Fire risk prevention and control and operation decision system for old-age homes based on knowledge graph

Abstract

The application discloses a pension agency fire risk prevention and control and operation decision system based on a knowledge graph, relates to the technical field of fire data processing and management of a pension agency, and obtains a central risk node by acquiring initial fire source position data and mapping; a first risk influence subgraph with the central risk node as the center is acquired; a passing risk coefficient is calculated through cyclic iteration, the first iteration is used for correcting a basic passing coefficient according to the difference in spatial geometric characteristics to calculate an environmental risk coefficient, the non-first iteration is used for correcting the passing risk coefficient of the last iteration according to the difference ratio of the environmental risk coefficient, and the iteration is continued until the convergence condition is met; and a prevention and control decision report is output according to the converged passing risk coefficient. The application realizes dynamic iterative evaluation and self-adaptive path decision of fire risk, and through iterative propagation and dynamic correction of the risk gradient, the evacuation path can respond to the evolution law of the fire environment in real time.

Description

Technical Field

[0001] This invention relates to the field of fire data processing and management technology for elderly care institutions, and more specifically, this application relates to a fire risk prevention and control and operational decision-making system for elderly care institutions based on knowledge graphs. Background Technology

[0002] In the field of fire risk prevention and control in elderly care facilities, existing decision-making systems typically rely on static building information and discrete sensor alarm points for path planning. When a fire alarm signal is triggered, a pre-set evacuation route map or shortest path algorithm is generally used to plan an escape route from the current location to a safe exit. However, limited by the fixed topological relationship of the building structure and the spatial location of the alarm points, these systems fail to fully consider the continuous distribution pattern of smoke concentration within the building space and its changes over time during the fire's development.

[0003] Existing systems struggle to quantify the risk gradient relationships between different areas. Furthermore, environmental risk perception and path planning decisions are statically correlated within existing systems. The system only performs a single risk assessment and path calculation at the initial moment based on the fire source location; the lack of a subsequent iterative update mechanism means that access risks cannot be dynamically adjusted to keep pace with the real-time evolution of the fire environment. When fire spreads or smoke diffusion paths change, the pre-defined access risks may become invalid, yet the system continues to use the original decision results, easily leading personnel to areas where the risk is increasing.

[0004] Existing technologies lack the accuracy to construct dynamic maps of access risk relationships that continuously approximate the real fire scene environment, driven by multi-dimensional data. Summary of the Invention

[0005] To address the aforementioned technical issues, this technical solution provides a knowledge graph-based fire risk prevention and operational decision-making system for elderly care institutions, resolving the problems mentioned in the background section.

[0006] In a first aspect, embodiments of this application provide a knowledge graph-based fire risk prevention and operational decision-making system for elderly care institutions, comprising: a data acquisition module: used to acquire initial fire source location data and map it from a pre-trained institutional knowledge graph according to preset mapping rules to obtain a central risk node; a risk impact subgraph acquisition module: used to acquire a first risk impact subgraph centered on the central risk node, wherein each risk node in the first risk impact subgraph contains spatial geometric features, and the edge weight of each connected edge of each risk node is initialized as a basic passage coefficient; and an iterative execution module: used to perform cyclic iteration: acquire the environmental risk coefficient of the previous iteration, and calculate the central risk node, the first risk subgraph, and the risk impact subgraph respectively. The environmental risk coefficients of all adjacent risk nodes within the risk impact subgraph are compared with the environmental risk coefficients of the previous iteration to obtain the adjacent iteration difference of the central risk node and the adjacent iteration difference of adjacent risk nodes. The ratio of the two is calculated and the passage risk coefficient of the previous iteration is corrected accordingly to obtain the passage risk coefficient of the current iteration. Iteration judgment module: used to determine the end of the iteration if the difference of the passage risk coefficients of adjacent iterations does not exceed the preset risk difference threshold, or if the current iteration number exceeds the maximum iteration number. The passage risk coefficient after the iteration ends is obtained. Output module: used to output the first prevention and control decision report based on the passage risk coefficient after the iteration ends.

[0007] Secondly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions.

[0008] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0009] 1. This invention calculates the environmental risk coefficient and corrects the basic access coefficient by acquiring the differences in spatial geometric characteristics between the central risk node and other risk nodes within the first risk impact sub-map, thus transforming discrete sensor alarm signals into continuous risk transmission relationships between building space nodes. This processing method overcomes the limitations of existing technologies that rely solely on threshold triggers for discrete risk assessment, enabling precise characterization of the gradual change in flue gas concentration in both vertical and horizontal directions, providing a refined risk distribution basis for subsequent path decision-making.

[0010] 2. This invention employs an iterative approach, calculating the environmental risk coefficient based on spatial geometric differences in the first iteration, and then adjusting the access risk coefficient based on the ratio of the environmental risk coefficient from the previous iteration to that of adjacent nodes. This mechanism ensures that the propagation of risk between building space nodes is no longer a unidirectional, static assignment, but rather a dynamic interaction and adaptive adjustment of risk distribution through iterative calculation of gradient ratios, guaranteeing that the access risk coefficient can evolve in real time following the changing patterns of the fire environment.

[0011] 3. This invention determines the iteration termination time by judging whether the number of times the difference between the current iteration's access risk coefficient and the previous iteration's access risk coefficient exceeds a preset risk difference threshold, and whether the cumulative number of iterations exceeds the maximum number of iterations. This convergence mechanism avoids the computational resource consumption caused by infinite iteration, while ensuring timely output of decision results when the risk distribution tends to stabilize, and capturing fluctuation characteristics by continuously exceeding the limit when the risk changes rapidly. This makes the first prevention and control decision report output by the output module based on the access risk coefficient after the iteration ends both real-time and reliable. Attached Figure Description

[0012] Figure 1 A schematic diagram of the structure of a knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions provided in this application embodiment;

[0013] Figure 2 This is a schematic diagram of the logical flow of a knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions, provided in an embodiment of this application. Detailed Implementation

[0014] This application provides a knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions, which solves the technical problem in the prior art where the accuracy of constructing a dynamic graph of access risk relationships that continuously approximates the real fire scene environment under multi-dimensional data-driven conditions is insufficient.

[0015] In the field of fire risk prevention and control in elderly care institutions, existing decision-making systems typically rely on static building information and discrete sensor alarm points for path planning. When a fire alarm signal is triggered, a pre-set evacuation route map or shortest path algorithm is generally used to plan an escape route from the current location to a safe exit. However, this approach is limited by the fixed topological relationship of the building structure and the spatial location of the alarm points, failing to fully consider the continuous distribution pattern of smoke concentration within the building space and its changes over time during the fire's development. Furthermore, environmental risk perception and path planning decisions are statically correlated in existing systems. The system only performs a risk assessment and path calculation once at the initial moment based on the fire source location. The lack of a subsequent iterative update mechanism means that access risks cannot be dynamically adjusted to keep pace with the real-time evolution of the fire environment. When the fire spreads or the smoke diffusion path changes, the pre-determined access risks may become invalid, but the system continues to use the original decision results, easily leading personnel to areas where the risk is increasing.

[0016] To address the aforementioned issues, this solution first acquires initial fire source location data and maps it from a pre-trained institutional knowledge graph according to preset mapping rules to obtain a central risk node. This process transforms the fire source alarm signal into a computable spatial entity in the knowledge graph. Subsequently, a first risk impact subgraph is obtained centered on this central risk node. Each risk node in the subgraph contains spatial geometric features, and the edge weight of each connected edge is initialized as a basic access coefficient, thereby constructing the spatial topology structure to be analyzed. Based on this, this scheme introduces a cyclical iterative mechanism: In the first iteration, the environmental risk coefficient is calculated based on the difference in spatial geometric characteristics between the central risk node and other risk nodes in the subgraph, and the basic access coefficient is corrected accordingly to obtain the access risk coefficient. This process transforms the spatial influence of the fire source location into a quantified risk value. The environmental risk coefficient of the previous iteration is obtained, and the environmental risk coefficients of the central risk node and all adjacent risk nodes in the first risk impact subgraph are calculated separately. The difference between these environmental risk coefficients and the environmental risk coefficients of the previous iteration is obtained to obtain the adjacent iteration difference of the central risk node and the adjacent iteration difference of adjacent risk nodes. The ratio of these two is calculated, and the access risk coefficient of the previous iteration is corrected accordingly to obtain the access risk coefficient of the current iteration. This process realizes the dynamic propagation and interaction of risk between adjacent nodes through the risk gradient ratio. If the difference between the access risk coefficient of the current iteration and the access risk coefficient of the previous iteration exceeds a preset risk difference threshold for more than a preset number of times, or if the current iteration exceeds the maximum number of iterations, the iteration is considered to have ended, and the access risk coefficient after the iteration ends is obtained. Finally, the first prevention and control decision report is output based on the access risk coefficient after the iteration ends. Through the above processing, this solution transforms discrete fire source alarm information into a continuous risk transmission relationship between building space nodes, and realizes the dynamic evolution of risk distribution through iterative calculation of gradient ratio. This enables the access risk coefficient to be adaptively adjusted in real time to follow the evolution of the fire environment, solving the problem of the disconnect between access risk and fire evolution in the existing technology.

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] like Figure 1The diagram shown is a structural schematic of a knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions provided in this application embodiment. It includes: a data acquisition module for acquiring initial fire source location data and mapping it from a pre-trained institutional knowledge graph according to preset mapping rules to obtain a central risk node; a risk impact subgraph acquisition module for acquiring a first risk impact subgraph centered on the central risk node, where each risk node in the first risk impact subgraph contains spatial geometric features, and the edge weight of each connected edge of each risk node is initialized as a basic passage coefficient; and an iterative execution module for performing loop iterations: acquiring the environmental risk coefficient of the previous iteration, and calculating the central risk node, the... The environmental risk coefficients of all adjacent risk nodes within a risk impact subgraph are compared with the environmental risk coefficients of the previous iteration to obtain the adjacent iteration difference of the central risk node and the adjacent iteration difference of adjacent risk nodes. The ratio of the two is calculated and the passage risk coefficient of the previous iteration is corrected accordingly to obtain the passage risk coefficient of the current iteration. Iteration judgment module: used to determine the end of the iteration if the difference between the passage risk coefficients of adjacent iterations does not exceed the preset risk difference threshold, or if the current iteration number exceeds the maximum iteration number. The passage risk coefficient after the iteration ends is obtained. Output module: used to output the first prevention and control decision report based on the passage risk coefficient after the iteration ends.

[0019] Figure 2 This is a schematic diagram of the logical flow of a knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions, provided in an embodiment of this application.

[0020] If it is the first iteration, the environmental risk coefficient is calculated based on the difference in spatial geometric characteristics between the central risk node and other risk nodes in the first risk impact subgraph, and the basic access coefficient is corrected accordingly to obtain the access risk coefficient.

[0021] Basic passage coefficient: A pre-stored base value for each connected edge, reflecting the ease with which people can pass through that edge under normal circumstances. It is set based on factors such as passage length, width, and slope, with a typical range of 0.1 to 1.0. Typical values ​​are preset according to the passage type, for example, 0.5 for a straight corridor and 0.8 for a staircase. An example formula for the basic passage coefficient is as follows: ,in, For channel length, For channel width, The slope coefficient is... and These are empirical weighting coefficients, with values ​​ranging from 0.1 to 0.5 and from 0.2 to 0.6, respectively. The typical value range is 0.1 to 1.0. The smaller the value, the smoother the passage, and the larger the value, the more difficult the passage.

[0022] A preset risk difference threshold is used to determine whether the passage risk coefficient tends to stabilize between two adjacent iterations. The threshold is set based on the system's sensitivity requirements to risk changes. Specifically, the threshold range is determined according to the rate of change of smoke concentration in actual fire simulation data or historical fire cases.

[0023] The typical value range is 0.01~0.05, with a typical value of 0.02, indicating that when the change in the passage risk coefficient is less than 2%, it is considered close to convergence;

[0024] The preset threshold number of iterations is used to capture the stable trend after risk fluctuations. The value range is 2 to 5, with a typical value of 3, which means that the risk coefficient changes exceed the threshold in 3 consecutive iterations before it is judged as non-convergence.

[0025] Preset frequency threshold: The threshold for the number of times the risk difference threshold is exceeded consecutively, used to capture stability after fluctuations. The typical value range is 2~5, with a typical value of 3.

[0026] Maximum number of iterations: A hard limit to prevent infinite loops. It is set according to computing resources, with a typical range of 20 to 50 and a typical value of 30.

[0027] This solution maps the location of the fire source to knowledge graph nodes, extracts a risk impact subgraph, and uses iterative loops to dynamically calculate the access risk coefficient, achieving continuous quantification and adaptive propagation of fire risk. In the specific scenario of elderly care facilities, it can reflect the impact of smoke diffusion and fire spread on evacuation routes in real time, solving the problems of static risk assessment and disconnection from fire evolution in existing technologies, and providing safer and more reliable evacuation decisions for the elderly.

[0028] Furthermore, the specific process for determining the central risk node is as follows: read the spatial geometric features of each risk node from the institutional knowledge graph. The spatial geometric features include at least the three-dimensional coordinates corresponding to the risk node; calculate the Euclidean distance between the initial fire source location data and the three-dimensional coordinates of each risk node as the spatial distance value; and determine the risk node with the smallest spatial distance value as the central risk node.

[0029] In this embodiment, the spatial geometric features of each risk node are read from the institutional knowledge graph. The spatial geometric features include at least the three-dimensional coordinates corresponding to the risk node. Here, the three-dimensional coordinates refer to the actual location coordinates of the risk node in the building space, usually in meters. The range of values ​​is determined according to the building size. For example, for common elderly care institutions, the coordinate range may be within 1 meter.

[0030] The Euclidean distance between the initial fire source location data and the three-dimensional coordinates of each risk node is calculated as the spatial distance value. Euclidean distance is a conventional method for calculating geometric distance.

[0031] The risk node with the smallest spatial distance value is identified as the central risk node. This step involves comparing all calculated spatial distance values ​​and selecting the node corresponding to the minimum value, without the need for a preset threshold.

[0032] This embodiment uses three-dimensional coordinates in spatial geometry to calculate Euclidean distance and determines the risk node closest to the fire source as the initial center, thereby accurately mapping discrete fire alarm signals to specific entities in the knowledge graph, providing an accurate starting point for subsequent risk propagation analysis.

[0033] Furthermore, the specific process of obtaining the first risk impact subgraph is as follows: taking the central risk node as the search starting point, using the breadth-first search algorithm, and taking the current maximum graph search depth as the traversal termination condition, all risk nodes whose search depth does not exceed the current maximum graph search depth are visited layer by layer; the visited risk nodes and all connected edges connecting the risk nodes are extracted to form the first risk impact subgraph.

[0034] In this embodiment, the central risk node is used as the starting point for the search. A breadth-first search algorithm is employed, with the current maximum graph search depth as the termination condition. All risk nodes whose search depth does not exceed the current maximum graph search depth are visited layer by layer. Breadth-first search is a classic algorithm in graph theory used to traverse all reachable nodes in a graph. The current maximum graph search depth is a preset parameter used to limit the search range. Its setting is based on the building size of the elderly care facility and the effective distance for fire risk propagation. A typical value range is 3 to 10 floors; for a typical multi-story elderly care facility, 5 floors is suitable.

[0035] Extract the visited risk nodes and all connected edges linking them to form the first risk impact subgraph. This step combines the traversal results into a subgraph, which serves as the basis for subsequent iterative analysis.

[0036] This embodiment starts from the central risk node and uses breadth-first search to extract the risk impact subgraph within a preset depth. This ensures that the potentially affected areas are covered, while avoiding the waste of resources caused by full-graph computation, and provides a reasonable spatial range for subsequent iterations.

[0037] Furthermore, the specific calculation process of the environmental risk coefficient is as follows: Obtain the absolute value of the difference between the vertical projection height values ​​of the spatial geometric features of the central risk node and any other risk node in the first risk impact sub-map, as the vertical distance value; obtain the absolute value of the difference between the horizontal projection distance values ​​of the spatial geometric features of the central risk node and any other risk node in the first risk impact sub-map, as the horizontal distance value; weight the vertical distance value according to a first preset weight, and weight the horizontal distance value according to a second preset weight, and use the weighted sum as the spatial risk transmission coefficient of the other risk node relative to the central risk node; obtain the preset baseline environmental risk value and spatial risk transmission coefficient of the other risk node to generate the initial environmental risk coefficient of the other risk node relative to the central risk node; sum the initial environmental risk coefficients of all other risk nodes relative to the central risk node to obtain the environmental risk coefficient of the central risk node.

[0038] In this embodiment, the absolute value of the difference between the vertical projection height values ​​of the spatial geometric features of the central risk node and any other risk node in the first risk impact sub-graph is obtained as the vertical distance value. The vertical projection height value is the floor height coordinate of the node, and the absolute value of the difference reflects the vertical distance between the two nodes, in meters. The value range is determined by the building floor height. For example, if the typical floor height is 3 meters, the difference may be between 0 and 30 meters.

[0039] Obtain the absolute value of the difference between the horizontal projected distance values ​​of the spatial geometric features of the central risk node and any other risk node in the first risk impact sub-graph, and use this as the horizontal distance value. The horizontal distance value refers to the straight-line distance between the two nodes on the plane, in meters, and its range is determined by the building plan dimensions, for example, 0 to 50 meters.

[0040] Vertical distance values ​​are weighted according to a first preset weight, and horizontal distance values ​​are weighted according to a second preset weight. The weighted sum is used as the spatial risk transmission coefficient of other risk nodes relative to the central risk node. The first and second preset weights are pre-set constants used to balance the influence of vertical and horizontal directions on risk propagation. The setting is based on the physical characteristics of smoke diffusion in a fire: vertical propagation is faster due to the chimney effect, therefore the vertical weight is usually greater than the horizontal weight. Typical values: first preset weight is 0.7, second preset weight is 0.3, and the sum is 1.

[0041] The system obtains the preset baseline environmental risk value and spatial risk transfer coefficient for each other risk node, and generates the initial environmental risk coefficient for that other risk node relative to the central risk node. The baseline environmental risk value is the inherent basic risk value of each risk node, reflecting the fire sensitivity of the area itself. For example, areas such as kitchens and electrical rooms have higher baseline values, ranging from 0 to 1. The generation method typically involves multiplying the baseline value by the spatial risk transfer coefficient.

[0042] The environmental risk coefficient of the central risk node is obtained by summing the initial environmental risk coefficients of all other risk nodes relative to the central risk node. This summation reflects the combined risk impact of all neighboring nodes on the central node.

[0043] This embodiment comprehensively considers the impact of vertical and horizontal distances on risk propagation, quantifies spatial geometric differences into risk transmission coefficients using weighted coefficients, and combines these with the baseline risk values ​​of each node to finally calculate the environmental risk coefficient of the central node. This process transforms the spatial influence of the fire source into a quantifiable numerical value, laying the foundation for subsequent risk gradient analysis.

[0044] Furthermore, the iterative execution module is also used to: obtain several boundary nodes marked as preset safety exits in the first risk impact subgraph; in each iteration, monitor whether the environmental risk coefficient of the boundary nodes exceeds the preset exit blocking threshold; the preset mapping rules include the maximum graph search depth determined based on the maximum difference between the initial fire source location data and the spatial geometric features of the risk nodes; when the environmental risk coefficient of any boundary node exceeds the preset exit blocking threshold, remove the boundary node from the first risk impact subgraph, and reduce and correct the current maximum graph search depth according to the difference between the environmental risk coefficient and the preset exit blocking threshold to obtain the first corrected maximum graph search depth; call the risk impact subgraph acquisition module to re-acquire the updated first risk impact subgraph with the central risk node as the center and the first corrected maximum graph search depth, and record it as the second risk impact subgraph; continue to execute the loop iteration based on the second risk impact subgraph.

[0045] In this embodiment, several boundary nodes marked as preset safety exits are obtained in the first risk impact subgraph. The safety exit nodes are pre-marked in the knowledge graph, such as the exterior doors of a building, stairwell exits, etc.

[0046] In each iteration, the environmental risk coefficient of the boundary nodes is monitored to see if it exceeds the preset exit blockade threshold. The preset exit blockade threshold is a key parameter used to determine whether an exit is unavailable due to fire risk. It is set based on the maximum acceptable risk level to ensure safe escape for personnel, and is typically determined according to safety standards such as smoke concentration and temperature. A typical value range is 0.6 to 0.8, with a typical value of 0.7.

[0047] The preset mapping rules include the maximum graph search depth determined based on the maximum difference between the initial fire source location data and the spatial geometric features of the risk nodes.

[0048] When the environmental risk coefficient of any boundary node exceeds the preset exit blockade threshold, the boundary node is removed from the first risk impact subgraph, and the current maximum graph search depth is reduced and corrected according to the difference between the environmental risk coefficient and the preset exit blockade threshold to obtain the first corrected maximum graph search depth. The correction method can be linear reduction.

[0049] The calculation logic for the first corrected maximum graph search depth is as follows: Subtract the preset exit blockade threshold from the environmental risk coefficient to obtain the environmental risk difference; then divide the environmental risk difference by the preset exit blockade threshold to obtain an environmental risk exceeding the threshold ratio; next, subtract this environmental risk exceeding the threshold ratio from 1 to obtain an adjustment coefficient; finally, multiply the current maximum graph search depth by this adjustment coefficient to obtain the first corrected maximum graph search depth; ensuring that the larger the difference, the greater the depth reduction. Subsequently, the risk impact subgraph acquisition module is called to re-acquire the updated first risk impact subgraph centered on the central risk node and using the first corrected maximum graph search depth, which is denoted as the second risk impact subgraph.

[0050] This embodiment dynamically monitors the risk status of safety exits during the iteration process. Once the risk of an exit exceeds the limit, it is excluded and the search range is narrowed, so that subsequent iterations focus on areas that are still passable, avoiding invalid calculations, while ensuring that the evacuation path always points to an actual safe exit.

[0051] Furthermore, the specific process for obtaining the passage risk coefficient is as follows: Obtain the environmental risk coefficient of the central risk node in the current iteration, and the environmental risk coefficient of each adjacent risk node; for each adjacent risk node, calculate the difference between the environmental risk coefficient of the central risk node and the environmental risk coefficient of that adjacent risk node, and divide it by the sum of the environmental risk coefficient of that adjacent risk node and a preset minimum positive number to obtain the environmental risk gradient ratio; normalize the environmental risk coefficient and the environmental risk gradient ratio of each adjacent risk node to obtain the normalized ratio of the environmental risk gradient of the central risk node to that adjacent risk node; when the environmental risk... When the gradient normalization ratio is positive, the travel risk coefficient pointing to the adjacent risk node is decreased according to the environmental risk gradient normalization ratio to obtain the first travel risk coefficient, which is used as the travel risk coefficient of the connected edge corresponding to the adjacent risk node in the current iteration; when the environmental risk gradient normalization ratio is negative, the travel risk coefficient pointing to the adjacent risk node is increased by the absolute value of the environmental risk gradient normalization ratio to obtain the second travel risk coefficient, which is used as the travel risk coefficient of the connected edge corresponding to the adjacent risk node in the current iteration; when the environmental risk gradient normalization ratio is zero, the travel risk coefficient pointing to the adjacent risk node remains unchanged.

[0052] In this embodiment, for each adjacent risk node, the difference between the environmental risk coefficient of the central risk node and the environmental risk coefficient of the adjacent risk node is calculated, and then divided by the sum of the environmental risk coefficient of the adjacent risk node and a preset minimum positive number to obtain the environmental risk gradient ratio. The preset minimum positive number is used to prevent division by zero.

[0053] The environmental risk coefficient is normalized to the ratio of the environmental risk gradient of each adjacent risk node, resulting in the normalized ratio of the environmental risk gradient of the central risk node to its adjacent risk nodes. The normalization method involves summing the absolute values ​​of the gradient ratios of all adjacent nodes, and then dividing each ratio by the sum. This yields a normalized ratio between 0 and 1.

[0054] When the environmental risk gradient normalization ratio is positive, the traversal risk coefficient pointing to the adjacent risk node is reduced according to the environmental risk gradient normalization ratio to obtain the first traversal risk coefficient, which is then used as the traversal risk coefficient of the connected edge corresponding to the adjacent risk node in the current iteration. The reduction is proportional to the normalization ratio.

[0055] When the environmental risk gradient normalization ratio is negative, the absolute value of the environmental risk gradient normalization ratio is increased to the access risk coefficient pointing to the adjacent risk node, resulting in a second access risk coefficient, which is then used as the access risk coefficient of the connected edge corresponding to the adjacent risk node in the current iteration. The increase is proportional to the absolute value, where β is a preset adjustment coefficient, typically 0.2.

[0056] When the environmental risk gradient normalization ratio is zero, the passage risk coefficient pointing to the adjacent risk node remains unchanged.

[0057] This embodiment quantifies the risk differences between adjacent nodes using the environmental risk gradient ratio and dynamically adjusts the passage risk coefficient through a normalized ratio. When the risk of an adjacent node is low (i.e., the gradient is positive), the passage cost is reduced to encourage movement in that direction; when the risk is high (i.e., the gradient is negative), the passage cost is increased to warn of avoidance. This mechanism enables risk gradient information to directly drive the iterative update of path costs, achieving adaptive propagation of risk distribution.

[0058] Furthermore, it also includes a continuous monitoring module: if, after the first prevention and control decision report is output, several new initial fire source location data are acquired within the first time window, then the corresponding new central risk node is mapped from the institutional knowledge graph according to a preset mapping rule for each new initial fire source location data; for each new central risk node, the corresponding new risk impact subgraph is obtained with the new central risk node as the center and the current maximum graph search depth; all new risk impact subgraphs are subjected to a union operation to obtain a merged risk impact subgraph and the risk nodes that simultaneously belong to at least two different new risk impact subgraphs are marked as cross-risk nodes; the iterative execution module is called according to the cross-risk nodes to obtain the cross-traffic risk coefficient after the iteration is completed; and the second prevention and control decision report is output with the cross-traffic risk coefficient.

[0059] In this embodiment, if several new initial fire source location data are acquired within the first time window after the first prevention and control decision report is output, then the corresponding new central risk node is mapped from the institutional knowledge graph according to a preset mapping rule for each new initial fire source location data. The time window is a preset duration used to monitor whether the fire source spreads or new fire points appear. Its setting is based on the time scale of fire development, with a typical value range of 30 seconds to 2 minutes, and a typical value of 1 minute.

[0060] For each newly added central risk node, obtain the corresponding newly added risk impact subgraph centered on the newly added central risk node and using the current maximum graph search depth.

[0061] Perform a union operation on all newly added risk impact subgraphs to obtain a merged risk impact subgraph, and mark the risk nodes that simultaneously belong to at least two different newly added risk impact subgraphs as cross-risk nodes. The union operation is simply taking the set of all nodes.

[0062] This embodiment can handle situations where fire spreads or multiple fires occur. It monitors new fire sources through time windows, merges multiple risk impact sub-maps, identifies overlapping areas affected by multiple fire sources, and then conducts specialized risk assessments on these areas to ensure that the decision report reflects the true risk distribution under complex fire conditions.

[0063] Furthermore, the specific process for obtaining the cross-traffic risk coefficient is as follows: For each cross-risk node, obtain the traffic risk coefficient corresponding to each risk impact subgraph to which it belongs and perform mean processing to obtain the average traffic risk coefficient; count the number of times the cross-risk node is referenced by different newly added risk impact subgraphs in the merged risk impact subgraph as the fire source coverage frequency; calculate the cross-risk weight factor based on the fire source coverage frequency; multiply the average traffic risk coefficient of the current iteration of the cross-risk node with the cross-risk weight factor and normalize it to obtain the cross-traffic risk coefficient of the cross-risk node.

[0064] In this embodiment, for each cross-risk node, the passage risk coefficient corresponding to each risk-affected subgraph to which it belongs is obtained and averaged to obtain the average passage risk coefficient. The averaged factor is an arithmetic mean; for example, if the node belongs to two subgraphs, the average of the two passage risk coefficients is taken.

[0065] The frequency of cross-risk nodes being referenced by different newly added risk impact subgraphs in the merged risk impact subgraph is used as the fire source coverage frequency. Frequency is determined by how many subgraphs the node appears in; for example, if it appears in two subgraphs, the frequency is 2.

[0066] The cross-risk weighting factor is calculated based on the frequency of fire source coverage. The weighting factor is typically positively correlated with the frequency, for example, set as the frequency itself, or the frequency multiplied by a predetermined coefficient set by expert experience. The cross-risk weighting factor is used to quantify the comprehensive risk intensity of nodes under the influence of multiple fire sources. A specific example formula is as follows: Fire source coverage frequency The number of different newly added risk-affected subgraphs to which a node belongs; The maximum fire source coverage frequency among all intersection nodes; , This represents the cross-risk weighting factor.

[0067] The cross-traffic risk coefficient of a cross-risk node is obtained by multiplying the mean coefficient of its current iteration's traffic risk by the cross-risk weight factor and then normalizing the result. Normalization can be achieved by dividing by the largest product value among all cross-risk nodes, or by using other methods to keep the result within a reasonable range, such as dividing by the maximum frequency value, so that the coefficient does not exceed 1.

[0068] This embodiment performs weighted processing on the cross-risk nodes, giving higher weights to nodes covered by more fire sources, thereby highlighting their high-risk characteristics and making the final output cross-traffic risk coefficient more accurately reflect the actual risk level under the influence of multiple fire sources.

[0069] Furthermore, the continuous monitoring module is also used to: after outputting the second prevention and control decision report, within the second time window, again obtain the cross-traffic risk coefficient of each cross-risk node in the merged risk impact sub-graph; calculate the maximum rate of change of the cross-traffic risk coefficient of the same cross-risk node within the second time window, and record it as the rate of change of the traffic risk coefficient; calculate the average of the rate of change of the traffic risk coefficient of all cross-risk nodes, and use it as the cross-region risk fluctuation index; if the cross-region risk fluctuation index exceeds the preset fluctuation threshold, then increase the current graph search depth according to the difference between the cross-region risk fluctuation index and the preset fluctuation threshold to obtain the second corrected maximum graph search depth.

[0070] In this embodiment, after outputting the second prevention and control decision report, within the second time window, the cross-traffic risk coefficient of each cross-risk node in the merged risk impact sub-graph is obtained again.

[0071] Calculate the maximum rate of change of the cross-traffic risk coefficient of the same cross-traffic risk node within the second time window, and denot it as the rate of change of the traffic risk coefficient.

[0072] The average rate of change of the travel risk coefficient at all cross-risk nodes is used as the risk volatility index for the cross-region. The average reflects the drastic degree of overall risk change.

[0073] If the cross-region risk volatility index exceeds a preset volatility threshold, the current chart search depth is increased based on the difference between the cross-region risk volatility index and the preset volatility threshold, resulting in a second maximum corrected chart search depth. The preset volatility threshold is used to determine whether the risk is changing rapidly, and it is set based on the allowable normal volatility range, typically ranging from 0.1 to 0.3, with a typical value of 0.2.

[0074] The calculation logic for the second modified maximum graph search depth is as follows: Subtract a preset volatility threshold from the cross-region risk volatility index to obtain the cross-region risk volatility difference; then add 1 to this cross-region risk volatility difference to obtain the cross-region risk volatility adjustment coefficient; finally, multiply the current maximum graph search depth by this cross-region risk volatility adjustment coefficient to obtain the second modified maximum graph search depth. This ensures that the greater the volatility, the greater the increase in search depth, in order to capture a wider range of risk changes.

[0075] This embodiment continuously monitors risk fluctuations in the cross-regional area. When the fluctuation exceeds the threshold, the search range is automatically expanded to more comprehensively assess the risk spread trend. This enables the system to adaptively adjust the analysis granularity, ensuring that it can still provide accurate decision-making basis when the fire situation changes rapidly.

[0076] This application also provides a computer-readable storage medium for storing a computer program, which, when executed by a processor, implements a knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions.

[0077] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0078] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0079] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0080] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0081] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0082] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A knowledge graph-based fire risk prevention and operational decision-making system for elderly care institutions, characterized in that, include: Data acquisition module: used to acquire initial fire source location data and map the central risk node from the pre-trained institutional knowledge graph according to preset mapping rules; Risk impact subgraph acquisition module: used to acquire the first risk impact subgraph centered on the central risk node. Each risk node in the first risk impact subgraph contains spatial geometric features, and the edge weight of each connected edge of each risk node is initialized as the basic passage coefficient. Iterative execution module: Used to perform loop iterations. Obtain the environmental risk coefficient of the previous iteration, calculate the environmental risk coefficients of the central risk node and all adjacent risk nodes in the first risk impact subgraph, and the difference between them and the environmental risk coefficient of the previous iteration. Obtain the adjacent iteration difference of the central risk node and the adjacent iteration difference of adjacent risk nodes. Calculate the ratio of the two and correct the passage risk coefficient of the previous iteration accordingly to obtain the passage risk coefficient of the current iteration. Iteration Judgment Module: Used to determine if the number of times the difference between the passage risk coefficients of adjacent iterations does not exceed the preset risk difference threshold, or if the current iteration number exceeds the maximum iteration number, then the iteration is determined to end and the passage risk coefficient after the iteration ends is obtained. Output module: Used to output the first prevention and control decision report based on the traffic risk coefficient after the iteration ends.

2. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 1, characterized in that, The specific process for determining the central risk node is as follows: The spatial geometric features of each risk node are read from the institutional knowledge graph, and the spatial geometric features include at least the three-dimensional coordinate values ​​corresponding to the risk node; Calculate the Euclidean distance between the initial fire source location data and the three-dimensional coordinates of each risk node, and use it as the spatial distance value; The risk node with the smallest spatial distance value is determined as the central risk node.

3. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 1, characterized in that, The specific process for obtaining the first risk impact subgraph is as follows: Starting from the central risk node, a breadth-first search algorithm is used, with the current maximum graph search depth as the traversal termination condition, to visit all risk nodes whose search depth does not exceed the current maximum graph search depth layer by layer. Extract the visited risk nodes and all connected edges that link the risk nodes to form the first risk impact subgraph.

4. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 1, characterized in that, The specific calculation process for the environmental risk coefficient is as follows: The absolute value of the difference between the vertical projection height values ​​of the spatial geometric features of the central risk node and any other risk node in the first risk impact sub-graph is obtained as the vertical distance value. The absolute value of the difference between the horizontal projected distance values ​​of the spatial geometric features of the central risk node and any other risk node in the first risk impact sub-graph is used as the horizontal distance value. The vertical distance value is weighted according to the first preset weight, and the horizontal distance value is weighted according to the second preset weight. The weighted sum is used as the spatial risk transmission coefficient of the other risk node relative to the central risk node. Obtain the preset baseline environmental risk value and spatial risk transmission coefficient of the other risk node, and generate the initial environmental risk coefficient of the other risk node relative to the central risk node; The environmental risk coefficient of the central risk node is obtained by summing the initial environmental risk coefficients of all other risk nodes with respect to the central risk node.

5. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 4, characterized in that, The iterative execution module is also used for: Obtain several boundary nodes marked as preset safety exits in the first risk impact subgraph; In each iteration, monitor whether the environmental risk coefficient of the boundary node exceeds the preset exit blocking threshold; The preset mapping rules include the maximum graph search depth determined based on the maximum difference between the initial fire source location data and the spatial geometric features of the risk nodes; When the environmental risk coefficient of any of the boundary nodes exceeds the preset exit blockade threshold, the boundary node is removed from the first risk impact subgraph, and the current maximum graph search depth is reduced and corrected according to the difference between the environmental risk coefficient and the preset exit blockade threshold to obtain the first corrected maximum graph search depth. The risk impact subgraph acquisition module is then called to re-acquire the updated first risk impact subgraph with the central risk node as the center and the first corrected maximum graph search depth, which is denoted as the second risk impact subgraph. The loop iteration continues based on the second risk impact subgraph.

6. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 4, characterized in that, The specific process for obtaining the traffic risk coefficient is as follows: Obtain the environmental risk coefficient of the central risk node in the current iteration, as well as the environmental risk coefficient of each adjacent risk node; For each adjacent risk node, calculate the difference between the environmental risk coefficient of the central risk node and the environmental risk coefficient of the adjacent risk node, and divide it by the sum of the environmental risk coefficient of the adjacent risk node and a preset minimum positive number to obtain the environmental risk gradient ratio. The environmental risk coefficient is normalized to the ratio of the environmental risk gradient of each adjacent risk node to obtain the normalized ratio of the environmental risk gradient of the central risk node to that adjacent risk node. When the environmental risk gradient normalization ratio is positive, the first passage risk coefficient is obtained by reducing the passage risk coefficient pointing to the adjacent risk node according to the environmental risk gradient normalization ratio, and is used as the passage risk coefficient of the connected edge corresponding to the adjacent risk node in the current iteration. When the environmental risk gradient normalization ratio is negative, the passage risk coefficient pointing to the adjacent risk node is increased by the absolute value of the environmental risk gradient normalization ratio to obtain the second passage risk coefficient, which is used as the passage risk coefficient of the connected edge corresponding to the adjacent risk node in the current iteration. When the environmental risk gradient normalization ratio is zero, the passage risk coefficient pointing to the adjacent risk node remains unchanged.

7. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 1, characterized in that, It also includes a continuous monitoring module: If, after the first prevention and control decision report is output, several new initial fire source location data are obtained within the first time window, then the corresponding new central risk node is mapped from the institutional knowledge graph according to the preset mapping rule for each new initial fire source location data. For each newly added central risk node, obtain the corresponding newly added risk impact subgraph centered on the newly added central risk node and using the current maximum graph search depth; Perform a union operation on all newly added risk impact subgraphs to obtain a merged risk impact subgraph and mark the risk nodes that simultaneously belong to at least two different newly added risk impact subgraphs as cross-risk nodes. The cross-traffic risk coefficient is obtained by calling the iterative execution module based on the cross-risk node after the iteration ends. The second prevention and control decision report is output based on the risk coefficient of cross-traffic.

8. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 7, characterized in that, The specific process for obtaining the cross-traffic risk coefficient is as follows: For each cross-risk node, obtain the passage risk coefficient corresponding to each risk impact subgraph to which it belongs and perform mean processing to obtain the passage risk mean coefficient. The number of times the cross-risk node is referenced by different newly added risk impact subgraphs in the merged risk impact subgraph is used as the fire source coverage frequency. Calculate the cross-risk weighting factor based on the frequency of fire source coverage; The cross-traffic risk coefficient of the cross-risk node is obtained by multiplying the mean coefficient of the current iteration of the cross-risk node with the cross-risk weight factor and normalizing the result.

9. The knowledge graph-based fire risk prevention and operation decision-making system for elderly care institutions according to claim 8, characterized in that, The continuous monitoring module is also used for: After outputting the second prevention and control decision report, within the second time window, the cross-traffic risk coefficient of each cross-risk node in the merged risk impact sub-graph is obtained again; Calculate the maximum rate of change of the cross-traffic risk coefficient of the same cross-traffic risk node within the second time window, and denot it as the rate of change of the traffic risk coefficient; The average rate of change of the passage risk coefficient of all cross-risk nodes is used as the risk fluctuation index of the cross-region. If the cross-region risk volatility index exceeds the preset volatility threshold, the current graph search depth is increased and corrected based on the difference between the cross-region risk volatility index and the preset volatility threshold, resulting in a second corrected maximum graph search depth.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the system as described in any one of claims 1-9.

Images