Accident consequence simulation calculation method based on mathematical physical model coupling solution

Abstract

The invention provides an accident consequence simulation calculation method based on mathematical physical model coupling solution, and the method comprises the steps: constructing an accident chain knowledge graph and a physical trigger graph, and building a causal relationship and physical constraints among equipment, states, events and consequences; gathering and mapping historical records, expert rules and online observation data into map entities and relationships, forming a baseline accident scene and calculating a baseline index; generating candidate paths by utilizing graph reasoning and coupled multi-physics field simulation, and realizing a closed loop of graph reasoning and physical solution through consistency check; performing scoring and disturbance simulation analysis on the candidate paths, and screening robust target paths; and carrying out high-fidelity simulation on the target path, identifying key nodes in combination with sensitivity analysis and a minimum cut-off set, and generating disposal suggestions and action priorities. According to the method, high-credibility prediction of accident evolution and emergency response closed-loop linkage are realized, and the method has high precision, high robustness and engineering implementability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of accident consequence simulation, in particular to an accident consequence simulation calculation method based on mathematical physical model coupling solution. BACKGROUND

[0002] With the continuous expansion of large-scale chemical, energy and storage facilities, the evolution process of accident consequences presents complex coupling, multi-stage transmission and strong nonlinearity. In actual accident situations, accidents are often accompanied by the coupling and superposition of multiple physical processes, such as the interaction of pressure relief and diffusion process, the parallel occurrence of thermal radiation and convection heat transfer, the coupling response of explosion shock wave and building structure, etc. These factors make the accident consequences have typical characteristics such as strong burst, high path diversity and difficult extrapolation prediction.

[0003] The existing accident consequence simulation has the following shortcomings: first, a single physical field simulation is mostly used, which lacks system modeling and causal reasoning of the accident chain, and it is difficult to reflect the dynamic evolution characteristics of complex accidents; second, the path extrapolation lacks a constraint mechanism coupled with physical accessibility, and the extrapolation result is easy to deviate greatly from the real accident, which is difficult to meet the application requirements of emergency practice; third, there is a lack of methods for evaluating the uncertainty and robustness of candidate accident paths, resulting in a lack of scientific basis for plan selection.

[0004] In view of the above shortcomings, the present application provides an accident consequence simulation calculation method based on mathematical physical model coupling solution.

[0005] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0006] The purpose of the present application is to provide an accident consequence simulation calculation method based on mathematical physical model coupling solution to solve the technical problems mentioned in the background art.

[0007] To achieve the above purpose, the present application provides the following technical solutions: An accident consequence simulation calculation method based on mathematical physical model coupling solution, comprising the following steps: S1, establishing an accident chain knowledge graph based on process flow and accident history information, representing the causal relationship between equipment, state, event and consequence; combining parameter threshold and conservation constraint to establish a physical trigger graph, forming accident trigger logic and physical accessibility constraint; S2, map historical accident records, expert rules, and online observation data into accident chain knowledge graph entities and relationships, trigger path after call coupling multi-physical field solver, establish baseline accident consequence field, and calculate baseline consequence index as extrapolation control reference; S3, use graph causal trigger relationship to traverse and reason to form candidate extrapolation path, map the path to simulation boundary conditions to solve the accident consequence field, and perform consistency check through mass conservation, energy conservation, and trigger satisfaction degree, realizing closed-loop iteration of graph reasoning and physical calculation; S4, comprehensive scoring of candidate paths based on consistency index, consequence index, and path complexity, using perturbation simulation and statistical analysis to obtain uncertainty interval and robustness index, and screening to obtain the final robust target path; S5, high-precision physical field simulation for the robust target path, calculation of accident consequence index, identification of key trigger nodes in the accident chain through sensitivity analysis and minimum cut set solution, and generation of disposal suggestions and action priority sequence; S6, output the path, key node, consequence index, and disposal suggestion in a standardized data format to the upper system, realize automatic linkage disposal through industrial protocol, and archive and version control the whole process of accident deduction, complete closed-loop emergency response and post-tracing.

[0008] S1 specifically includes: Establish an accident chain knowledge graph to represent devices, states, events, and causal relationship structure; Establish a physical trigger graph to represent threshold conditions, conservation constraints, and trigger logic; Establish a coupled multi-physical field solver to calculate the accident consequence field based on source term parameters and boundary conditions; Define accident scenario parameter vector, extrapolation path set, consequence index, consistency check index, convergence criterion, and constraint set; Set a rollback mechanism for abnormal recovery and linkage to the upper system.

[0009] S2 specifically includes: Map historical records, expert rules, and real-time observations into entities, relationships, and initial trigger conditions in the accident chain knowledge graph; Input the accident scenario parameter vector into the physical trigger graph to complete the parameterization of the trigger logic; Call the coupled multi-physical field solver to calculate the baseline accident consequence field based on the parameters and trigger conditions; Calculate and record the baseline consequence index as a reference for subsequent extrapolation path comparison; Store the baseline state for reference in subsequent screening and rollback.

[0010] S3 specifically includes: A candidate extrapolation path is generated according to a causal relationship reasoning on the accident chain knowledge graph; The candidate extrapolation path is mapped to a boundary condition and a source term of a coupled multi-physical field solver, and a simulation accident consequence field is calculated; A simulation consequence index is calculated, and the candidate extrapolation path is checked for physical reachability and conservation consistency based on consistency checking index; The checking result is used to update the relationship weight in the accident chain knowledge graph and the trigger condition in the physical trigger graph in reverse; A bidirectional iterative link of graph reasoning and physical calculation is formed to continuously optimize the credibility of the candidate extrapolation path.

[0011] S4 specifically includes: The candidate extrapolation path is scored according to the consistency checking index and the consequence index in the constraint set; A sequential screening method is used to retain a target extrapolation path set; The target extrapolation path set is subjected to perturbation simulation within the uncertainty range of the accident scene parameter vector; The robustness and consequence index interval characteristics of the target extrapolation path are evaluated; If the convergence criterion is not met or the constraint set is violated, a rollback mechanism is triggered to adjust the relationship weight, trigger condition or extrapolation path, and the iteration is restarted in S3.

[0012] S5 specifically includes: A coupled multi-physical field solver is called to perform high-fidelity coupled solving on the target extrapolation path set; The target accident consequence field and the target consequence index are obtained; Based on the sensitivity analysis of the target consequence index, a key trigger node in the accident chain knowledge graph is identified; Based on the graph structure centrality, a minimum cut set is extracted to represent the key accident control point; A treatment suggestion sequence and corresponding priority are generated based on the target extrapolation path.

[0013] S6 specifically includes: The target extrapolation path set, key trigger node, minimum cut set, target accident consequence field and target consequence index are output to the upper system; The treatment suggestion sequence and corresponding priority are output for linkage control and emergency response; The graph state, trigger graph parameters, solver results and consistency checking evidence are archived.

[0014] The extrapolation path iteration convergence process and rollback trajectory are recorded for online calibration and traceability analysis; A closed loop of knowledge graph, physical trigger and numerical solution is formed to provide reusable reasoning basis for subsequent accident warning and disposal.

[0015] The application has the beneficial effects that: The application expresses the logical relationship between the device running state, process flow, accident event and consequence in a structured manner by constructing the accident chain knowledge graph and physical trigger graph, and is coupled with the multi-physical field simulation model, so that the accident deduction is no longer dependent on a single path assumption, but supports multi-path, multi-stage, dynamic evolution of full-link reasoning and simulation. This technology breaks through the scene static nature of traditional accident consequence simulation, realizes the transformation from single-point prediction to multi-path dynamic extrapolation, and greatly improves the coverage and accuracy of accident prediction.

[0016] The application can strictly screen the candidate accident evolution path by introducing multi-dimensional consistency checking indexes such as mass conservation, energy conservation and trigger logic satisfaction, and eliminate the "pseudo path" that does not conform to the physical constraint, forming a "real path" set verified by physical calculation. Compared with the traditional method which depends on experience rules, this mechanism significantly reduces the deduction error and uncertainty, so that the accident path reasoning has self-correction and convergence ability, and improves the stability and reproducibility of the model.

[0017] The application carries out disturbance sampling and uncertainty propagation calculation on the accident input parameters, constructs a robustness index, selects the extrapolation path that still maintains high credibility under multiple working conditions and multiple disturbances, and enhances the anti-disturbance ability and generalization ability of the prediction result. This method effectively overcomes the problem that the path screening in the prior art is too sensitive to the initial conditions, so that the system can still maintain stable and reliable early warning and deduction effect under actual complex working conditions.

[0018] The application automatically identifies the trigger node with the greatest impact on the consequence in the accident chain based on sensitivity analysis and minimum cut set solving, realizes the identification of high-value intervention points of accident propagation. The system can automatically generate disposal action suggestions and priority sequences according to the importance and influence degree of the node, provide scientific basis and clear action path for accident emergency disposal, and significantly improve the response efficiency and prevention and control effect.

[0019] The application organically integrates graph reasoning, physical calculation, disturbance robustness analysis and engineering linkage, realizes the integration of "from prediction to action", can adapt to various complex accident scenes (such as storage and transportation leakage, explosion, fire, etc.), has good engineering generality, real-time performance and scalability, and provides technical support for the essential safety protection of major hazard sources. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 It is a schematic diagram of the accident consequence simulation calculation method based on mathematical and physical model coupling solving of the application. DETAILED DESCRIPTION

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example 1: As Figure 1 As shown in the figure, this embodiment provides a method for simulating and calculating accident consequences based on the coupled solution of mathematical and physical models, including the following steps: S1. Construct an accident chain knowledge graph and a physical trigger graph. Based on the process flow, equipment ledger and accident history information, establish an accident chain knowledge graph to represent the causal relationship between equipment, status, events and consequences; and combine monitoring parameter thresholds and conservation constraints to establish a physical trigger graph to form accident triggering logic and physical reachability constraints. S2. Perform data aggregation and baseline accident scenario simulation, map historical accident records, expert rules and online observation data into accident chain knowledge graph entities and relationships, trigger the path and call the coupled multiphysics solver to establish the baseline accident consequence field, and calculate the baseline consequence index as an extrapolation reference. S3. Based on graph reasoning and physical decomposition, candidate paths are generated. The graph causal triggering relationship is used to traverse and reason to form candidate extrapolation paths. The paths are mapped to simulation boundary conditions to solve the accident consequence field. Consistency is checked through mass conservation, energy conservation and trigger satisfaction. Paths that do not meet physical constraints are eliminated to realize the closed-loop iteration of graph reasoning and physical decomposition. S4. Conduct path scoring and uncertainty analysis. The candidate paths are comprehensively scored based on consistency indicators, consequence indicators and path complexity. The uncertainty interval and robustness indicators are obtained by perturbation simulation and statistical analysis. The final robust target path is then selected. S5. Conduct high-fidelity simulation and key node identification, perform high-precision physical field simulation for robust target paths, calculate accident consequence indicators, identify key triggering nodes in the accident chain by combining sensitivity analysis and minimum cut-off set solution, and generate disposal suggestions and action priority sequences. S6. Achieve result output, linkage and traceability. Output the path, key nodes, consequence indicators and handling suggestions to the upper system in a standardized data format. Achieve automatic linkage and handling through industrial protocols. Archive and version control the entire process of accident simulation to complete closed-loop emergency response and post-event traceability.

[0023] S1 specifically includes the following sub-steps: S110, establish an accident chain knowledge graph: based on the process flow, equipment account and accident history information, establish an accident chain knowledge graph for representing the causal relationship between the equipment, state, event and consequence related to the accident; The accident chain knowledge graph includes the following entity types: Device entities (such as valves, pipelines, storage tanks, cooling towers, reaction kettles, etc.); State entities (such as pressure over-limit, temperature over-limit, leakage, ignition failure, etc.); Event entities (such as diffusion, explosion, combustion, rupture, etc.); Consequence entities (such as personnel injury, facility damage, chain reaction, etc.); The relationship types include causal trigger relationship, state conversion relationship and functional dependency relationship, and the relationship weights The weighted directed graph structure is adopted, the number of typical entities ranges from 50 to 300, the number of relationship types ranges from 3 to 6, and the graph level depth ranges from 2 to 5; The weight initialization can be calculated by the following formula: wherein represents the trigger edge weight from node i to node j; represents the accident trigger edge from node i to node j in the accident chain knowledge graph, which corresponds to the evolution path of a specific accident event; represents an element in the set of all triggerable candidate accident evolution edges at the same node i; represents the attribute function value of edge , which can be defined according to the intensity of accident triggering, over-threshold degree, occurrence probability or danger level, etc.; represents the sum of all trigger edge attribute function values of node i.

[0024] Through the above calculation, the trigger intensity of all evolution paths of node i is standardized, so that the weight sum of each path satisfies: The normalization processing not only ensures the comparability of the relative weights between different paths, but also provides a quantitative basis for subsequent path extrapolation and simulation mapping. In the path extrapolation process, the path with higher trigger intensity will obtain higher weight, so as to be given priority in multi-path evolution reasoning, thereby improving the rationality and stability of accident evolution prediction; The method provides a calculable and expandable quantitative basis for subsequent accident chain path extrapolation, disturbance simulation and risk warning through the normalization processing of the trigger weight of the accident evolution path, which significantly improves the stability and accuracy of accident evolution prediction.

[0025] In a typical embodiment, the attribute function can be determined as follows: ; wherein represents the probability of occurrence of the accident trigger event, which can be obtained from historical data or real-time monitoring systems; represents the threshold intensity of the trigger event, such as the degree of overpressure or the proportion of overconcentration; represents the risk level of the occurrence of the accident type; , , is an empirical weight or a coefficient obtained by training a statistical learning model.

[0026] S120, establishing a physical trigger graph: based on the monitoring points, sensors and engineering parameters related to the accident, a physical trigger graph is established to represent the threshold conditions, conservation constraints and trigger logic; The threshold conditions include but are not limited to: pressure threshold : 0.6-1.2 MPa; temperature threshold : 50-90°C; concentration threshold : 10%-20% LEL; leakage rate : 0.05-1.0 kg / s.

[0027] The trigger logic is represented by a Boolean expression or a rule matrix, for example: where Y and C represent the pressure and concentration monitoring values of the accident scene, respectively; and represent the corresponding trigger thresholds. Only when the pressure and concentration both exceed the threshold, the accident node is activated and enters the downstream path extrapolation link. Here, the logical AND ∧ is used, which means: when the pressure Y exceeds the trigger threshold , and the concentration C exceeds the trigger threshold , it is considered that the accident trigger event occurs (Trigger=True), which realizes the accident trigger determination based on physical quantities, and improves the accuracy and automation level of accident reasoning; The conservation constraints include mass conservation, energy conservation and momentum conservation, with an allowed error: where is the mass conservation check error, which is used to measure the proportion of the difference between the input and output mass of the system to the input mass, represents the total input mass of the system, represents the total output mass of the system, represents the allowed relative error threshold. When is less than or equal to , it is considered that the simulation calculation result meets the mass conservation requirement; otherwise, the calculation parameters need to be adjusted. Through the above check, the credibility and stability of the simulation result of the accident consequences can be improved.

[0028] S130, establish a coupled multi-physics solver: according to the characteristics of the accident scene, a coupled multi-physics solver is established for multi-field coupled calculation of fluid diffusion, heat transfer, pressure fluctuation and chemical reaction.

[0029] The solver can be implemented by finite volume method (FVM) or finite element method (FEM), and the applicable computing platforms include OpenFOAM, ANSYS Fluent or COMSOL Multiphysics, etc.

[0030] The typical calculation domain grid size is 10,000-100,000 nodes, the time step range is 0.01-1s, and the simulation time window is 0-600s.

[0031] The boundary conditions include: leakage rate, environmental meteorological conditions, building obstacle layout and wind field disturbance.

[0032] Through iterative convergence criterion Control the calculation precision, and the convergence residual is set to 10 -4 order of magnitude.

[0033] S140, define key variables and parameters: Define the accident scene parameter vector , including initial pressure, temperature, concentration, leakage rate and accident occurrence time; Define the extrapolation path set P, the number of elements of which ranges from 10 to 100; Define the consequence index R, including concentration peak, thermal radiation dose, explosion shock radius and personnel injury level; Define the consistency check index , which is used to represent the consistency of the candidate path and the physical trigger diagram. Define the constraint set , including the safety boundary and the trigger range; define the convergence criterion , including the error upper limit and the iteration upper limit; define the rollback mechanism B for path recovery in abnormal situations.

[0034] S150, establish a linkage upper system interface: realize data interaction with the upper system through OPCUA or Modbus-TCP industrial communication protocol, transmission frequency 1-5Hz, communication delay less than 2s; The upper system includes DCS (Distributed Control System), ECS (Emergency Control System) or SCADA system; The data interface supports synchronous transmission of accident state pushing, path identification, index value and disposal suggestion, and the interface supports event backtracking identification and linkage trigger signal, ensuring the real-time and traceability of subsequent emergency decision-making.

[0035] S2 specifically comprises the following sub-steps: S210, data aggregation and graph instantiation: historical accident records, expert rules and online observation data are unified and aggregated to form a solvable data set; The historical accident records are stored in the form of a structured database or a CSV file, and each record contains at least the following fields: accident occurrence timestamp, device ID, state code, leakage rate, temperature rise curve, and environmental meteorological parameters. The typical record size is 500-5000.

[0036] The expert rules are in the form of IF-THEN or Boolean expressions, for example: IF (T>60°C AND Q>0.1kg / s) THEN state="thermal diffusion warning" The above expression means: if the temperature exceeds 60°C and the leakage rate exceeds 0.1kg / s, the system enters a thermal diffusion warning state; where "IF" means "if", "T>60°C" means "trigger when temperature is greater than 60°C", "AND" means logical "and", "Q>0.1kg / s" means "trigger when leakage rate is greater than 0.1kg / s"; "THEN state" means "once the condition is met, the system state is marked as thermal diffusion warning".

[0037] The online observation data is accessed in the form of a time series stream with a sampling frequency of 1-5Hz, including real-time signals of pressure sensors, temperature sensors and gas concentration sensors.

[0038] The above data is converted into entities and relationships in the accident chain knowledge graph through mapping logic: Device ID→Device Entity; State code / sensor alarm→State Entity; IF-THEN rule→Causal relationship edge; Consequence type→Consequence entity.

[0039] The initial instantiation of the graph is completed, forming a node set, a relationship set and a trigger state initial value, providing a structural basis for subsequent calculations.

[0040] S220, trigger state recognition and path activation: the accident scene parameter vector Input physical trigger graph; for each trigger node, judge whether it is true. If true, the trigger state is marked as 1, otherwise as 0; where represents the i-th trigger node corresponding to the accident scene input parameter value, represents the trigger threshold value corresponding to the node; The trigger logic automatically identifies the paths that meet the conditions based on the Boolean expression and generates an activated path set ; Typically, there are 10-50 activation paths, corresponding to multiple possible initial evolution directions of the accident. The activation states and activation paths are stored in memory buffers and databases synchronously, providing input conditions for baseline simulation calculation.

[0041] S230, Baseline post-accident consequence field solving: calling a coupled multi-physical field solver to perform simulation calculation of the baseline accident scenario based on the boundary conditions; ; The calculation domain grid size is 10,000-100,000 nodes, the time step range is 0.01-1s, and the simulation duration is 0-600s; typical boundary conditions include: leakage rate Q: 0.05-1.0 kg / s; ambient wind speed : 0-10 m / s; ambient temperature : 20-35°C; initial pressure Y: 0.6-1.2 MPa; The solving process outputs three types of results: temperature field, concentration field, and pressure field, which are used to depict the effects of accident diffusion, heat radiation, and shock wave; the calculation process uses residual control convergence, and the target convergence error is set to 10 -4 .

[0042] S240, Baseline consequence index calculation and storage: based on the simulation output results, calculating the baseline consequence index , including: maximum concentration peak: wherein is the maximum concentration value, specifically the highest pollution concentration in the entire calculation region at the end time , t represents the calculation region, x represents the spatial position, , and t represents the end time of simulation or monitoring. represents the concentration value at position x and time ; through the above formula, the peak concentration at the end of the accident can be quickly obtained as a key index for risk analysis, emergency decision-making, and threshold comparison. Heat radiation dose:

[0043] wherein represents the cumulative heat radiation energy, used to represent the total amount of heat radiation borne by a point or a spatial region during the entire accident process; represents the instantaneous heat radiation intensity at position x at time t, usually in units of watts per square meter (W / m 2 ); x represents the spatial position subjected to heat radiation; t is the time variable; ​The formula represents the end time of the accident evolution or the end time of the calculation. By time integrating the thermal radiation intensity of the whole accident process, the cumulative value of the radiation energy of a certain point during the entire accident duration can be obtained. The thermal radiation contribution of each time step is accumulated, thereby reflecting the overall thermal load effect of the accident process on the target point.

[0044] For example, in the accident scenarios of fireball explosion, pool fire or jet fire, the thermal radiation intensity changes significantly with time. Through the above formula, the total radiation energy of the target area can be accurately quantified, providing reliable basis for thermal hazard classification, equipment protection design and emergency response; Explosion shock wave radius : refers to the overpressure action range generated by the explosion shock wave in the explosion accident. Its determination method is: by calculating the spatial distribution of the shock wave pressure field, and extracting the isobaric surface under a certain overpressure threshold (such as 0.02 MPa, 0.07 MPa, etc.), then performing spatial geometric analysis on the isobaric surface to obtain the radius corresponding to the isobaric surface, i.e. the explosion shock wave radius ; At the same time, the personnel injury level, building damage index and other derived indexes can be calculated for risk classification; the calculation results are stored in JSON or relational database table format, and the fields include: index name, value, timestamp, calculation version number; as a reference value for subsequent extrapolation path comparison and inversion algorithm.

[0045] S250, baseline state solidification and control mechanism: package the baseline activation path set , accident consequence field , consequence index , trigger state vector and graph structure state into the database; The storage method can include a database (PostgreSQL), a binary model cache (HDF5) or an engineering file (JSON+BIN hybrid format); the typical storage volume is 1-5 MB / scene; these baseline data are used for subsequent extrapolation path screening, simulation comparison and rollback mechanism, providing a stable reference frame for the entire accident deduction algorithm; the storage process supports version number and checksum to ensure the traceability and consistency of the simulation results.

[0046] S3 specifically includes the following sub-steps: S310, accident chain knowledge graph reasoning to generate candidate extrapolation paths: taking the initial state of the accident chain knowledge graph in S210-S250 as input, and performing path reasoning according to the causal trigger relationship in the graph.

[0047] Backward reasoning from the active node, search all reachable paths that satisfy the trigger condition, form a candidate extrapolation path set .

[0048] Each path is composed of a node sequence and a corresponding relationship weight sequence , where the node sequence is used to represent the spatio-temporal logical order of accident evolution, and the relationship weight sequence is used to quantify the strength or occurrence probability of the trigger relationship between adjacent nodes. Through the above structured representation, the overall risk level of different paths can be calculated and compared, providing a data basis for accident chain extrapolation reasoning and path optimization; The number of typical candidate paths is 50-300, and the length of a single path is 3-10 nodes. When generating paths, cyclic paths are pruned to ensure that the path is directed acyclic (DAG), preventing repeated reasoning and calculating redundancy.

[0049] S320, map the candidate paths to the simulation boundary conditions and perform solving: For each candidate extrapolation path , according to the node attributes and trigger logic in the path, map it to the boundary conditions of the coupled multi-physics solver. For example: node "tank leakage" → leakage rate boundary ; Node "gas diffusion" → environmental wind field and concentration field source term; node "flash explosion" → transient pressure wave boundary ; In the mapping process, maintain the same grid size (10,000-100,000 nodes), time step (0.01-1s) and solving accuracy (residual ≤10 -4 ) as in S230; by calling the solver, output the corresponding accident consequence field for each , including temperature field T(x, t), concentration field C(x, t), and pressure field Y(x, t).

[0050] S330, calculate the simulation consequence indicators and perform consistency check; according to the simulation results of each candidate path , calculate the corresponding simulation consequence indicators , including: Concentration peak: Where represents the maximum concentration value in the calculation region at the end of the accident under the i-th candidate path ; represents the position x and time The concentration field distribution of the hazardous substance; through this index, the peak value of the spatial distribution of the hazardous substance at the end of the accident can be characterized, which is used to determine the hazard level and warning range.

[0051] Thermal radiation dose: Wherein represents the total amount of thermal radiation accumulated at the corresponding position of path i during the entire process of the accident; represents the instantaneous thermal radiation intensity of the path at time t, which reflects the cumulative effect of energy during the accident process and is the core parameter of thermal hazard assessment; Explosion shock radius represents the explosion shock radius obtained by analyzing the explosion shock wave overpressure surface in the explosion accident triggered by path i, which is used to characterize the explosion damage range and protection radius.

[0052] Consistency check is performed for each candidate path to determine whether the path reasoning and physical algorithm meet the conservation and triggering conditions. The consistency check index is defined as: Wherein , , are weight coefficients, is the mass conservation error, is the energy conservation error, is the triggering condition matching confidence; when the path meets the physical conservation and triggering conditions, the consistency check index value is higher, and it is preferentially entered into the accident extrapolation and risk assessment link; if it is lower, the path is eliminated or given a lower weight; and indicate that the smaller the error, the larger the value obtained by subtracting the error from 1, and the higher the consistency; the larger the error, the lower the consistency score.

[0053] Used to indicate whether the path meets the triggering conditions (such as pressure threshold, concentration threshold, etc.), if it fully meets, the value can be set to 1; if it partially meets or does not meet, it can be a decimal between 0 and 1 or 0.

[0054] Wherein is calculated with reference to , which is calculated from the deviation of the input mass and the output mass of the system. Specifically, by mass statistics of the accident source term, the diffusion area and the boundary flux, the total input mass of the system and the total output mass of the system are obtained, and the calculation formula is: ; Secondly, the energy conservation error The way of obtaining and the quality conservation are similar, by the statistics of the input and output of the energy in the accident process, the total input energy and the total output energy are obtained, including thermal radiation, shock wave energy and other dissipation forms. The calculation formula is: ; Trigger condition matching confidence According to the accident trigger judgment logic calculation. By comparing the accident key state parameters (such as pressure, concentration, etc.) with the set trigger threshold, it is judged whether the accident trigger condition is met. If the threshold is completely met, The value is 1; if it is partially met or close to the threshold, the confidence between 0 and 1 is calculated by normalization mapping; if it is not met, it is 0.

[0055] Consistency judgment rule: ; Where is the set threshold (such as 0.85), it is considered that the path passes the consistency check; otherwise, it is rejected.

[0056] S340, update the atlas weight and trigger state: for the path that passes the consistency check, increase the weight of the corresponding relationship in the path To strengthen the credibility; for the path that does not pass, reduce the relationship weight.

[0057] The weight update formula is: Where represents the updated value of the relationship weight from node i to node j; represents the weight value before updating; is the update step or learning rate, which controls the sensitivity of weight adjustment; represents the consistency check index calculated by the ith candidate path. is the consistency threshold, used to judge whether the path needs to be enhanced or weakened, and the difference between the two represents the deviation of the relative threshold of the path consistency, the larger the deviation, the stronger the adjustment; if Below 0.1, mark the relationship as "invalid" and exclude it in subsequent reasoning; at the same time, adjust the corresponding trigger state in the trigger graph synchronously to ensure that the atlas structure and the physical process are consistent.

[0058] When the consistency check index of the path exceeds the threshold, the weights of all edges in the path are updated synchronously; when the consistency check index If only local threshold is exceeded, only the corresponding local edge weight is increased to realize local reinforcement; if threshold is not exceeded, no adjustment is made to the edge weight.

[0059] S350, bidirectional iteration closed loop of graph reasoning and object resolution algorithm: input the updated graph weight and trigger state into the path reasoning module again to regenerate the candidate path set ; form a closed loop iteration with the simulation and checking process of S320-S330 until the convergence criterion Γ is met or the iteration upper limit is reached.

[0060] The convergence criterion includes: Wherein represents the consistency checking index value of the ith candidate path at the kth iteration; represents the consistency checking index value of the corresponding path at the k-1th iteration; is a preset convergence threshold, preferably a very small positive number; represents the maximum value of all path consistency change amounts; k is the iteration step number; when the maximum change amount is less than the threshold, it is considered that the path consistency checking index has tended to be stable, the iteration process is terminated, repeated calculation is avoided, the path reasoning efficiency is improved, and the stability and uniqueness of the final reasoning result are ensured; output the final target path set that passes the consistency checking , and store the corresponding accident consequence field and consequence index ; The path that does not converge will trigger the rollback mechanism B to backtrack to the last stable iteration state to ensure the stability and credibility of the reasoning.

[0061] S4 specifically includes the following sub-steps: S410, candidate path comprehensive score calculation: for the candidate extrapolation path set that passes the consistency checking obtained through S310-S350 , calculate the comprehensive score of each path ; the score dimensions include the consistency checking index , the consequence severity index , and the path length , and the comprehensive score function is defined as: Wherein represents the comprehensive score of the ith candidate path; represents the consistency checking index of the ith candidate path (reflecting whether the path satisfies the constraints such as mass conservation, energy conservation, trigger condition confidence, etc.); ​represents the consequence severity indicator (e.g. maximum concentration peak, blast radius, thermal radiation dose, etc.) associated with the i-th path; represents the path length or complexity indicator of the i-th path (which can be represented as the number of nodes, the number of propagation steps, the path topology complexity, etc.); , is a function for quantitatively mapping the consequence severity and path length, such as normalization, logarithmic mapping or weight scaling; , , respectively represent the weight coefficients of the consistency check indicator, the consequence severity indicator and the path length indicator, used to adjust the contribution proportion of each item in the comprehensive score. By weighting and superimposing the multi-dimensional attributes of the candidate paths, the quantitative scoring and ranking of the paths are realized, so as to preferentially select the key accident evolution path with high credibility, high hazard and low complexity.

[0062] wherein is a normalization function for the accident severity, which can be: wherein represents the initial value of the accident consequence severity indicator associated with the i-th path; is the maximum value among all candidate path consequence severity indicators; is the minimum value among all candidate path consequence severity indicators; is the normalized consequence severity score, ranging from [0, 1].

[0063] is a penalty term for the path complexity, for example: wherein is the path length, is the maximum path length in the candidate path set; represents the normalized path length score, ranging from [0, 1]. Through this linear normalization function, the quantitative processing of the path length can be realized, so that shorter paths obtain higher scores, which is helpful for preferentially selecting the key path with high accident propagation efficiency and strong importance; the typical parameter setting range is . After scoring all candidate paths, they are ranked from high to low according to the value to form a ranked set .

[0064] S420, target extrapolation path screening: the ranked set is screened according to the following conditions: wherein Score threshold (typical value 0.6-0.8); Ensure the physical plausibility of the path; Be in the high-risk quantile (e.g. above 80% quantile); The set of retained target extrapolated paths is denoted as Typical number of paths is in the range of 5-20; record the path ID, score, threshold hit in the screening process to form a screening record table for traceability and verification.

[0065] S430, perturbation simulation and uncertainty propagation analysis: in the process of accident evolution path risk calculation, perturbation simulation and uncertainty propagation analysis mechanism is introduced to fully consider the result uncertainty caused by input parameter perturbation; for each , perturbation is performed on the accident scene parameters to generate a set of parameter samples ; denotes the jth accident evolution extrapolated path obtained after scoring and screening.

[0066] The perturbation method can include: Monte Carlo random sampling (typical sample number N=200-1000); Sobol global sensitivity analysis; Gaussian perturbation or interval perturbation. The perturbation range is set to ±10%-±30% of the original parameter; call the coupled multi-physics solver for each perturbed sample to calculate the corresponding consequence index ; statistic the index distribution under perturbed samples to get the uncertainty interval , calculate the confidence level CL by the ratio of the number of perturbed samples that meet the warning condition to the total number of samples: This confidence level is used to reflect the probability of the accident path triggering major consequences under input uncertainty, thereby providing a basis for subsequent path weight correction and risk optimization, where the warning condition is defined as the event set where the simulated consequence index exceeds the regulatory threshold (such as explosion shock wave radius, maximum concentration, thermal radiation dose, etc.).

[0067] S440, robustness evaluation and risk ranking: define the path robustness index : where denotes the robustness index of the jth path; CL is the confidence level, which represents the probability of the path triggering the warning in the perturbation simulation (see S430); denotes the coefficient of variation of the consequence index, is the standard deviation, is the mean; Weighting factor to control the impact of confidence level and coefficient of variation in robustness calculation; the higher the CL, the easier the path triggers significant consequences, indicating higher risk; The smaller the CV, the more stable and reliable the path consequences, indicating less simulation uncertainty; The larger the CV, the more risky and stable the path, which is more worthy of attention; For , sort the paths according to value, and prioritize paths with high robustness and high risk; after screening, form the final robust target path set , typically 3-10 paths; record the perturbation range, distribution interval, confidence level, and robustness index of each path, and generate a robustness analysis report structure.

[0068] S450, rollback mechanism and termination condition: if the number of paths in is less than the preset lower limit (e.g. 3), or the robustness index (typically threshold 0.6-0.8), trigger rollback mechanism B. The rollback mechanism performs the following steps: Lower or threshold, relax path screening criteria; adjust perturbation range (e.g. ±30%→±15%); increase iteration number or expand candidate path set.

[0069] If the convergence criterion is met: or the iteration limit is reached , terminate the rollback process, output the final robust target path set and its uncertainty interval and robustness results for subsequent S510-S550 high-fidelity solving; where represents the robustness index of the jth path in the kth iteration; represents the robustness index of the corresponding path in the k-1th iteration; is the preset robustness convergence threshold, usually a small positive number (e.g. 10 -3 ); represents the maximum value of the change in all path robustness indices; Preferably, when the maximum change in path robustness index in two consecutive iterations is less than the threshold, the calculation result is considered to have converged, the iteration is terminated, the stability and convergence of the robustness evaluation process are ensured, and the analysis and calculation efficiency is improved.

[0070] S5 specifically includes the following sub-steps: S510, high-fidelity post-accident consequence field solving: solve the final robust target path set obtained in S450 Input coupling multi-physics solver, perform high-fidelity simulation calculation; Simulation setup: grid scale 50,000-200,000 nodes; time step 0.005-0.05s; total simulation duration 0-900s; convergence residual threshold 10 -5 Boundary conditions and source term mapping consistent with S320, such as leakage rate Q, meteorological wind field , pressure wave , etc. are defined by path nodes; Simulation results include: temperature field T(x, t); concentration field C(x, t); pressure field Y(x, t); flame propagation profile (if applicable); store simulation results as structured data, spatial resolution 0.1-0.5m, temporal resolution 0.05-0.5s, for subsequent sensitivity analysis and key node identification.

[0071] S520, target consequence index calculation and extraction: based on high-fidelity simulation results, calculate target consequence index for each path , including but not limited to: maximum concentration peak , thermal radiation dose , explosion shock wave radius ; Store results as a three-tuple data structure (path ID, , spatiotemporal location) to support multi-path comparison; spatial interpolation and temporal smoothing of index data to eliminate the influence of grid differences on sensitivity calculation.

[0072] S530, sensitivity analysis and key trigger node identification: perform sensitivity analysis on the parameter vector of each path to evaluate the degree of influence of parameter changes on .

[0073] Sensitivity calculation can use: Sobol global sensitivity analysis (recommended); local partial derivative method ; Morris method (if acceleration is required).

[0074] Calculate first-order sensitivity index : Where represents the first-order Sobol sensitivity index of input parameter ; represents the i-th of the model input parameters; represents all other parameters except all other parameter sets; R is the model output (i.e. accident consequence indicators such as blast radius, concentration peak, radiant heat dose, etc.); denotes the variance operator, which measures the overall uncertainty of the result; denotes the conditional expectation of other parameters under the fixed ; denotes the variance under the variation of , which can be used to quantitatively evaluate the degree of contribution of input parameters to the uncertainty of output results, so as to identify key sensitive parameters and optimize the accuracy of accident scene simulation; Sensitivity mapping is performed on the nodes in the graph, and the trigger nodes corresponding to high sensitivity parameters are marked as "key nodes". The number of typical key nodes is 3-10; when the sensitivity index , it is determined that the node is a "high-impact node".

[0075] S540, minimum cut set solving: based on the accident chain knowledge graph, a directed weighted graph G=(V,E) is established, where V is the trigger node set, and E is the weighted causal relationship edge.

[0076] Taking the path between the accident starting node and the consequence node as the object, the minimum cut set is solved by using the maximum flow-minimum cut theorem: In detail: in order to identify the key control nodes in the accident propagation network, a minimum cut set solving model is introduced. For the accident propagation graph G, the cut value of the cut set is defined, which represents the total sum of propagation edge weights across the cut set after cutting the node set C; The above can obtain the minimum cut set , that is, the key node set that blocks the accident propagation at the minimum cut cost. This method can be used to identify the weakest link of risk propagation, and provide quantitative basis for accident prevention and control strategy and emergency decision-making; The nodes in the minimum cut set represent the key control points that can interrupt the propagation of the accident chain once blocked; the solving algorithm can use Edmonds-Karp or Push-Relabel algorithm, and the calculation time is usually less than 1s / path; the nodes in are labeled as "key control nodes", and cross-verified with the sensitivity analysis results to eliminate false positives or weak impact nodes.

[0077] S550, treatment suggestion sequence generation and priority setting: according to the key trigger nodes and the minimum cut set, a treatment action suggestion sequence A is automatically generated; the treatment suggestion is one-to-one corresponding to the node type, for example: Node "tank leakage" -> action "close feed valve + start pressure relief"; Node "gas diffusion" -> action "start spray + dilution ventilation"; Node "flame propagation" -> action "start fire extinguishing system + isolate area".

[0078] According to the sensitivity index and the position (hierarchical depth) of the node in the accident chain, the action priority is calculated: where, represents the sensitivity index of the node, which characterizes the sensitivity of the node to the accident propagation consequences or uncertainty indicators; represents the depth of the node in the accident propagation topology, reflecting its distance from the initial source of propagation level; and are weight coefficients for balancing the relative importance of sensitivity and topological factors in priority evaluation, typically 0.7, 0.3; By calculating and sorting the value, the key nodes with high sensitivity and significant impact on accident propagation and consequences can be identified in priority, and the early nodes located upstream of the accident source or at critical propagation levels.

[0079] The output results include: the minimum cut set ; corresponding action recommendations A; priority list ; action location and time recommendations (such as closing valves within 30s, starting spray within 60s); the results are output in JSON / database structured format, providing a direct execution interface for upper system linkage (DCS / ECS / SCADA).

[0080] S6 specifically includes the following sub-steps: S610, accident deduction result output: the final robust target path set , the minimum cut set , the disposal action sequence A, the priority list , the accident consequence field and the consequence indicators are packaged into structured data objects; The data structure adopts JSON format or relational database table, and typical fields include: path_id (path ID); node_id (key node ID); action (treatment suggestion); priority (action priority); effect_radius (effect radius); time_window (recommended treatment time limit); confidence (path confidence); robustness (robustness index); hazard_index (risk level); The output frequency is determined according to the emergency response time limit, and the typical frequency is 1-5 Hz, which ensures that the system can meet the second-level treatment response.

[0081] The data output module is physically separated from the computing module, and high reliability transmission is realized through a message bus (such as MQTT, Kafka or ZeroMQ).

[0082] S620, linkage upper system interface: the communication between the upper system (DCS / ECS / SCADA) adopts industrial protocol OPCUA or Modbus-TCP, and the transmission delay is controlled within 2s.

[0083] The interface definition includes: receiving fields: action_id, trigger_time, execution_status; sending fields: path_id, node_id, action, priority, confidence.

[0084] The linkage strategy adopts an "event-action" mechanism, when the system output treatment suggestion meets the preset trigger condition, the corresponding linkage instruction is automatically triggered, such as: automatically closing the leakage source valve; starting the spraying and ventilation system; issuing an alarm to the personnel evacuation broadcast system; the interface supports synchronous and asynchronous communication modes, which can be dynamically switched according to the severity of the accident; the execution result of the linkage action is returned through the interface, which is used to update the system state and subsequent decision-making.

[0085] S630, data archiving and version control: all intermediate data, simulation results and linkage action logs generated during accident deduction are archived.

[0086] The archiving format adopts HDF5 or Parquet binary format, and the typical single-scene storage capacity is 10-50MB; a unique version number version_id and a check code hash_code are generated each time the accident deduction is performed, which is used to ensure data integrity and traceability.

[0087] The archived content includes: all path score results and perturbation parameters; simulation parameters and boundary conditions; key nodes and minimum cut sets ; handling action execution log and feedback information; final output linkage record; support automatic compression and block storage, ensure archiving efficiency and retrieval performance.

[0088] S640, iteration convergence and rollback record: record the key intermediate state in the iteration process of path screening and robustness evaluation, including the number of paths in each iteration, the change of score , the change of robustness index , the change of key node set The record format uses time series database (such as InfluxDB) or text index database (such as Elasticsearch) for easy retrieval and visualization; each iteration automatically generates an "iteration snapshot", including: iteration round number; pass / eliminate path ID; convergence error epsilon; rollback trigger reason; calculation time; if trigger rollback mechanism B, roll back the snapshot to the last stable convergence point, ensure system stability and result consistency.

[0089] S650, form a closed-loop linkage and trace mechanism: through the result output of S610-S640, linkage trigger and log archiving, realize the "prediction-linkage-record-trace" closed-loop emergency response system.

[0090] The upper system can perform event review, emergency drill and model retraining based on the archived data, forming a continuous improvement cycle; this method supports online linkage and post-tracing of the whole process, with typical response time <5s, and trace records can be preserved for ≥5 years; all operations have timestamp signature and digital signature mechanism, ensuring the traceability and evidence effectiveness of the accident handling process; the closed-loop linkage mechanism is designed integrally with the technical links of S110-S550, ensuring that the deduction results can be executed in real time, avoiding the problem of "staying in the calculation results".

[0091] Path extrapolation in this embodiment means using accident chain knowledge and physical models to automatically extrapolate multiple possible evolution routes from the current state of the accident, and selecting credible paths through physical and mathematical calculations to provide decision basis for emergency response and risk control.

[0092] The above formulas are dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the latest real situation, and the preset parameters and threshold values in the formula are set by the person skilled in the art according to the actual situation.

[0093] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0094] Those skilled in the art can appreciate that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0095] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device, and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0096] In several embodiments provided in the present application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed each other can be indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0097] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0098] In addition, the functional modules in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0099] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0100] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for simulating and calculating the consequences of an accident based on the coupled solution of a mathematical-physical model, characterized in that, Includes the following steps: S1. Establish an accident chain knowledge graph based on process flow and accident history information to represent the causal relationship between equipment, status, events and consequences; combine parameter thresholds and conservation constraints to establish a physical trigger graph to form accident triggering logic and physical reachability constraints. S2. Map historical accident records, expert rules, and online observation data into accident chain knowledge graph entities and relationships. After triggering the path, call the coupled multiphysics solver to establish a baseline accident consequence field and calculate the baseline consequence index as an extrapolation reference. S3. Using graph causal triggering relationships to traverse and reason to form candidate extrapolation paths, the paths are mapped to simulation boundary conditions to solve the accident consequence field. Consistency is checked through mass conservation, energy conservation and trigger satisfaction, realizing closed-loop iteration of graph reasoning and physical solution. S4. The candidate paths are comprehensively scored based on consistency index, consequence index and path complexity. The uncertainty interval and robustness index are obtained by perturbation simulation and statistical analysis, and the final robust target path is selected. S5. Perform high-precision physical field simulation for robust target paths, calculate accident consequence indicators, identify key triggering nodes in the accident chain by combining sensitivity analysis and minimum cut-off set solution, and generate disposal suggestions and action priority sequences.

2. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 1, characterized in that, It also includes S6, which outputs the path, key nodes, consequences indicators and handling suggestions to the upper system in a standardized data format, realizes automatic linkage handling through industrial protocols, and archives and versions the entire process of accident simulation to complete closed-loop emergency response and post-event traceability.

3. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 1, characterized in that, S1 specifically includes: Establish an incident chain knowledge graph to represent the structure of equipment, state, events, and causal relationships; Establish a physical triggering graph to represent threshold conditions, conservation constraints, and triggering logic; Establish a coupled multiphysics solver to calculate the accident consequence field based on source term parameters and boundary conditions; Define the accident scenario parameter vector, extrapolation path set, consequence index, consistency check index, convergence criterion and constraint set; Configure a rollback mechanism and linkage with the higher-level system for recovery in case of anomalies.

4. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 1, characterized in that, S2 specifically includes: Historical records, expert rules, and real-time observations are mapped to entities, relationships, and initial triggering conditions in an incident chain knowledge graph. Input the accident scenario parameter vector into the physical trigger diagram to complete the parameterization of the trigger logic; Call the coupled multiphysics solver to calculate the baseline accident consequence field based on parameters and triggering conditions; Calculate and record baseline consequence metrics as a benchmark for subsequent extrapolation path comparisons; Store the baseline state for reference in subsequent filtering and rollback.

5. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 1, characterized in that, S3 specifically includes: Candidate extrapolation paths are generated based on causal relationships in the accident chain knowledge graph. The candidate extrapolation path is mapped to the boundary conditions and source terms of the coupled multiphysics solver to calculate the simulated accident consequence field. Calculate the simulated consequences index and perform physical reachability and conservation consistency checks on the candidate extrapolation paths based on the consistency check index; The verification results are used to update the relationship weights in the accident chain knowledge graph and the triggering conditions in the physical triggering graph. A bidirectional iterative link between graph reasoning and physical decomposition is formed to continuously optimize the credibility of candidate extrapolation paths.

6. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 1, characterized in that, S4 specifically includes: Within the constraint set, a comprehensive score is calculated for candidate extrapolation paths based on consistency verification indicators and consequence indicators; A sequential filtering method is used to retain the set of target extrapolation paths; Perform perturbation simulations on the target extrapolation path set within the uncertainty range of the accident scenario parameter vector; Evaluate the robustness of the target extrapolation path and the range characteristics of the consequence indicators; If the convergence criterion is not met or the constraint set is violated, a backoff mechanism is triggered to adjust the relation weights, triggering conditions, or extrapolation paths, and the process returns to S3 for re-iteration.

7. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 1, characterized in that, S5 specifically includes: The coupled multiphysics solver is invoked to perform a high-fidelity coupled solution on the target extrapolation path set; Obtain the target accident consequence scene and target consequence indicators; Key triggering nodes in the accident chain knowledge graph are identified based on sensitivity analysis of target consequence indicators. The minimum cut-off set is extracted based on graph structure centrality to characterize critical accident control points; A sequence of proposed actions and their corresponding priorities are generated by combining the target extrapolation path.

8. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 2, characterized in that, S6 specifically includes: Output the target extrapolation path set, key triggering nodes, minimum cutoff set, target accident consequence field, and target consequence indicators to the upper-level system; Output a sequence of suggested actions and their corresponding priorities for coordinated control and emergency response. Archived graph status, trigger graph parameters, solver results, and consistency verification evidence.

9. The method for simulating and calculating accident consequences based on the coupled solution of a mathematical-physical model according to claim 8, characterized in that, S6 also includes: recording the extrapolation path iterative convergence process and backtracking trajectory for online calibration and traceability analysis; This forms a closed loop of knowledge graph, physical trigger, and numerical solution, providing a reusable simulation basis for subsequent accident early warning and handling.

Images