A comprehensive pipe gallery intelligent monitoring risk coupling analysis method

Abstract

The application discloses a kind of comprehensive pipe gallery wisdom monitoring risk coupling analysis methods.It includes the following steps: S1: constructing comprehensive pipe gallery operation safety risk field ontology knowledge base;S2: risk quantitative evaluation is carried out based on improved FMEA method;S3: the risk propagation dynamic model based on SEIRS model is constructed;S4: risk propagation path search is carried out based on ant colony algorithm, and the key propagation path set that system risk influence is maximum is obtained;S5: model parameter optimization is carried out based on particle swarm algorithm;S6: risk coupling simulation and sensitivity analysis are carried out based on system dynamics.It is advantageous that: by constructing the risk propagation dynamic model based on SEIRS model, the modeling of the propagation process of risk between different nodes is realized, the dynamic evolution law that risk state changes with time can be described, so as to make up for the deficiency that traditional FMEA method can only carry out static evaluation, improve the time sequence and dynamics of risk analysis.

Description

Technical Field

[0001] This invention relates to the field of intelligent computing and complex system modeling technology, and in particular to a method for risk coupling analysis of intelligent monitoring of integrated utility tunnels. Background Technology

[0002] With the continuous expansion of urban scale and the increasing complexity of utility tunnel systems, the types of equipment inside the tunnels are numerous, the operating environment is complex and changeable, and various risk factors are intertwined, exhibiting significant coupling and dynamic evolution characteristics, posing a significant challenge to traditional operation and maintenance monitoring methods.

[0003] Existing technologies in the field of integrated utility tunnel risk monitoring still have the following problems: First, risk assessment methods tend to focus on static analysis, making it difficult to reflect the dynamic propagation process of risks; second, model parameters rely on experience settings and lack adaptive optimization capabilities; third, there is a lack of effective means to identify the coupling relationship and propagation path between multiple risk factors; and fourth, the utilization of knowledge resources is low, making it difficult to achieve systematic and intelligent risk analysis. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a comprehensive intelligent monitoring risk coupling analysis method for utility tunnels that improves the timeliness and dynamism of risk analysis and the pertinence of risk early warning and control.

[0005] To address the aforementioned technical problems, the integrated utility tunnel intelligent monitoring risk coupling analysis method of the present invention includes the following steps:

[0006] S1: Construct a domain ontology knowledge base for the operation and maintenance safety risks of integrated utility tunnels, collect data from integrated utility tunnels, establish a domain ontology model, and transform the domain ontology model into a risk coupling graph structure through semantic mapping rules. The risk coupling graph is as follows:

[0007] ;

[0008] Where V represents the set of risk nodes, E represents the set of risk-coupled edges, and ω ij η represents the coupling weight between node i and node j. ij This represents the risk propagation probability from node i to node j; the risk coupling graph is used as the basic data structure for subsequent risk quantification assessment and propagation modeling.

[0009] S2: Quantitative risk assessment based on an improved FMEA method, for each risk node v i The corresponding failure modes are analyzed, their risk priority values ​​are calculated, and the risk priority values ​​are normalized. The risk priority values ​​are used as the initial weights of the nodes in the risk coupling graph.

[0010] The calculation formula is as follows:

[0011] ;

[0012] Among them, Q i Let α represent the risk priority value of the i-th risk node. i β represents the failure severity coefficient. i γ represents the failure frequency coefficient. i The failure detectability coefficient is represented by the risk priority value, which is then input into the risk propagation model as an importance representation parameter of the risk node.

[0013] S3: Construct a dynamic risk propagation model based on the SEIRS model, mapping each risk node to a set of propagation states, and establish a risk propagation state transition model:

[0014] ;

[0015] Among them, S i (t) represents the susceptible state, E i (t) represents the exposed state, I i (t) represents the infection state, R i (t) represents the recovery state, λ0 represents the risk propagation coefficient, δ represents the risk activation rate, μ represents the risk recovery rate, and σ represents the conversion coefficient for the risk to re-enter the susceptible state.

[0016] The risk propagation coefficient λ0 is corrected based on the risk priority value Qi, and the time-varying propagation coefficient λij(t) is constructed by combining the coupling relationship between risk nodes and the differences in risk states:

[0017]

[0018] The risk propagation probability ηij of edge (i,j) is determined based on the time-varying propagation coefficient λij(t) and used for subsequent path search model;

[0019] S4: Risk propagation path search is performed based on the ant colony algorithm. A path search model is constructed in the risk coupling graph, and a path evaluation function is defined.

[0020] ;

[0021] Among them, F k P represents the comprehensive evaluation value of the k-th risk transmission path. k Let θ represent the set of edges contained in the k-th path. ij η represents the pheromone concentration on edge (i,j). ij Let denot (i,j) represent the risk propagation probability of edge (i,j), ρ represent the pheromone influence factor, κ represent the heuristic function influence factor, and ξ represent the risk priority modulation factor.

[0022] Path selection probability:

[0023] ;

[0024] Where s represents the index of the neighboring node of node i, Ω i Represents the set of optional adjacent nodes of node i;

[0025] The path is iteratively optimized using pheromone update rules:

[0026] ;

[0027] ;

[0028] Where, θ ij (t) represents the pheromone concentration at the t-th iteration. Δθ represents the pheromone evaporation coefficient. ij Indicates the incremental contribution of the path;

[0029] Obtain the set of critical propagation paths that have the greatest impact on system risk, and use the critical propagation paths as important constraint information for model parameter optimization;

[0030] S5: Model parameter optimization based on particle swarm optimization (PSO) algorithm. The PSO algorithm is used to jointly optimize the model parameters constituting the time-varying propagation coefficient λij(t) and the activation rate δ and recovery rate μ in the SEIRS model, constructing an optimized model with risk prediction error as the objective function.

[0031] ;

[0032] Where J represents the global error function, U i This represents the actual risk value of the i-th monitoring point. This represents the model prediction value calculated by the model, and N represents the number of monitoring points;

[0033] The optimal parameter combination is obtained and the optimized parameters are fed back into the risk propagation model for updating.

[0034] S6: Based on system dynamics, perform risk coupling simulation and sensitivity analysis, construct a system dynamics model, simulate the risk evolution process, and obtain the overall system risk level function:

[0035] ;

[0036] Where R(t) represents the overall system risk level at time t, Q i S represents the risk priority value of the i-th risk node. i(t) represents the state strength of the i-th risk node at time t, and M represents the total number of risk nodes;

[0037] By performing perturbation analysis on the model parameters, the impact of each parameter on the system risk level is calculated, and key risk factors and main risk propagation paths are identified.

[0038] The data collected in step S1 in the integrated utility tunnel includes equipment units, environmental status, operating conditions and historical accident data; the domain ontology model includes risk source entities, risk event entities and their relationships.

[0039] In step S3, based on the risk priority value Q i The process of correcting the propagation coefficient λ specifically includes:

[0040] ① Construct the corrected propagation coefficient λ for node association i ′, its expression is:

[0041] ;

[0042] Where, λ i ′ represents the modified propagation coefficient corresponding to the i-th risk node, λ0 represents the basic propagation coefficient, and ϵ i Q represents the sensitivity adjustment factor for the i-th risk node. i This represents the risk priority value of the i-th risk node;

[0043] ② Further, by combining the coupling relationship between risk nodes, the propagation coefficient is modified at the edge level to obtain the inter-node propagation coefficient λ. ij '', its expression is:

[0044] ;

[0045] Where, λ ij λ'' represents the correction coefficient propagating from node i to node j, λi' represents the corrected propagation coefficient of node i, and ϕ ij ψi represents the structural coupling strength between node i and node j, and ψj represents the susceptibility weight factor of node j.

[0046] ③ The propagation coefficient is dynamically updated over time to construct the time-varying propagation coefficient λ. ij (t), whose expression is:

[0047] ;

[0048] Where, χ ij Indicates the dynamically adjusted weighting coefficient, Δ ij (t) represents the risk state difference function between nodes;

[0049] ④ The risk state difference function is defined as follows:

[0050] ;

[0051] Among them, S i (t) represents the risk state intensity of node i at time t, S j (t) represents the risk state intensity of node j at time t.

[0052] In the risk propagation path search in step S4, a risk priority value is introduced to modulate the pheromone increment, so that high-risk paths obtain higher pheromone accumulation:

[0053] ;

[0054] Where L k This represents the path cost of the k-th path.

[0055] Advantages of this invention:

[0056] 1. By constructing a risk propagation dynamic model based on the SEIRS model, the modeling of the risk propagation process between different nodes is realized, which can depict the dynamic evolution of risk status over time, thereby making up for the shortcomings of the traditional FMEA method which can only perform static assessments, and improving the temporal and dynamic nature of risk analysis.

[0057] 2. Particle swarm optimization algorithm is introduced to adaptively optimize key parameters in the risk propagation model. By minimizing the error between the model's predicted values ​​and the actual monitoring data, the model parameters are automatically calibrated, reducing the reliance on manual experience-based assignment and improving the model's accuracy and adaptability.

[0058] 3. By constructing a domain ontology knowledge base and mapping it to a risk coupling graph structure, a structured expression of risk nodes and their relationships is realized. At the same time, by performing node-level, edge-level, and dynamic multi-level corrections on the propagation coefficients, the risk propagation capability can reflect the characteristics of nodes and the coupling relationships between nodes, thereby improving the ability to characterize complex risk coupling relationships.

[0059] 4. By using the ant colony algorithm to perform a global search of the propagation paths in the risk coupling graph and modulating the path evaluation in combination with the risk priority value, the key risk propagation paths can be identified, thereby locating the risk links that have the greatest impact on the overall system risk and improving the pertinence of risk warning and control.

[0060] 5. Based on the system dynamics model, the risk evolution process is simulated and analyzed. Combined with the sensitivity analysis of parameter disturbances, the key factors affecting the system risk level and the main propagation paths can be identified, thus providing a basis for the formulation of risk control strategies. Attached Figure Description

[0061] Figure 1 This is a flowchart of the integrated utility tunnel intelligent monitoring risk coupling analysis method of the present invention. Detailed Implementation

[0062] The risk coupling analysis method for intelligent monitoring of integrated utility tunnels of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0063] Example:

[0064] The integrated utility tunnel intelligent monitoring risk coupling analysis method includes the following steps:

[0065] S1: Construct an ontology knowledge base for the safety risks of integrated utility tunnel operation and maintenance, collecting data from the integrated utility tunnel, including equipment units, environmental status, operating conditions, and historical accident data; establish a domain ontology model containing risk source entities, risk event entities, and their relationships, and transform the domain ontology model into a risk coupling graph structure through semantic mapping rules. This achieves structured expression and relational modeling of risk knowledge, providing unified data support for risk analysis, improving knowledge reusability and analysis efficiency. The risk coupling graph is as follows:

[0066] ;

[0067] Where V represents the set of risk nodes, E represents the set of risk-coupled edges, and ω ij η represents the coupling weight between node i and node j. ij This represents the probability of risk propagation from node i to node j;

[0068] S2: Risk quantification assessment is conducted based on an improved FMEA (Failure Mode and Effects Analysis) method, targeting each risk node v. i The corresponding failure modes are analyzed, their risk priority values ​​are calculated, and the risk priority values ​​are normalized and used as the initial weights of the nodes in the risk coupling graph.

[0069] The calculation formula is as follows:

[0070] ;

[0071] Among them, Q i Let α represent the risk priority value of the i-th risk node. i β represents the failure severity coefficient. i γ represents the failure frequency coefficient. i This represents the failure detectability coefficient;

[0072] Determination of the severity coefficient α1 of failure: The impact of risk events on the operational safety of the integrated utility tunnel is assessed in a graded manner, using a scoring standard of 1 to 10, where: 1 indicates minimal impact and no impact on system operation; 5 indicates some impact on the local system; and 10 indicates potential serious accidents or system failure. The specific scoring criteria include factors such as the type of accident consequences, the scope of impact, and the difficulty of recovery.

[0073] Failure frequency coefficient β i Determination: Based on historical operational data and monitoring data, the frequency of risk events is statistically analyzed and quantified by combining experience-based classification. Specifically, this includes: counting the number of times such risk events occur per unit time; and mapping the frequency of occurrence to a scoring range of 1 to 10.

[0074] Failure detectability coefficient γ i Determination: The risk is assessed based on the ease with which it is detected by the monitoring system, specifically including: sensor coverage; detection sensitivity; alarm response time; the scoring rule is: 1 indicates extremely easy to detect; 10 indicates extremely difficult to detect or impossible to detect.

[0075] The above methods enable a mechanism that combines empirical assessment with data analysis to obtain parameters such as failure severity, frequency of occurrence, and detectability, thereby improving the objectivity and accuracy of risk assessment results.

[0076] S3: Construct a dynamic risk propagation model based on the SEIRS model, mapping each risk node to a set of propagation states, and establish a risk propagation state transition model:

[0077] ;

[0078] Among them, S i (t) represents the susceptible state, E i (t) represents the exposed state, I i (t) represents the infection state, R i (t) represents the recovery state, λ0 represents the risk propagation coefficient, δ represents the risk activation rate, μ represents the risk recovery rate, and σ represents the conversion coefficient for the risk to re-enter the susceptible state.

[0079] Based on risk priority value Q i The risk propagation coefficient λ0 is corrected;

[0080] The time-varying propagation coefficient λij(t) is constructed by combining the coupling relationship between risk nodes and the differences in risk states:

[0081]

[0082] Among them, ϵ i Let ϕ represent the sensitivity adjustment factor for the i-th risk node.ij ψj represents the structural coupling strength between node i and node j, ψj represents the susceptibility weight factor of node j, and χj represents the structural coupling strength between node i and node j. ij Indicates the dynamically adjusted weighting coefficient, Δ ij (t) represents the risk state difference function between nodes;

[0083] The risk propagation probability ηij is determined based on the time-varying propagation coefficient λij(t) and used for subsequent path search models;

[0084] The corrected propagation coefficient λ for constructing node associations i ′, its expression is:

[0085] ;

[0086] Where, λ i ′ represents the modified propagation coefficient corresponding to the i-th risk node, λ0 represents the basic propagation coefficient, and ϵ i Q represents the sensitivity adjustment factor for the i-th risk node. i This represents the risk priority value of the i-th risk node;

[0087] Further considering the coupling relationship between risk nodes, the propagation coefficient is modified at the edge level to obtain the inter-node propagation coefficient λ. ij '', its expression is:

[0088] ;

[0089] Where, λ ij λ'' represents the correction coefficient propagating from node i to node j, λi' represents the corrected propagation coefficient of node i, and ϕ ij ψi represents the structural coupling strength between node i and node j, and ψj represents the susceptibility weight factor of node j.

[0090] The propagation coefficient is dynamically updated over time, constructing a time-varying propagation coefficient λ. ij (t), whose expression is:

[0091] ;

[0092] Where, λ ij (t) represents the propagation coefficient from node i to node j at time t, λ ij χ' represents the propagation coefficient after static correction. ij Indicates the dynamically adjusted weighting coefficient, Δ ij (t) represents the risk state difference function between nodes;

[0093] The risk status difference function is defined as follows:

[0094] ;

[0095] Among them, S i (t) represents the risk state intensity of node i at time t, S j (t) represents the risk state intensity of node j at time t.

[0096] The time-varying propagation coefficient λij(t) is used to characterize the difference in the dynamic propagation ability of risk between different nodes, to update the parameters of the risk propagation model, and to feed back to the path search module to achieve dynamic closed-loop optimization of risk propagation analysis.

[0097] S4: Use the ant colony algorithm to perform path search on the risk coupling graph to identify key risk propagation paths, construct a path search model in the risk coupling graph, and define a path evaluation function:

[0098] ;

[0099] Among them, F k P represents the comprehensive evaluation value of the k-th risk transmission path. k Let θ represent the set of edges contained in the k-th path. ij η represents the pheromone concentration on edge (i,j). ij Let denot (i,j) represent the risk propagation probability, ρ represent the pheromone influence factor, κ represent the heuristic function influence factor, and ξ represent the risk priority modulation factor.

[0100] Where the risk propagation probability η ij The time-varying propagation coefficient λ in the propagation model ij (t) is mapped to show that, during the path search process, high-risk paths are strengthened through the pheromone update mechanism, so that key risk propagation paths can be identified first.

[0101] The path is iteratively optimized using pheromone update rules:

[0102] ;

[0103] ;

[0104] Where θij(t) represents the pheromone concentration at the t-th iteration. L represents the pheromone evaporation coefficient, Δθij represents the path contribution increment; k Let L represent the path cost of the k-th path. k The risk is determined based on the risk transmission capacity of each side of the path; the stronger the transmission capacity, the smaller the corresponding cost.

[0105] The set of critical propagation paths that have the greatest impact on system risk is obtained. In the risk propagation path search, a risk priority value is introduced to modulate the pheromone increment, so that high-risk paths can obtain higher pheromone accumulation, thereby improving the identification accuracy of critical risk paths.

[0106] Through the above improvements, this invention achieves deep coupling of risk priority, propagation capability and path search process, enabling the ant colony algorithm to quickly converge to the critical propagation path with the greatest impact on system risk in complex risk networks, thereby improving the accuracy and stability of path identification.

[0107] S5: Model parameter optimization based on particle swarm optimization (PSO) algorithm. The PSO algorithm is used to jointly optimize the propagation coefficient λ, activation rate δ, and recovery rate μ in the SEIRS model, constructing an optimized model with risk prediction error as the objective function.

[0108] ;

[0109] Where J represents the global error function, U i This represents the actual risk value of the i-th monitoring point. This represents the model prediction value calculated by the model, and N represents the number of monitoring points;

[0110] The propagation coefficient λ, activation rate δ, and recovery rate μ are optimized and updated using the particle swarm optimization algorithm to obtain the optimal parameter combination;

[0111] S6: Based on system dynamics-based risk coupling simulation and sensitivity analysis, a system dynamics model is constructed to simulate the risk evolution process and obtain the overall system risk level function.

[0112] ;

[0113] Where R(t) represents the overall system risk level at time t, Q i S represents the risk priority value of the i-th risk node. i (t) represents the state strength of the i-th risk node at time t, and M represents the total number of risk nodes;

[0114] By performing perturbation analysis on the model parameters, the impact of each parameter on the system risk level is calculated, and key risk factors and main risk propagation paths are identified.

[0115] Specific implementation:

[0116] Using an urban underground utility tunnel as an example, the tunnel is designed to include an electrical compartment, a gas compartment, and a communication compartment, and is equipped with various monitoring devices such as temperature sensors, gas concentration sensors, and humidity sensors.

[0117] Step 1: Ontology Modeling

[0118] Collect equipment operation data, environmental parameters, and historical accident records to construct a risk ontology, including:

[0119] Risk sources: cable aging, gas leak; risk events: fire, explosion; relationships: causal relationships, spatial adjacency relationships; and are transformed into a risk coupling diagram.

[0120] Step 2: FMEA Analysis

[0121] Assessment of the "gas leak" failure mode:

[0122] Severity α1=9, incidence β1=6, detectability γ1=5;

[0123] Calculate: Q1 = 9 × 6 × 5 = 270;

[0124] Step 3: Propagation Modeling

[0125] (1) Node-level propagation coefficient correction: Let the basic propagation coefficient be λ0 = 0.2; for the "gas leak" risk node, its risk priority value is Q1 = 270; let the sensitivity adjustment factor be ϵ1 = 0.01; then the node-level corrected propagation coefficient is:

[0126] λ1′=λ0(1+ϵ1Q1)=0.2×(1+2.7)=0.74;

[0127] (2) Correction of the edge propagation coefficient: Considering the coupling relationship between nodes, let node i: gas leakage; node j: gas accumulation;

[0128] Let: ϕ ij =0.8 (structural coupling strength); ψ j =1.2 (node ​​susceptibility weight); then the propagation coefficient between nodes is:

[0129] λ ij ′′=λ i ′·ϕ ij ·ψ j =0.74×0.8×1.2=0.7104;

[0130] (3) Correction of dynamic propagation coefficient

[0131] At time t, let X: i (t) = 0.9 (Intensity of gas leak risk status); X j (t) = 0.4 (Intensity of gas accumulation risk state).

[0132] The state difference function is then: Δ ij (t)=∣S i (t)−S j(t)∣=0.5;

[0133] Let the dynamic adjustment coefficient be: χ ij =0.6; then the time-varying propagation coefficient is:

[0134] λ ij (t)=0.7104×(1+0.6×0.5)=0.92352;

[0135] (4) Application of the propagation model

[0136] The time-varying propagation coefficient λ ij (t) Substituted into the SEIRS model, it is used to describe the dynamic propagation process of risk between different nodes, thereby achieving a refined characterization of risk diffusion behavior.

[0137] Through the above-mentioned multi-level correction mechanism of node level, edge level and dynamic propagation coefficient, the refined modeling of risk propagation capability is realized, so that the model can more realistically reflect the coupled propagation characteristics of risks in the integrated utility tunnel.

[0138] Step 4: Path Search

[0139] Using the ant colony algorithm for search: After multiple iterations, pheromones gradually concentrate on a few high-quality paths, the algorithm converges, and finally selects the path with the highest evaluation value as the critical risk propagation path, thus discovering the critical path:

[0140] Gas leak → gas accumulation → explosion; this path has the highest evaluation value.

[0141] Step 5: Parameter Optimization

[0142] Based on actual monitoring data: Measured risk value U i Model predictions The particle swarm optimization algorithm is used to minimize the error function J.

[0143] Step 6: System Simulation

[0144] Simulation results show that the gas leak node contributes the most to the overall risk of the system, and there is a significant coupling amplification effect between adjacent nodes. This verifies the effectiveness of the constructed risk propagation model and path search method, which can accurately identify key risk paths and dominant risk factors.

Claims

1. A method for comprehensive pipe gallery intelligent monitoring risk coupling analysis, characterized in that, Includes the following steps: S1: Construct a domain ontology knowledge base for the operation and maintenance safety risks of integrated utility tunnels, collect data from the integrated utility tunnels, establish a domain ontology model, and transform the domain ontology model into a risk coupling graph structure through semantic mapping rules. The risk coupling graph is as follows: ； wherein V represents a risk node set, E represents a risk coupling edge set, ω ij represents a coupling weight between node i and node j, η ij represents a risk propagation probability from node i to node j; taking the risk coupling graph as a basic data structure for subsequent risk quantification assessment and propagation modeling; S2: Quantitative risk assessment based on an improved FMEA method, for each risk node v i The corresponding failure modes are analyzed, their risk priority values ​​are calculated, and the risk priority values ​​are normalized. The risk priority values ​​are used as the initial weights of the nodes in the risk coupling graph. The calculation formula is as follows: ； Among them, Q i Let α represent the risk priority value of the i-th risk node. i β represents the failure severity coefficient. i γ represents the failure frequency coefficient. i The failure detectability coefficient is represented by the risk priority value, which is then input into the risk propagation model as an importance representation parameter of the risk node. S3: Construct a dynamic risk propagation model based on the SEIRS model, mapping each risk node to a set of propagation states, and establish a risk propagation state transition model: ； Among them, S i (t) represents the susceptible state, E i (t) represents the exposed state, I i (t) represents the infection state, R i (t) represents the recovery state, λ0 represents the risk propagation coefficient, δ represents the risk activation rate, μ represents the risk recovery rate, and σ represents the conversion coefficient for the risk to re-enter the susceptible state. The risk propagation coefficient λ0 is corrected based on the risk priority value Qi, and the time-varying propagation coefficient λij(t) is constructed by combining the coupling relationship between risk nodes and the differences in risk states: ； Among them, ϵ i Let ϕ represent the sensitivity adjustment factor for the i-th risk node. ij ψj represents the structural coupling strength between node i and node j, ψj represents the susceptibility weight factor of node j, and χj represents the structural coupling strength between node i and node j. ij Indicates the dynamically adjusted weighting coefficient, Δ ij (t) represents the risk state difference function between nodes; The risk propagation probability ηij of edge (i,j) is determined based on the time-varying propagation coefficient λij(t) and used for subsequent path search model; S4: Risk propagation path search is performed based on the ant colony algorithm. A path search model is constructed in the risk coupling graph, and a path evaluation function is defined. ； Among them, F k P represents the comprehensive evaluation value of the k-th risk transmission path. k Let θ represent the set of edges contained in the k-th path. ij η represents the pheromone concentration on edge (i,j). ij Let denot (i,j) represent the risk propagation probability of edge (i,j), ρ represent the pheromone influence factor, κ represent the heuristic function influence factor, and ξ represent the risk priority modulation factor. Path selection probability: ； where s denotes the index of the neighboring node of node i, Ω i denotes the set of optional neighboring nodes of node i; The path is iteratively optimized using pheromone update rules: ； ； where θ ij (t) denotes the pheromone concentration at the tth iteration, denotes the pheromone evaporation coefficient, Δθ ij denotes the path contribution increment; Obtain the set of critical propagation paths that have the greatest impact on system risk, and use the critical propagation paths as important constraint information for model parameter optimization; S5: Model parameter optimization based on particle swarm optimization (PSO) algorithm. The PSO algorithm is used to jointly optimize the model parameters constituting the time-varying propagation coefficient λij(t) and the activation rate δ and recovery rate μ in the SEIRS model, constructing an optimized model with risk prediction error as the objective function. ； wherein J represents an overall error function, U i represents the actual risk value of the i-th monitoring point, represents the model prediction value calculated by the model, and N represents the number of monitoring points. The optimal parameter combination is obtained and the optimized parameters are fed back into the risk propagation model for updating. S6: Based on system dynamics, perform risk coupling simulation and sensitivity analysis, construct a system dynamics model, simulate the risk evolution process, and obtain the overall system risk level function: ； wherein R(t) represents the total risk level of the system at time t, Q i represents the risk priority value of the i-th risk node, S i (t) represents the state intensity of the i-th risk node at time t, and M represents the total number of risk nodes. By performing perturbation analysis on the model parameters, the impact of each parameter on the system risk level is calculated, and key risk factors and main risk propagation paths are identified. 2.The method of claim 1, wherein: The data collected in step S1 in the integrated utility tunnel includes equipment units, environmental status, operating conditions and historical accident data; the domain ontology model includes risk source entities, risk event entities and their relationships. 3.The method of claim 1, wherein: In the step S3, the risk priority value Q is calculated according to the following equation: i The process of modifying the propagation coefficient λ specifically includes: ① the modified propagation coefficient λ associated with the construction node i , which is expressed as: ； Where, λ i ′ represents the modified propagation coefficient corresponding to the i-th risk node, λ0 represents the basic propagation coefficient, and ϵ i Q represents the sensitivity adjustment factor for the i-th risk node. i This represents the risk priority value of the i-th risk node; ②Further combined with the coupling relationship between the risk nodes, the edge-level correction is made to the propagation coefficient λ ij The expression is: ; Where, λ ij λ'' represents the correction coefficient propagating from node i to node j, λi' represents the corrected propagation coefficient of node i, and ϕ ij ψi represents the structural coupling strength between node i and node j, and ψj represents the susceptibility weight factor of node j. iii. said propagation coefficient is dynamically updated over time, building a time-varying propagation coefficient λ ij (t), whose expression is: ； Where, χ ij Indicates the dynamically adjusted weighting coefficient, Δ ij (t) represents the risk state difference function between nodes; ④ The risk state difference function is defined as follows: ； Among them, S i (t) represents the risk state intensity of node i at time t, S j (t) represents the risk state intensity of node j at time t.

4. The integrated utility tunnel intelligent monitoring risk coupling analysis method according to claim 1, characterized in that: In the risk propagation path search in step S4, a risk priority value is introduced to modulate the pheromone increment, so that high-risk paths obtain higher pheromone accumulation: ； where L k denotes the path cost of the kth path.

Images