A risk management method, device, equipment and medium for limited space operation

Abstract

This application provides a risk management method, apparatus, equipment, and medium for confined space operations, applicable to the field of confined space operation safety technology. The method includes: constructing a risk semantic association graph and determining the neighborhood subgraph of spatial nodes; performing a weighted average of target risk values ​​based on similarity to obtain a target cross-site risk value, and extracting fields to obtain a target information set; calculating conditional probabilities of the target information set and environmental vectors to obtain a target unsupervised risk value, and determining a connection matrix; iterating the initial propagation risk value according to the connection matrix and preset iteration thresholds to obtain a target propagation risk value, and performing a weighted summation of the target cross-site risk value, the target unsupervised risk value, and the target propagation risk value to obtain a target risk value; determining the target risk level of spatial nodes based on the target risk value and multiple preset risk thresholds, and outputting control prompt information; thereby improving the safety rate and accuracy of risk management in confined space operations.

Description

Technical Field

[0001] This application relates to the field of confined space operation safety technology, and in particular to a risk control method, device, equipment and medium for confined space operations. Background Technology

[0002] Confined space operations are high-risk activities in industrial production and municipal operations. They refer to operations carried out in enclosed or semi-enclosed spaces with narrow entrances and exits, poor natural ventilation, and a tendency to accumulate toxic, harmful, flammable, and explosive gases, as well as oxygen deficiency. Confined spaces are prone to accumulating toxic and flammable gases and pose a risk of oxygen deficiency. There are many sudden accidents and a high rate of casualties in these operations. Therefore, precise risk management is required to improve the identification and assessment of hazardous factors in the work environment.

[0003] Existing risk management methods for confined space operations rely on expert experience, making it difficult to address unknown risks brought about by new processes and materials; risk assessments are based on static rules and cannot be dynamically adjusted according to real-time environmental data; common risks in similar spaces across different locations cannot be effectively shared, resulting in a waste of knowledge; and there is a lack of collaborative analysis capabilities for space-related risks.

[0004] Therefore, existing risk management methods for confined space operations suffer from low environmental adaptability and control accuracy. Summary of the Invention

[0005] This application provides a risk management method, apparatus, equipment, and medium for confined space operations, which addresses the problems of low environmental adaptability and control accuracy in existing risk management methods for confined space operations.

[0006] Firstly, this application provides a risk management method for confined space operations, the method comprising: Based on the nodes and their corresponding relationships in the operational scenario to be managed, a risk semantic association graph is constructed, and based on the risk semantic association graph, the neighborhood subgraph corresponding to the spatial node is determined. Based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, the target risk value is weighted and averaged to obtain the target cross-location risk value. Fields are extracted from the risk semantic association graph to obtain the target information set of spatial nodes. Conditional probability calculations are performed on the target information set and environment vector to obtain the unsupervised risk value of the target. Based on the preset connection relationship, the risk semantic association graph is traversed to obtain the connection matrix corresponding to the spatial node. Based on the connection matrix and the preset iteration threshold, the initial propagation risk value is iterated to obtain the target propagation risk value. The target cross-location risk value, the target unsupervised risk value, and the target propagation risk value are then weighted and summed to obtain the target risk value. Based on the target risk value and multiple preset risk thresholds, the target risk level of the spatial node is determined, and corresponding control prompts are output according to the target risk level.

[0007] In some embodiments of this application, determining the neighborhood subgraph corresponding to a spatial node based on a risk semantic association graph includes: Based on the node relationships in the risk semantic association graph, the associated nodes corresponding to the spatial nodes are determined, and the associated nodes are extended based on the preset number of extension steps to obtain multiple associated nodes and their respective node relationships. Based on the associated nodes and node relationships, determine the neighborhood subgraph corresponding to the spatial node.

[0008] In some embodiments of this application, a weighted average of the target risk value is performed based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold to obtain the target cross-location risk value, including: Based on the scenario type of the operation scenario to be managed, determine the corresponding target operation scenario, and based on the multiple neighborhood subgraphs corresponding to the target operation scenario, obtain the target neighborhood subgraph and its corresponding risk value. Feature extraction is performed on the neighborhood subgraph and the target neighborhood subgraph to obtain their respective feature vectors. The similarity between the feature vectors is then calculated to obtain the similarity between the neighborhood subgraph and the target neighborhood subgraph. By comparing the similarity value with the preset similarity threshold, a target similarity value greater than the preset similarity threshold is obtained, and the risk value corresponding to the target similarity value is determined as the target risk value. Based on the corresponding target similarity, determine the weight value corresponding to each target risk value, and calculate the weighted average of each target risk value according to its corresponding weight value to obtain the corresponding target cross-site risk value.

[0009] In some embodiments of this application, conditional probability calculations are performed on the target information set and the environment vector to obtain the target unsupervised risk value, including: Based on the sensor data corresponding to the spatial nodes, the corresponding environmental vector is determined, and conditional probability calculations are performed on the target information set and the environmental vector based on the preset probability distribution model to obtain the initial probability value; wherein, the probability distribution model is a Gaussian distribution model or a kernel density estimation model; Logarithmic calculation of the initial probability value yields the target unsupervised risk value.

[0010] In some embodiments of this application, based on preset connection relationships, the risk semantic association graph is traversed to obtain the connection matrix corresponding to the spatial nodes, including: Based on the preset connection relationship, the risk semantic association graph is traversed to obtain the associated spatial nodes corresponding to the spatial nodes, and the similarity between the spatial nodes and each associated spatial node is calculated to obtain the corresponding association similarity. Based on the association similarity, the corresponding association weights are determined, and based on the association weights, the corresponding connection matrix is ​​determined.

[0011] In some embodiments of this application, the initial propagation risk value is iterated based on the connection matrix and a preset iteration threshold to obtain the target propagation risk value, including: Based on the connection matrix and propagation coefficients, the initial propagation risk value is iterated to obtain the corresponding iterative propagation risk value. The iterative propagation risk value is then compared with the value of the preset iteration threshold to obtain the corresponding comparison result. If the comparison result shows that the iterative propagation risk value is less than the preset iterative threshold, then the iterative propagation risk value is determined to be the target propagation risk value. If the comparison result shows that the iterative propagation risk value is not less than the preset iterative threshold, then the iterative propagation risk value is iterated again until an iterative propagation risk value less than the preset iterative threshold is obtained, and the iterative propagation risk value is determined to be the target propagation risk value.

[0012] In some embodiments of this application, the target risk level of a spatial node is determined based on a target risk value and multiple preset risk thresholds, including: Based on a preset risk threshold, multiple corresponding risk intervals are determined, and the target risk value and the preset risk threshold are compared to obtain the corresponding target risk interval. Based on the risk level corresponding to each risk interval, determine the target risk level corresponding to the target risk interval.

[0013] Secondly, this application provides a risk management device for confined space operations, the device comprising: The construction module is used to build a risk semantic association graph based on the nodes and corresponding node relationships in the operation scenario to be managed, and to determine the neighborhood subgraphs corresponding to the spatial nodes based on the risk semantic association graph. The extraction module is used to perform a weighted average of the target risk value based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, to obtain the target cross-location risk value, and to extract fields from the risk semantic association graph to obtain a set of target information for spatial nodes. The calculation module is used to perform conditional probability calculation on the target information set and environment vector to obtain the target unsupervised risk value, and based on the preset connection relationship, to traverse the risk semantic association graph to obtain the connection matrix corresponding to the spatial node. The iteration module is used to iterate the initial propagation risk value according to the connection matrix and the preset iteration threshold to obtain the target propagation risk value, and to perform a weighted summation of the target cross-location risk value, the target unsupervised risk value and the target propagation risk value to obtain the target risk value. The determination module is used to determine the target risk level of spatial nodes based on the target risk value and multiple preset risk thresholds, and output corresponding control prompt information according to the target risk level.

[0014] Thirdly, this application provides a computer device, including: a processor, and a memory communicatively connected to the processor; The memory stores the instructions that the computer executes; The processor executes computer execution instructions stored in memory to implement the method of this application.

[0015] Fourthly, this application provides a computer-readable storage medium storing program code, which, when executed by a processor, is used to implement the method of this application.

[0016] Compared with existing technologies, the method in this application constructs a risk semantic association graph by combining various nodes and relationships within the operational scenario to be managed. It then extracts neighborhood subgraphs for each spatial node based on the association graph, thereby characterizing the entity composition, attribute features, and association logic of the finite space scenario, achieving a structured expression of scenario risk elements. By calculating the similarity between the neighborhood subgraph and the target neighborhood subgraph and using a threshold for numerical filtering, a weighted average is calculated on the similarities that meet the threshold requirements to obtain cross-site risk values. This effectively reuses historical risk patterns and hidden danger characteristics from similar scenarios, compensating for the current lack of risk identification dimensions in single scenarios and enabling cross-scenario risk experience transfer and prediction. Furthermore, by extracting attribute fields from the risk semantic association graph, a unique information set for each spatial node is obtained. Combined with real-time environmental vectors, conditional probability modeling is performed to calculate unsupervised risk values, realizing the inherent loop based on different finite spaces. By differentiating the characteristics of the environment and determining the degree of anomalies in real-time monitoring data, the problem of poor flexibility and adaptability of fixed evaluation standards is solved, enabling accurate identification of hidden environmental anomalies. Based on the actual connection relationships between spatial nodes, a semantic association graph is traversed to construct a connection matrix. Combined with an iterative convergence threshold, the initial risk data is iteratively calculated to solve for the propagation risk value, thereby simulating the risk transmission law under conditions such as physical connectivity and spatial proximity, and quantifying the risk linkage impact between spaces. The cross-site risk value, unsupervised risk value, and propagation risk value are weighted and fused to obtain a comprehensive target risk value. Based on multi-level thresholds, risk levels are divided and corresponding control prompts are output, realizing the integrated quantitative assessment and hierarchical differentiated control of multi-dimensional risks. This improves the comprehensiveness of risk identification in confined space operations, the accuracy of assessment, and the pertinence and proactivity of safety control, thereby preventing various safety accidents in confined spaces. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] Figure 1 A flowchart illustrating a risk management method for confined space operations provided in this application embodiment; Figure 2 A schematic diagram of a risk management method for confined space operations provided in this application embodiment; Figure 3 A schematic diagram of a risk management device for confined space operations provided in this application embodiment; Figure 4 This is a structural block diagram of an apparatus for performing a risk management method for confined space operations according to an embodiment of this application. Detailed Implementation

[0019] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0020] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart illustrating a risk management method for confined space operations provided in an embodiment of this application. Figure 1 As shown, the risk management method for this type of confined space operation may include the following steps: S110. Based on the nodes in the operation scenario to be controlled and their corresponding node relationships, construct a risk semantic association graph, and based on the risk semantic association graph, determine the neighborhood subgraph corresponding to the spatial node.

[0022] Among them, the operation scenarios to be managed refer to confined space operation sites that require risk assessment. Confined space operations pose significant safety risks and require accurate risk management. Examples include operation environments with a concentrated distribution of confined spaces such as municipal pipe networks, enclosed silos in factory areas, sewage pools, underground pipe corridors, and enclosed storage tanks. In practical applications, this scenario includes all confined space locations, supporting facilities, environmental conditions, work activities, hazard factors, and related ancillary units within the area. Examples include confined space locations such as manholes, underground pipe networks, and septic tanks; supporting facilities such as ventilation fans, underground ladders, and testing instruments; environmental conditions such as enclosed and unventilated space structures, humid and waterlogged environments, and high-temperature enclosed environments; work activities such as underground dredging, pipe network maintenance, and pool cleaning; hazard factors such as sewage sludge decomposition, organic matter decomposition, and pipe leaks; and related ancillary units such as surrounding connecting pipe networks, adjacent enclosed structures, and factory sewage systems. This integrates all risk-related entities, environmental data, work information, and related rules within the scenario to obtain a complete operation scenario to be managed, enabling subsequent risk assessment and management.

[0023] A node is a basic unit formed by digitally abstracting all independent physical entities, facilities, environmental elements, hazardous substances, and work behaviors related to risks within a confined space operation scenario. Each type of independent risk object corresponds to a separate node, which is used to transform complex on-site elements in reality into standardized graphical structure elements. Specifically, these include location nodes, such as schools, sewage treatment plants, and metallurgical enterprises; confined space nodes, such as silos, manholes, and septic tanks; material nodes, such as grain, mineral powder, sludge, and hazardous chemicals; hazard factor nodes, such as methane, hydrogen sulfide, dust, and oxygen deficiency; and work behavior nodes, such as entry, cleaning, welding, and maintenance.

[0024] Node relationships exist between different nodes and represent the logical correspondence between entities in terms of interaction, affiliation, spatial location, risk impact, or business association. Specifically, they include inclusion relationships, storage relationships, release relationships, exposure relationships, and influence relationships.

[0025] Risk semantic association graph refers to a directed attribute graph constructed from nodes and node relationships, thereby uniformly expressing multi-source heterogeneous information in confined space safety management. Specifically, for the operation scenario to be controlled, it uses various entity nodes as vertices and node relationships as connecting edges, integrating spatial topology, entity attribute information, risk association logic and business rule constraints to construct a structured knowledge graph.

[0026] Spatial nodes refer to enclosed / semi-enclosed work sites that exist independently within a scenario and possess typical characteristics of a finite space, such as manholes, septic tanks, sludge silos, sealed storage tanks, underground pipe trenches, and sewage pools. Each spatial node corresponds to an independent physical finite space and is the target object for risk calculation.

[0027] The neighborhood subgraph is a local network structure obtained by traversing the associated nodes and edges layer by layer outward from the complete risk semantic association graph with a single spatial node as the central origin. The subgraph retains only the central spatial node, the associated nodes within a specified range and their corresponding connections, and removes irrelevant nodes and redundant links from the entire domain, thereby representing the structural features, association attributes and local risk association information around the spatial node.

[0028] Based on this, by abstracting the location, limited space, materials, hazard factors and work behavior into a unified graph structure, a corresponding risk semantic association graph is obtained. For any spatial node, a subgraph containing all nodes with a shortest path distance of no more than k and the edges between them is determined with that node as the center. That is, the local network that can be reached in k steps is expanded outward to obtain the corresponding neighborhood subgraph.

[0029] S120. Based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, the target risk value is weighted and averaged to obtain the target cross-location risk value. Fields are extracted from the risk semantic association graph to obtain the target information set of spatial nodes.

[0030] Among them, the target neighborhood subgraph refers to the standardized neighborhood subgraph corresponding to each standard finite space node in other operation scenarios with the same category attributes. It can be stored in historical data and used to compare features with the neighborhood subgraph of the current spatial node to be evaluated. The higher the similarity between the two, the closer the risk semantics of the two spaces are, and the more worthy the historical risk value is to be referenced.

[0031] The preset similarity threshold is a pre-determined critical value used to determine whether there is similarity in risk knowledge between finite spaces, thereby determining whether the risk semantics between nodes are sufficiently similar, so as to carry out knowledge transfer. When the similarity calculation result of two neighborhood subgraphs is greater than or equal to the threshold, it is determined that the risk patterns of the two are highly consistent and that they have the conditions for risk transfer, so as to carry out subsequent calculations; if it is less than the threshold, it is determined that the two do not have reliable transfer conditions.

[0032] The target risk value refers to the known risk score of the target spatial node that has been verified in historical / similar scenarios. It can be the risk value determined by expert experience in risk labeling, industry standard risk level mapping value, historical accident statistical risk, or the output results of other mature risk assessment models.

[0033] Weighted average refers to the calculation process in which the similarity between the target neighborhood subgraph and the neighboring subgraphs is used as the weight coefficient. All similar spatial target risk values ​​that meet the threshold conditions are weighted and summed according to the similarity weight, and then divided by the total weight for normalization, so as to obtain a stable migration risk result.

[0034] The target cross-site risk value refers to the risk score determined by a spatial node based on the risk experience of other similar scenario nodes. It is a predictive risk value migrated from other similar locations. It represents the inherent risk level of the target space inferred from similar spatial experience in the absence of real-time perception data. It can be understood as: if there are risks in similar scenarios, the current space is likely to have similar risks as well.

[0035] Field extraction refers to the process of extracting information related to spatial nodes from a risk semantic association graph based on preset field categories, such as stored materials, hazardous gases, operational behaviors, and environmental conditions, to form structured data.

[0036] The target information set refers to the set of structured key attributes of the target's finite space nodes obtained from the risk semantic association graph through field extraction. It includes all information used for risk calculation, such as the target space's own attributes, location, stored materials, hazard factors, related operations, connected spaces, adjacent spaces, historical risks, and environmental conditions.

[0037] Based on this, the similarity between the neighborhood subgraph and the target neighborhood subgraph is calculated and compared with a preset similarity threshold. The weights are determined according to the similarity that meets the threshold requirements, so as to calculate the weighted average of the target risk value corresponding to the target neighborhood subgraph, obtain the target cross-site risk value, and extract fields from the risk semantic association graph to obtain the target information set of spatial nodes.

[0038] S130. Calculate the conditional probability of the target information set and the environment vector to obtain the unsupervised risk value of the target. Based on the preset connection relationship, traverse the risk semantic association graph to obtain the connection matrix corresponding to the spatial node.

[0039] Conditional probability calculation refers to calculating the probability of an event occurring given that a certain condition is true. In practical applications, by using the target information set as known conditions, probabilistic modeling is performed on the real-time environmental vector of the current finite space, thereby calculating the probability value that the current environmental state belongs to normal working conditions under the constraints of the space's inherent attributes, stored materials, hazard factors, and work behavior.

[0040] The target unsupervised risk value refers to the quantitative value of abnormal risk obtained by mapping the result of conditional probability calculation. It can characterize the degree of deviation and danger of the current environment from the historical normal pattern of the space under the premise of no expert rules, no labeled data, and no prior threshold. The greater the deviation, the higher the risk value. In practical applications, since the determination of the target unsupervised risk value does not depend on the given abnormal samples, but only on learning the normal conditional pattern, any information that deviates from the normal pattern, regardless of whether it has been trained, will be automatically judged as abnormal and quantified as risk. Therefore, the target unsupervised risk value can realize the discovery of unknown anomalies under unsupervised conditions, thereby solving the problem of identifying unknown risks and improving the adaptability to real-time environment and the accuracy of identification.

[0041] Preset connection relationships refer to the pre-determined association types used to determine whether there is a risk propagation relationship between two finite space nodes, thereby determining whether the space is connected, adjacent, of the same type, or has risk coupling. These include physical connectivity, spatial proximity, similar structural relationships, and risk impact relationships. If any of these relationships is satisfied, it is determined that a connection exists.

[0042] The matrix elements in the connection matrix are determined based on preset connection relationships. That is, if there are preset connection relationships between nodes, the connection weights between nodes can be further determined according to the preset connection relationships, so as to determine the connection weights as matrix elements in the connection matrix, thereby obtaining the connection matrix. Specifically, the connection weights can be obtained by setting or normalizing based on distance, connectivity strength, similarity, or empirical parameters. The connection matrix is ​​used to quantify the probability or strength of risk propagation from one space to another. This matrix is ​​directly used for the calculation of the risk propagation equation.

[0043] Based on this, by using the target information set as known conditions, conditional probability calculation is performed on the real-time environment vector in the current limited space to obtain the target unsupervised risk value. Based on the preset connection relationship, the risk semantic association graph is traversed to obtain the nodes that have connection relationships with the spatial nodes. Furthermore, the strength of the connection relationship can be quantified according to parameters such as the distance between nodes, connectivity strength, similarity, or empirical parameters to obtain the corresponding connection weights, thereby determining the connection matrix.

[0044] S140. Based on the connection matrix and the preset iteration threshold, iterate the initial propagation risk value to obtain the target propagation risk value, and then perform a weighted summation of the target cross-location risk value, the target unsupervised risk value, and the target propagation risk value to obtain the target risk value.

[0045] The preset iteration threshold is a pre-determined critical value used to determine whether the risk propagation iteration process has converged. It is used to control the number of iterations and convergence accuracy of the risk propagation equation. It is usually a very small positive number used to determine whether the difference between the risk vectors output by two adjacent iterations is small enough. When the difference between the two iteration results is less than the threshold, it is determined that the risk propagation has reached a stable equilibrium state and the iteration stops; otherwise, the iteration continues.

[0046] The initial propagation risk value refers to the original risk of a single point in a spatial node without considering the impact of spatial diffusion. It is the initial vector for risk propagation iteration and represents the inherent risk distribution of each spatial node before propagation begins. In practical applications, it can be the target cross-site risk value and the target unsupervised risk value, which together constitute the initial risk vector for propagation calculation. It can also include other inherent risk assessment values, such as static risk assessment, historical accident statistical risk, or expert experience assignment.

[0047] The target propagation risk value is a stable propagation risk value obtained by iterating multiple times from the initial propagation risk value based on the connection matrix until the difference between the propagation risk values ​​of two adjacent iterations is less than a preset iteration threshold. This value represents the comprehensive secondary risk value after the risk has fully diffused and synergistically affected between connected / adjacent / similar spaces, reflecting the chain risk, coupling risk, and diffusion risk between spaces. In practical applications, leakage in one space can lead to a synchronous increase in risk in connected spaces. Therefore, it is necessary to identify secondary risks, chain risks, and hidden risks based on the target propagation risk value, thereby solving the problem of traditional methods that only consider single points and not connections.

[0048] The target risk value is a final comprehensive quantitative risk score obtained by weighting and summing three types of risks—target cross-site risk value, target unsupervised risk value, and target propagation risk value—according to their credibility weights. This integrates experience transfer risk, real-time anomaly risk, and spatial propagation risk into a decision-making score.

[0049] Based on this, the initial propagation risk value is iteratively calculated according to the connection matrix. After multiple iterations, the target propagation risk value is obtained when the difference between the propagation risk values ​​of two adjacent iterations is less than the preset iteration threshold. The target cross-location risk value, the target unsupervised risk value, and the target propagation risk value are then weighted and summed to obtain the target risk value.

[0050] S150. Based on the target risk value and multiple preset risk thresholds, determine the target risk level of the spatial node, and output the corresponding control prompt information according to the target risk level.

[0051] Among them, the preset risk threshold is a predetermined numerical boundary point used to divide the continuously distributed comprehensive risk value into several discrete risk level intervals. For example, if the preset risk thresholds are A and B, and B is greater than A, then the intervals include interval 1, which is less than A, as a low risk level; interval 2, which is not less than A but less than B, as a medium risk level; and interval 3, which is not less than B, as a high risk level.

[0052] The target risk level refers to the risk level of a spatial node determined by comparing its target risk value with a preset threshold. In other words, it is the risk level corresponding to the interval where the target risk value is located. For example, if the target risk value is less than the preset risk threshold A, it corresponds to interval 1, and the risk level corresponding to interval 1 is low risk. Therefore, the target risk level corresponding to the target risk value can be determined to be low risk.

[0053] Control and management prompts refer to specific safety instructions, early warning notices, or equipment control signals triggered according to risk levels, used to guide on-site personnel or automated systems to take corresponding measures. In practical applications, different levels correspond to different control and management prompts. For example, a high-risk level indicates that the system is in a significantly abnormal or high-risk state, and operations should be immediately stopped and mandatory protective measures should be initiated. A medium-risk level indicates that there is a potential risk, and monitoring should be strengthened or operations should be restricted. A low-risk level is considered a normal risk range, and operations are permitted. The information can take various forms, including text prompts, audible and visual alarms, voice broadcasts, APP push notifications, platform pop-ups, and equipment linkage commands.

[0054] Based on this, by setting a risk threshold, multiple risk ranges and their corresponding risk levels are determined, thereby determining the target risk level corresponding to the target risk value, and outputting control prompt information corresponding to the target risk level.

[0055] Based on the feasible implementation of S110 described above, this application further provides a method for determining the neighborhood subgraph corresponding to a spatial node based on a risk semantic association graph, including: Based on the node relationships in the risk semantic association graph, the associated nodes corresponding to the spatial nodes are determined, and the associated nodes are extended based on the preset number of extension steps to obtain multiple associated nodes and their respective node relationships. Based on the associated nodes and node relationships, determine the neighborhood subgraph corresponding to the spatial node.

[0056] Among them, associated nodes refer to other nodes that are directly or indirectly connected to spatial nodes in the completed risk semantic association graph, that is, reachable through one or more edges; associated nodes can be any type of node, including material nodes, hazard factor nodes, work behavior nodes, other spatial nodes, and location nodes.

[0057] The preset extension step number, also known as the k-hop extension layer number or k-hop number, is the maximum number of outward traversal layers pre-set when constructing the neighborhood subgraph. It is used to limit how many layers of relationships can be extended outward from the target space node in the risk semantic association graph, thereby controlling the size of the neighborhood subgraph range and the completeness of information.

[0058] The association expansion is a process of recursively traversing the risk semantic association graph layer by layer outward from the target finite space node as the starting point, according to a preset number of expansion steps. It sequentially obtains the association nodes and their corresponding node relationships of each layer until the preset number of steps is reached, thus stopping the complete graph traversal process. Specifically, starting from the target space node, the first layer of association nodes and their relationship edges are obtained. Using the first layer of association nodes as the new starting point, the second layer of association nodes and their relationship edges are obtained. The above steps are repeated until the preset number of expansion steps is reached.

[0059] Based on this, in the risk semantic association graph, with the spatial node as the center, a subgraph containing all nodes with a shortest path distance of no more than k and the edges between them is determined. That is, the local network that can be reached in k steps is expanded outward, thereby obtaining the neighborhood subgraph corresponding to the spatial node.

[0060] Based on the feasible implementation of S120 described above, this application further provides a weighted average of the target risk value based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, to obtain the target cross-location risk value, including: Based on the scenario type of the operation scenario to be managed, determine the corresponding target operation scenario, and based on the multiple neighborhood subgraphs corresponding to the target operation scenario, obtain the target neighborhood subgraph and its corresponding risk value. Feature extraction is performed on the neighborhood subgraph and the target neighborhood subgraph to obtain their respective feature vectors. The similarity between the feature vectors is then calculated to obtain the similarity between the neighborhood subgraph and the target neighborhood subgraph. By comparing the similarity value with the preset similarity threshold, a target similarity value greater than the preset similarity threshold is obtained, and the risk value corresponding to the target similarity value is determined as the target risk value. Based on the corresponding target similarity, determine the weight value corresponding to each target risk value, and calculate the weighted average of each target risk value according to its corresponding weight value to obtain the corresponding target cross-site risk value.

[0061] Among them, the target operation scenario refers to the standard operation scenario that belongs to the same scenario type as the current operation scenario to be managed. It is a stored and verified similar operation scenario, and the risk knowledge of its internal spatial nodes has been verified to be reliable. It can be used as a reference object or knowledge source for the current scenario to be evaluated.

[0062] Risk value refers to a credible and quantifiable risk value obtained from historical statistics, model calculations, expert evaluations, or actual operation of the finite space associated with the corresponding neighborhood subgraph in the target operation scenario, thereby characterizing the inherent risk level of the reference space under standard conditions.

[0063] Based on this, and based on scenario type, target scenarios similar to the current scenario can be identified from existing mature operational scenarios. The credible risk values ​​of similar scenarios are then transferred to the current scenario in a weighted manner according to similarity. This enables the rapid acquisition of the initial risk of the current scenario without manual annotation or expert rules, and achieves risk knowledge sharing and reuse across locations and spaces. The similarity can be calculated using cosine similarity, Euclidean distance transformation, Mahalanobis distance, or other learned similarity functions.

[0064] Based on the feasible implementation of S130 described above, this application further provides a method for calculating the conditional probability of the target information set and the environment vector to obtain the unsupervised risk value of the target, including: Based on the sensor data corresponding to the spatial nodes, the corresponding environmental vector is determined, and conditional probability calculations are performed on the target information set and the environmental vector based on the preset probability distribution model to obtain the initial probability value; wherein, the probability distribution model is a Gaussian distribution model or a kernel density estimation model; Logarithmic calculation of the initial probability value yields the target unsupervised risk value.

[0065] The preset probability distribution model refers to a pre-determined model used to describe the statistical regularity of sensor observation data under normal operating conditions in a finite space. In practical applications, the model is trained based on sensor data under historical normal operating conditions so that the model can determine the normal distribution regularity of environmental parameters when the inherent properties, materials, hazard factors, and work behaviors of the space are fixed. This allows for further judgment on whether the current data conforms to the normal regularity under the environmental conditions. The model can be a Gaussian distribution model or a kernel density estimation model.

[0066] The initial probability value refers to the original conditional probability directly output after inputting the current environmental vector and the target information set using a preset probability distribution model. It represents the probability that the current set of sensor observation data belongs to the normal historical pattern under the fixed conditions of inherent attributes, materials, hazard factors, and operational behavior in the current limited space.

[0067] Based on this, by deploying multiple types of environmental sensors within each finite spatial node to continuously collect operational environmental parameters, the sampled values ​​from each sensor are spliced ​​together in a predetermined order to form a multidimensional observation vector, thus obtaining the corresponding environmental vector. Based on a preset probability distribution model, conditional probability calculations are performed on the target information set and the environmental vector to obtain an initial probability value, i.e., the probability that the environmental vector of the spatial node belongs to a normal environmental state under the currently associated target information. Furthermore, the initial probability value is logarithmically calculated to obtain the unsupervised risk value of the target.

[0068] Based on the feasible implementation of S130 described above, this application further provides a method for traversing a risk semantic association graph based on preset connection relationships to obtain a connection matrix corresponding to spatial nodes, including: Based on the preset connection relationship, the risk semantic association graph is traversed to obtain the associated spatial nodes corresponding to the spatial nodes, and the similarity between the spatial nodes and each associated spatial node is calculated to obtain the corresponding association similarity. Based on the association similarity, the corresponding association weights are determined, and based on the association weights, the corresponding connection matrix is ​​determined.

[0069] Among them, associated spatial nodes refer to other finite spatial nodes that are determined to have physical connectivity, spatial proximity, similar structure, and mutual risk influence after being traversed and filtered according to preset connection relationships, with the current spatial node as the center. These include physical connectivity relationships, such as direct connection through pipes, trenches, channels, or valves; spatial proximity relationships, such as the same area, the same pipe gallery, the same floor, or adjacent arrangement; similar structure relationships, such as similar finite spaces; and risk influence relationships, such as an increase in the risk of one party directly or indirectly affecting the safety of another party. In other words, other finite spaces that are connected to, close to, similar to, and can mutually infect the risk of the current space are associated spatial nodes.

[0070] Association weight is a quantitative connection strength value calculated based on the association similarity between spatial nodes and associated spatial nodes. It is used to represent the probability of risk propagation, the strength of the impact, and the tightness of coupling between two finite spaces. The weight can be set or normalized according to the distance, connectivity strength, similarity, or empirical parameters, and there is no limit to the specific calculation method.

[0071] Based on this, by identifying the associated spatial nodes that have a preset connection relationship with the spatial nodes, the similarity between the spatial nodes and each associated spatial node is calculated to obtain the corresponding association similarity. Based on the association similarity, the corresponding association weight is determined, and the association weight is determined as the corresponding matrix element to obtain the connection matrix.

[0072] Based on the feasible implementation of S140 described above, this application further provides a method for iterating the initial propagation risk value according to the connection matrix and a preset iteration threshold to obtain the target propagation risk value, including: Based on the connection matrix and propagation coefficients, the initial propagation risk value is iterated to obtain the corresponding iterative propagation risk value. The iterative propagation risk value is then compared with the value of the preset iteration threshold to obtain the corresponding comparison result. If the comparison result shows that the iterative propagation risk value is less than the preset iterative threshold, then the iterative propagation risk value is determined to be the target propagation risk value. If the comparison result shows that the iterative propagation risk value is not less than the preset iterative threshold, then the iterative propagation risk value is iterated again until an iterative propagation risk value less than the preset iterative threshold is obtained, and the iterative propagation risk value is determined to be the target propagation risk value.

[0073] The propagation coefficient is a coefficient between 0 and 1 that is pre-set in the risk propagation iterative calculation or can be obtained through adaptive learning. It is used to control the propagation intensity, diffusion speed and iterative stability of risk between associated spaces. It is a key adjustment parameter in the risk propagation equation used to balance the risk of the current space itself and the propagation risk from neighboring spaces.

[0074] The iterative propagation risk value refers to the temporary risk value output after each round of iterative calculation during the risk propagation process.

[0075] Therefore, in practical applications, the propagation coefficient is... The initial propagation risk value is The connection matrix is ​​A, and the iterative propagation risk value after the Kth iteration is... The iteration propagation risk value after the (k+1)th iteration is ,but: ; Thus, the higher the risk of neighboring nodes, the higher the risk of the node in this space, thereby determining the secondary risk of the connected region.

[0076] Based on the feasible implementation of S150 described above, this application further provides a method for determining the target risk level of a space node based on a target risk value and multiple preset risk thresholds, including: Based on a preset risk threshold, multiple corresponding risk intervals are determined, and the target risk value and the preset risk threshold are compared to obtain the corresponding target risk interval. Based on the risk level corresponding to each risk interval, determine the target risk level corresponding to the target risk interval.

[0077] Among them, the risk range refers to a numerical range that is continuous and non-overlapping, covering all possible risk values, formed by dividing multiple preset risk thresholds in order of magnitude.

[0078] Risk level refers to the safety level label corresponding to a risk range, such as low risk, medium risk, and high risk.

[0079] Based on this, by setting a risk threshold, multiple risk ranges and their corresponding risk levels are determined, thereby determining the target risk level corresponding to the target risk value, and outputting control prompt information corresponding to the target risk level.

[0080] Please refer to Figure 2 , Figure 2 A schematic diagram of a risk management method for confined space operations provided in this application embodiment; as shown. Figure 2 As shown, a finite set of spaces is generated by inputting the location, and a risk semantic graph is constructed. Three types of calculations are performed: similarity transfer, anomaly detection, and risk propagation. After risk fusion of the results, it is determined whether the comprehensive risk exceeds the threshold. If it exceeds the threshold, blocking, warning, or protection measures are implemented. If it does not exceed the threshold, the operation is allowed. Both cases are updated synchronously so that the updated information can be fed back to the risk semantic graph.

[0081] Based on the above steps, it can be seen that this application constructs a risk semantic association graph by combining various nodes and relationships within the operational scenario to be controlled, and extracts neighborhood subgraphs of each spatial node based on the association graph, thereby characterizing the entity composition, attribute features, and association logic of the limited spatial scenario, and realizing the structured expression of scenario risk elements; by calculating the similarity between the neighborhood subgraph and the target neighborhood subgraph and combining it with a threshold for numerical filtering, a weighted average calculation is performed on the similarity that meets the threshold requirements to obtain cross-site risk values, realizing the effective reuse of historical risk patterns and hidden danger characteristics of similar scenarios, making up for the current problem of insufficient risk identification dimensions in a single scenario, and realizing cross-scenario risk experience transfer and prediction; by extracting attribute fields from the risk semantic association graph, a unique information set of spatial nodes is obtained, and combined with real-time environmental vectors, conditional probability modeling is performed to calculate unsupervised risk values, realizing the inherent risk values ​​based on different limited spaces. By differentiating environmental characteristics and assessing the degree of anomalies in real-time monitoring data, this approach addresses the limitations of fixed evaluation standards in terms of flexibility and adaptability, enabling precise identification of hidden environmental anomalies. Based on the actual connections between spatial nodes, a semantic association graph is traversed to construct a connection matrix. Combined with an iterative convergence threshold, initial risk data is iteratively calculated to determine the propagation risk value, thereby simulating risk transmission patterns under conditions of physical connectivity and spatial proximity, and quantifying the interconnected impact of risks between spaces. By weightedly fusing cross-site risk values, unsupervised risk values, and propagation risk values, a comprehensive target risk value is obtained. Risk levels are then categorized based on multi-level thresholds, and corresponding control and management prompts are output. This achieves multi-dimensional risk fusion and quantitative assessment, as well as differentiated hierarchical control, improving the comprehensiveness of risk identification, the accuracy of assessment, and the targeted and proactive nature of safety management in confined space operations, thus preventing various safety accidents in confined spaces.

[0082] Figure 3 This is a schematic diagram of a risk management device for confined space operations provided in an embodiment of this application. Figure 3 As shown, the risk management device for this type of confined space operation includes: a construction module, an extraction module, a calculation module, an iteration module, and a determination module; wherein: The construction module is used to build a risk semantic association graph based on the nodes and corresponding node relationships in the operation scenario to be managed, and to determine the neighborhood subgraphs corresponding to the spatial nodes based on the risk semantic association graph. The extraction module is used to perform a weighted average of the target risk value based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, to obtain the target cross-location risk value, and to extract fields from the risk semantic association graph to obtain a set of target information for spatial nodes. The calculation module is used to perform conditional probability calculation on the target information set and environment vector to obtain the target unsupervised risk value, and based on the preset connection relationship, to traverse the risk semantic association graph to obtain the connection matrix corresponding to the spatial node. The iteration module is used to iterate the initial propagation risk value according to the connection matrix and the preset iteration threshold to obtain the target propagation risk value, and to perform a weighted summation of the target cross-location risk value, the target unsupervised risk value and the target propagation risk value to obtain the target risk value. The determination module is used to determine the target risk level of spatial nodes based on the target risk value and multiple preset risk thresholds, and output corresponding control prompt information according to the target risk level.

[0083] In this embodiment of the application, the construction module can also be specifically used for: Based on the node relationships in the risk semantic association graph, the associated nodes corresponding to the spatial nodes are determined, and the associated nodes are extended based on the preset number of extension steps to obtain multiple associated nodes and their respective node relationships. Based on the associated nodes and node relationships, determine the neighborhood subgraph corresponding to the spatial node.

[0084] In this embodiment of the application, the extraction module can also be specifically used for: Based on the scenario type of the operation scenario to be managed, determine the corresponding target operation scenario, and based on the multiple neighborhood subgraphs corresponding to the target operation scenario, obtain the target neighborhood subgraph and its corresponding risk value. Feature extraction is performed on the neighborhood subgraph and the target neighborhood subgraph to obtain their respective feature vectors. The similarity between the feature vectors is then calculated to obtain the similarity between the neighborhood subgraph and the target neighborhood subgraph. By comparing the similarity value with the preset similarity threshold, a target similarity value greater than the preset similarity threshold is obtained, and the risk value corresponding to the target similarity value is determined as the target risk value. Based on the corresponding target similarity, determine the weight value corresponding to each target risk value, and calculate the weighted average of each target risk value according to its corresponding weight value to obtain the corresponding target cross-site risk value.

[0085] In this embodiment of the application, the calculation module can also be specifically used for: Based on the sensor data corresponding to the spatial nodes, the corresponding environmental vector is determined, and conditional probability calculations are performed on the target information set and the environmental vector based on the preset probability distribution model to obtain the initial probability value; wherein, the probability distribution model is a Gaussian distribution model or a kernel density estimation model; Logarithmic calculation of the initial probability value yields the target unsupervised risk value.

[0086] In this embodiment of the application, the calculation module can also be specifically used for: Based on the preset connection relationship, the risk semantic association graph is traversed to obtain the associated spatial nodes corresponding to the spatial nodes, and the similarity between the spatial nodes and each associated spatial node is calculated to obtain the corresponding association similarity. Based on the association similarity, the corresponding association weights are determined, and based on the association weights, the corresponding connection matrix is ​​determined.

[0087] In this embodiment of the application, the iteration module can also be specifically used for: Based on the connection matrix and propagation coefficients, the initial propagation risk value is iterated to obtain the corresponding iterative propagation risk value. The iterative propagation risk value is then compared with the value of the preset iteration threshold to obtain the corresponding comparison result. If the comparison result shows that the iterative propagation risk value is less than the preset iterative threshold, then the iterative propagation risk value is determined to be the target propagation risk value. If the comparison result shows that the iterative propagation risk value is not less than the preset iterative threshold, then the iterative propagation risk value is iterated again until an iterative propagation risk value less than the preset iterative threshold is obtained, and the iterative propagation risk value is determined to be the target propagation risk value.

[0088] In this embodiment of the application, the determining module can also be specifically used for: Based on a preset risk threshold, multiple corresponding risk intervals are determined, and the target risk value and the preset risk threshold are compared to obtain the corresponding target risk interval. Based on the risk level corresponding to each risk interval, determine the target risk level corresponding to the target risk interval.

[0089] Figure 4 This is a schematic diagram of the structure of an apparatus for performing a risk management method for confined space operations according to an embodiment of this application. Figure 4 As shown, the device includes: The device may include one or more processors with processing cores, one or more computer-readable storage media such as memory, communication components, etc. The processor, memory, and communication components are connected via a bus.

[0090] In the specific implementation process, at least one processor executes computer execution instructions stored in memory, causing at least one processor to perform the risk management method for confined space operations as described above.

[0091] The specific implementation process of the processor can be found in the above method embodiments, and its implementation principle and technical effect are similar, so it will not be repeated here.

[0092] Furthermore, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0093] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0094] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0095] In some embodiments, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps in any of the above-described risk management methods for confined space operations.

[0096] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0097] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0098] Therefore, embodiments of this application provide a computer-readable storage medium storing multiple lines of program code that can be loaded by a processor to execute steps in any of the risk management methods for confined space operations provided in embodiments of this application.

[0099] The storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0100] According to one aspect of this application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium.

[0101] Since the instructions stored in the storage medium can execute the steps in any of the risk management methods for confined space operations provided in the embodiments of this application, the beneficial effects that any of the risk management methods for confined space operations provided in the embodiments of this application can achieve can be realized. For details, please refer to the previous embodiments, which will not be repeated here.

[0102] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope of this application is indicated by the appended claims.

[0103] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A risk management method for confined space operations, characterized in that, The method includes: Based on the nodes and their corresponding relationships in the operational scenario to be managed, a risk semantic association graph is constructed, and based on the risk semantic association graph, the neighborhood subgraph corresponding to the spatial node is determined. Based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, the target risk value is weighted and averaged to obtain the target cross-location risk value. Fields are extracted from the risk semantic association graph to obtain the target information set of the spatial node. Conditional probability calculations are performed on the target information set and environment vector to obtain the target unsupervised risk value. Based on preset connection relationships, the risk semantic association graph is traversed to obtain the connection matrix corresponding to the spatial node. Based on the connection matrix and the preset iteration threshold, the initial propagation risk value is iterated to obtain the target propagation risk value. The target cross-location risk value, the target unsupervised risk value, and the target propagation risk value are then weighted and summed to obtain the target risk value. Based on the target risk value and multiple preset risk thresholds, the target risk level of the space node is determined, and corresponding control prompt information is output according to the target risk level.

2. The method according to claim 1, characterized in that, The step of determining the neighborhood subgraph corresponding to the spatial node based on the risk semantic association graph includes: Based on the node relationships in the risk semantic association graph, the associated nodes corresponding to the spatial nodes are determined, and the associated nodes are extended based on a preset number of extension steps to obtain multiple associated nodes and their respective node relationships; Based on the associated nodes and the node relationships, the neighborhood subgraph corresponding to the spatial node is determined.

3. The method according to claim 1, characterized in that, The step of calculating a weighted average of the target risk value based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold to obtain the target cross-location risk value includes: Based on the scenario type of the operation scenario to be managed, the corresponding target operation scenario is determined, and based on the multiple neighborhood subgraphs corresponding to the target operation scenario, the target neighborhood subgraph and its corresponding risk value are obtained. Feature extraction is performed on the neighborhood subgraph and the target neighborhood subgraph to obtain their respective feature vectors, and similarity calculation is performed on the feature vectors to obtain the similarity between the neighborhood subgraph and the target neighborhood subgraph. By comparing the value of the similarity with the preset similarity threshold, a target similarity greater than the preset similarity threshold is obtained, and the risk value corresponding to the target similarity is determined to be the target risk value; Based on the corresponding target similarity, the weight value corresponding to each target risk value is determined, and a weighted average is calculated for each target risk value based on its corresponding weight value to obtain the corresponding target cross-location risk value.

4. The method according to claim 1, characterized in that, The step of calculating the conditional probability of the target information set and the environment vector to obtain the unsupervised risk value of the target includes: Based on the sensor data corresponding to the spatial node, the corresponding environmental vector is determined, and based on a preset probability distribution model, conditional probability calculation is performed on the target information set and the environmental vector to obtain an initial probability value; wherein, the probability distribution model is a Gaussian distribution model or a kernel density estimation model; The initial probability value is logarithmically calculated to obtain the target unsupervised risk value.

5. The method according to claim 1, characterized in that, The step of traversing the risk semantic association graph based on preset connection relationships to obtain the connection matrix corresponding to the spatial nodes includes: Based on the preset connection relationship, the risk semantic association graph is traversed to obtain the associated spatial nodes corresponding to the spatial nodes, and the similarity between the spatial nodes and each of the associated spatial nodes is calculated to obtain the corresponding association similarity. Based on the association similarity, the corresponding association weight is determined, and based on the association weight, the corresponding connection matrix is ​​determined.

6. The method according to claim 1, characterized in that, The step of iterating the initial propagation risk value based on the connection matrix and a preset iteration threshold to obtain the target propagation risk value includes: Based on the connection matrix and propagation coefficient, the initial propagation risk value is iterated to obtain the corresponding iterative propagation risk value, and the iterative propagation risk value is compared with the value of the preset iteration threshold to obtain the corresponding comparison result. If the comparison result shows that the iterative propagation risk value is less than the preset iteration threshold, then the iterative propagation risk value is determined to be the target propagation risk value. If the comparison result indicates that the iterative propagation risk value is not less than the preset iteration threshold, then the iterative propagation risk value is iterated again until the iterative propagation risk value is less than the preset iteration threshold, and the iterative propagation risk value is determined to be the target propagation risk value.

7. The method according to claim 1, characterized in that, The step of determining the target risk level of the space node based on the target risk value and multiple preset risk thresholds includes: Based on the preset risk threshold, multiple corresponding risk intervals are determined, and the target risk value and the preset risk threshold are compared to obtain the corresponding target risk interval. The target risk level corresponding to the target risk interval is determined based on the risk level corresponding to each of the risk intervals.

8. A risk management device for confined space operations, characterized in that, The device includes: The construction module is used to construct a risk semantic association graph based on the nodes and corresponding node relationships in the operation scenario to be managed, and to determine the neighborhood subgraph corresponding to the spatial node based on the risk semantic association graph. The extraction module is used to perform a weighted average of the target risk value based on the similarity between the neighborhood subgraph and the target neighborhood subgraph and a preset similarity threshold, to obtain the target cross-location risk value, and to extract fields from the risk semantic association graph to obtain the target information set of the spatial node; The calculation module is used to perform conditional probability calculation on the target information set and the environment vector to obtain the target unsupervised risk value, and to traverse the risk semantic association graph based on the preset connection relationship to obtain the connection matrix corresponding to the spatial node. The iteration module is used to iterate the initial propagation risk value according to the connection matrix and the preset iteration threshold to obtain the target propagation risk value, and to perform a weighted summation of the target cross-location risk value, the target unsupervised risk value and the target propagation risk value to obtain the target risk value. The determination module is used to determine the target risk level of the space node based on the target risk value and multiple preset risk thresholds, and output corresponding control prompt information based on the target risk level.

9. A computer device, characterized in that, include: One or more processors; Memory; One or more programs, wherein the one or more programs are stored in memory and configured to be executed by one or more processors, the one or more programs being configured to perform the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code that can be called by a processor to perform the method as described in any one of claims 1 to 7.

Images