relationship networks in the index data layer, job standard layer, impact analysis layer, and accident case layer, storing the data in a relational database, storing the entity relationship networks in a graph database, and generating a hierarchical knowledge base based on the stored relational database and the stored graph database; Calculating the semantic similarity of cross-level entities in the hierarchical knowledge base through a pre-trained model, and associating the entities with semantic similarity exceeding a preset threshold across levels to generate an initial knowledge graph.

3. The method according to claim 2, wherein The index data of the job environment elements includes the real-time data of job environment elements, and the real-time data of job environment elements is obtained in the following manner: Distributively deploying a sensor network in a job environment monitoring scenario, and collecting the real-time data of job environment elements from multiple sensors of the sensor network. The real-time data of job environment elements includes a timestamp, a sensor measurement value, and a measurement location; Taking a preset reference time starting point or the initial acquisition moment of the first sensor as a benchmark, compensating for the time offset of the timestamps of the remaining sensors to generate the real-time data of job environment elements under a unified time benchmark; If there are missing values in the real-time data of the job environment elements, calculating supplementary values through bidirectional linear interpolation based on the data points before and after the missing position to generate complete real-time data of job environment elements.

4. The method according to claim 3, characterized in that, The job standard layer includes a safety range; the index data of the job environment elements also includes the historical data of job environment elements; The comparing the index data of job environment elements obtained in advance with the hierarchical knowledge base to identify risk indicators, calculating the risk values of the risk indicators of the job environment elements at the same time, and constructing a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values includes: Comparing the real-time data of the job environment elements with the safety range of the job standard layer, and identifying the real-time data of job environment elements not within the safety range as risk indicators; Based on the historical data of job environment elements, use the attention mechanism to assign two-dimensional weights to the risk indicators, and calculate the risk values corresponding to the risk indicators using a weighted calculation method based on the two-dimensional weights; Establish an entity association between the risk indicators and the corresponding risk values, construct a safety risk analysis and evaluation set based on the entity association, and store the safety risk analysis and evaluation set in the indicator data layer.

5. The method according to claim 4, characterized in that, The two-dimensional weights include the historical data attention weight of job environment elements and the real-time data attention weight of job environment elements; The method of using the attention mechanism to assign two-dimensional weights to the risk indicators based on the historical data of job environment elements and calculating the risk values corresponding to the risk indicators using a weighted calculation method based on the two-dimensional weights includes: Calculate the historical data attention weight of job environment elements based on a pre-trained gated recurrent unit model; Calculate the abnormality degree and change rate of the risk indicator, calculate the real-time emergency score based on the abnormality degree and change rate of the risk indicator, and calculate the real-time data attention weight of job environment elements based on the real-time emergency score; Calculate the comprehensive attention weight of the risk indicator based on the historical data attention weight of the job environment elements and the real-time data attention weight of the job environment elements; Calculate the risk value corresponding to the risk indicator using a weighted calculation method based on the comprehensive attention weight of the risk indicator.

6. The method according to claim 2, wherein The method of constructing a risk knowledge graph for evaluating job environment safety based on the safety risk analysis and evaluation set and the initial knowledge graph includes: Semantically match each risk indicator in the safety risk analysis and evaluation set with the entity nodes in the indicator data layer, and establish an association relationship between the risk indicator and the entity nodes at its corresponding level; Adjust the visualization identification attributes of the risk indicators according to the size of the risk value, and the visualization identification attributes include color and size; Calculate the semantic similarity between the risk indicators and the entity nodes in the job standard layer, impact analysis layer, and accident case layer respectively based on a pre-trained model, and establish a cross-level association relationship based on the relationship between the semantic similarity and a preset threshold; Construct a risk knowledge graph for evaluating job environment safety based on the cross-level association relationship.

7. The method according to claim 6, characterized in that, The method of calculating the semantic similarity between the risk indicators and the entity nodes in the job standard layer, impact analysis layer, and accident case layer respectively based on a pre-trained model and establishing a cross-level association relationship based on the relationship between the semantic similarity and a preset threshold includes: Calculate the semantic similarity between the risk indicator and the entity in the job standard layer through the pre-trained model, and establish a compliance or violation relationship; Based on the semantic matching of the risk indicator to the potential consequence description in the impact analysis layer, form a risk propagation path, and construct a causal relationship between risks and consequences; Perform multi-dimensional matching between the current risk indicator and the historical accident cases in the accident case layer. When the matching degree exceeds the preset threshold, establish a similarity relationship and associate the historical treatment measures to form risk prediction and disposal suggestions; the multi-dimensional matching includes index type similarity matching, numerical range similarity matching, and time pattern similarity matching.

8. An operating environment safety analysis and evaluation device, characterized in that, The device includes: An initialization module for establishing a hierarchical knowledge base of job environment elements, establishing entity associations between its levels, and generating an initial knowledge graph; An attention module for comparing the index data of job environment elements obtained in advance with the hierarchical knowledge base to identify risk indicators, calculating the risk values of the risk indicators, and constructing a safety risk analysis and evaluation set based on the risk indicators and the corresponding risk values; A safety evaluation module for constructing a risk knowledge graph for evaluating the safety of the job environment based on the safety risk analysis and evaluation set and the initial knowledge graph, and performing safety analysis and evaluation on the job environment based on the risk knowledge graph.

9. A computer device, characterized in that, Comprising: A memory and a processor, which are communicatively connected to each other. The memory stores computer instructions, and the processor executes the computer instructions to execute the job environment safety analysis and evaluation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, Computer instructions are stored on the computer-readable storage medium, and the computer instructions are used to cause a computer to execute the job environment safety analysis and evaluation method according to any one of claims 1 to 7.

Images