A method for post-disaster emergency risk grade evaluation of landslide in tunnel address area based on DPSIR-FISM-BN coupling

Abstract

The application discloses a kind of based on DPSIR-FISM-BN coupling's tunnel site area landslide post-disaster emergency risk grade evaluation method, belong to landslide post-disaster risk assessment field, including the following steps: S1, obtaining risk assessment initial index set;S2, after eliminating redundant index, standardization risk assessment index system is constructed;S3, FISM multistage recursive cause-effect structure is constructed;S4, is converted into bayesian network topological structure;S5, risk state division;S6, the root node priori probability table and child node conditional probability table are calculated;S7, evaluation model is constructed;S8, measured data are input into evaluation model, determine the final risk grade of target area;S9, identify the sensitivity index of influence risk grade.The above-mentioned one kind based on DPSIR-FISM-BN coupling's tunnel site area landslide post-disaster emergency risk grade evaluation method, realizes the accurate grading and key factor identification of mountain railway tunnel site area landslide post-disaster hidden danger area risk, provides scientific, efficient decision-making support for emergency rescue.

Description

Technical Field

[0001] This invention relates to the field of landslide post-disaster risk assessment technology, and in particular to a method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling. Background Technology

[0002] Landslides are one of the most common and widespread geological hazards during the construction and operation of railways in mountainous areas. With the continuous development of western my country and the ongoing construction of railway infrastructure, mountain railway tunnel sites are highly susceptible to landslides due to complex topography, active geological structures, concentrated heavy rainfall, and engineering disturbances. These landslides can lead to problems such as tunnel structural damage, line interruption, regional traffic disruption, and difficulties in emergency rescue.

[0003] Following a landslide, the affected area and adjacent sections remain in a state of deformation or near-stability, making them potential emergency rescue zones prone to secondary disasters. Failure to quickly and scientifically assess the risk level of these zones during the emergency phase can directly lead to inaccurate rescue decisions, potentially triggering secondary disasters and threatening the lives of rescue personnel. Against the backdrop of national transportation infrastructure development and continuous improvement in railway safety capabilities, risk identification, grading assessment, and scientific prevention and control in post-landslide emergency rescue in mountainous railway tunnel areas have become core technical requirements for ensuring safe railway operation and enhancing disaster emergency response capabilities.

[0004] Currently, relevant technological achievements have been made in the field of railway landslide disaster risk assessment and emergency management: In terms of railway landslide disaster risk level assessment, domestic and foreign research has gradually developed from early disaster experience judgment and line area hazard analysis to spatial risk assessment based on GIS and remote sensing technologies, as well as landslide susceptibility analysis driven by machine learning and deep learning. Some studies have achieved landslide risk zoning along railway lines through multi-factor comprehensive evaluation; In terms of railway landslide disaster emergency management, existing technologies mainly focus on monitoring and early warning, intelligent emergency decision support, on-site emergency engineering treatment, and the construction of multi-source sensor monitoring platforms, realizing pre-event early warning, post-event response and engineering treatment of railway landslide disasters.

[0005] It is known that the existing technology has the following drawbacks: 1. Insufficient focus of research subjects and disconnect between risk assessment and emergency response: Existing landslide risk assessments mostly focus on the susceptibility and hazard assessment of disaster areas, and railway research is mostly oriented towards the entire line or a large area along the line. There is an extreme lack of specific risk assessment research on landslide emergency response hazard areas in mountainous railway tunnel sites. At the same time, emergency management research focuses on response and governance after disasters occur, and does not closely integrate the risk level assessment of post-disaster hazard areas with emergency response decision-making and allocation of emergency response resources.

[0006] 2. The assessment methods cannot characterize the hierarchy and dynamic relationship of complex risks: Existing railway landslide risk assessment methods tend to focus on indicator weighting and result zoning, making it difficult to clearly express the causal hierarchy, action path and dynamic evolution characteristics between risk factors; the risks of post-disaster hidden danger areas in mountainous railway tunnel sites are affected by the coupling effect of multiple factors such as natural environment, slope condition, engineering structure and emergency response, and a single assessment method cannot fully reflect its risk formation and transmission mechanism.

[0007] 3. Inability to meet the practical needs of emergency rescue after landslides: Existing technologies do not fully consider the secondary, dynamic, and sensitive characteristics of landslide hazard areas, and lack rapid risk assessment models adapted to emergency rescue scenarios; at the same time, existing methods are difficult to quantitatively calculate risk levels and identify sensitivity indicators, and cannot provide accurate risk judgment basis and decision support for front-line emergency rescue, which may easily lead to safety hazards in rescue operations. Summary of the Invention

[0008] The purpose of this invention is to provide a method for assessing the post-disaster emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling, thereby solving the aforementioned technical problems.

[0009] To achieve the above objectives, this invention provides a method for assessing the post-disaster emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling, comprising the following steps: S1. Based on the DPSIR framework, the initial indicators for risk level assessment of landslide disaster emergency rescue hazard areas in mountainous railway tunnel sites are initially identified from five dimensions: driving force, pressure, state, impact, and response, forming an initial indicator set for risk assessment; S2. Calculate the coefficient of variation, independence value, and sensitivity value of the initial risk assessment indicator set obtained in S1. Based on the cumulative contribution rate threshold, complete two rounds of indicator screening, eliminate redundant indicators, and construct a standardized risk assessment indicator system. S3. Based on the standardized risk assessment indicator system constructed by S2, the correlation data matrix is ​​constructed by statistically analyzing the relationship between the indicators. After normalization, the fuzzy correlation matrix is ​​obtained. Then, through threshold transformation and reachability matrix calculation, a multi-level hierarchical causal structure of FISM is constructed. S4. Combining the topological characteristics of Bayesian networks, the FISM multi-level hierarchical causal structure constructed in S3 is transformed into a Bayesian network topology structure, clarifying the network nodes and the directed associations between nodes. S5. Based on the railway tunnel engineering risk management specifications, classify the risk status of all nodes in the Bayesian network topology obtained in S4, and determine a unified node status classification standard. S6. The EM algorithm is used to learn the parameters of the root node of the Bayesian network topology obtained in S4, and the prior probability table of the root node is calculated. At the same time, the triangular fuzzy number combined with the Noisy-Max model is used to calculate the conditional probability table of the Bayesian network child nodes. S7 integrates the Bayesian network topology of S4, the node state classification standard of S5, and the root node prior probability table and child node conditional probability table of S6 to construct a FISM-BN coupled landslide post-disaster emergency risk level assessment model. S8. Assign status values ​​to the measured data of the landslide hazard area in the target tunnel site according to the node status classification standard determined in S5. Input the assigned data into the landslide post-disaster emergency risk level assessment model constructed in S7. Calculate the risk level probability distribution through Bayesian network forward inference to determine the final risk level of the target area. S9. Based on the final risk level determined in S8, set the target risk level state in the landslide post-disaster emergency risk level assessment model constructed in S7, and calculate the probability change range of each indicator node through Bayesian network inverse reasoning to identify sensitive indicators affecting the risk level.

[0010] Therefore, the present invention employs the above-mentioned method for assessing the post-disaster emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling, which has the following beneficial effects: 1. Scientifically Adapted Indicator System: An initial indicator set was constructed based on the DPSIR framework. Redundant indicators were eliminated through two rounds of quantitative screening using the coefficient of variation and the cumulative contribution rate of the comprehensive value. This resulted in a standardized indicator system adapted to the emergency response scenario after landslides in mountainous railway tunnel areas. This system ensured the systematic nature and representativeness of the indicators and improved the pertinence and efficiency of subsequent assessments. It also overcame the shortcomings of traditional landslide risk assessment indicators being generalized and disconnected from emergency scenarios. 2. Rigorous model topology logic: The Fuzzy Interpretive Structure Model (FISM) is used to clarify the hierarchical causal relationship between risk indicators and directly transform them into a Bayesian network topology. This avoids the subjective defects of traditional Bayesian network modeling that relies on experience to set causal relationships, ensuring that the model structure is consistent with the risk transmission mechanism and improving the interpretability and scientificity of the model. 3. Accurate assignment of probability parameters: To address the challenge of assigning parameters to Bayesian networks, the EM algorithm is used to process the prior probability of the root node to solve the problem of missing samples. The triangular fuzzy number and the Noisy-Max model are combined to calculate the conditional probability of child nodes to solve the parameter explosion problem under the coupling of multiple parent nodes. This fully utilizes objective data and reasonably incorporates expert experience, greatly improving the objectivity and accuracy of probability parameters. 4. Comprehensive risk decision support: Through forward inference using Bayesian networks, the probability distribution of risk levels in hazard areas is quickly output, and reverse sensitivity analysis accurately identifies key influencing indicators, achieving the dual function of "quantitative judgment of risk levels + precise positioning of key control points". This provides direct and implementable decision-making basis for the allocation of emergency rescue resources and the formulation of disposal plans, overcoming the shortcomings of traditional assessment methods that can only perform qualitative classification and cannot support refined decision-making. 5. Highly adaptable to emergency scenarios: The indicator system, model parameters, and evaluation process of the solution are all adapted to the scenario characteristics of landslide emergency rescue and disaster relief areas in mountainous railway tunnel sites. Data collection, model input, and result output can be completed quickly after the disaster, meeting the timeliness requirements of emergency response and avoiding the problems of long cycle and poor operability of traditional evaluation methods, effectively ensuring the safety of rescue operations and railway operation. 6. Strong applicability and scalability in engineering: The solution has been verified through actual engineering cases, and the assessment results are highly consistent with the actual risk status of the project. It can provide a standardized technical process for emergency risk assessment after landslides in similar mountainous railway tunnel sites, and has good prospects for promotion and application.

[0011] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0012] Figure 1 This is a flowchart of a method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling, as described in this invention. Figure 2 This is a flowchart of the DPSIR classification index system for a method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling, as described in this invention. Figure 3 This is a binary adjacency matrix diagram of each risk indicator in a method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling, as described in this invention. Figure 4 This is a five-level hierarchical causal structure diagram of the risk indicators of the post-disaster emergency risk level assessment method for landslides in tunnel sites based on DPSIR-FISM-BN coupling described in this invention. Figure 5 The geological plan of the Daping landslide described in the engineering verification example; Figure 6 A risk level assessment result diagram of the Daping landslide hazard area described in the engineering verification example; Figure 7 A percentage diagram of risk levels at each stage in the Daping landslide hazard area, as described in an engineering verification example; Figure 8A comparison diagram of the probability distribution of the state of each parent node before and after reverse reasoning of node T, as described in the engineering verification example; Figure 9 This is a comparison diagram of the probability distribution of the state of each parent node before and after the reverse inference of node I4, as described in the engineering verification example. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the embodiments of the present invention and are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout.

[0014] It should be noted that the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, such as a process, method, system, product, or server that includes a series of steps or units, not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or device.

[0015] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0016] like Figures 1-4 As shown, a method for assessing the post-disaster emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling includes the following steps: S1. Based on the DPSIR framework, the initial indicators for risk level assessment of landslide disaster emergency rescue hazard areas in mountainous railway tunnel sites are initially identified from five dimensions: driving force, pressure, state, impact, and response, forming an initial indicator set for risk assessment.

[0017] S2. Calculate the coefficient of variation, independence value, and sensitivity value of the initial risk assessment indicator set obtained in S1. Based on the cumulative contribution rate threshold, complete two rounds of indicator screening, eliminate redundant indicators, and construct a standardized risk assessment indicator system.

[0018] S3. Based on the standardized risk assessment indicator system constructed by S2, the correlation data matrix is ​​constructed by statistically analyzing the relationship between the indicators. After normalization, a fuzzy correlation matrix is ​​obtained. Then, through threshold transformation and reachability matrix calculation, a multi-level hierarchical causal structure of FISM is constructed.

[0019] S4. Combining the topological characteristics of Bayesian networks, the FISM multi-level hierarchical causal structure constructed in S3 is transformed into a Bayesian network topology structure, clarifying the network nodes and the directed relationships between nodes.

[0020] S5. Based on the railway tunnel engineering risk management specifications, classify the risk status of all nodes in the Bayesian network topology obtained in S4, and determine a unified node status classification standard.

[0021] S6. The EM algorithm is used to learn the parameters of the root nodes of the Bayesian network topology obtained in S4, and the prior probability table of the root nodes is calculated. At the same time, the triangular fuzzy number combined with the Noisy-Max model is used to calculate the conditional probability table of the Bayesian network child nodes.

[0022] S7 integrates the Bayesian network topology of S4, the node state classification standard of S5, and the root node prior probability table and child node conditional probability table of S6 to construct a FISM-BN coupled landslide post-disaster emergency risk level assessment model.

[0023] S8. Assign status values ​​to the measured data of the landslide hazard area in the target tunnel site according to the node status classification standard determined in S5. Input the assigned data into the landslide post-disaster emergency risk level assessment model constructed in S7. Calculate the risk level probability distribution through forward inference of Bayesian network to determine the final risk level of the target area.

[0024] S9. Based on the final risk level determined in S8, set the target risk level state in the landslide post-disaster emergency risk level assessment model constructed in S7, and calculate the probability change range of each indicator node through Bayesian network inverse reasoning to identify sensitive indicators affecting the risk level.

[0025] In step S1, the initial assessment indicators for driving forces include rainfall data, rainfall duration data, rainfall intensity data, and river and ditch water level rise data; the initial assessment indicators for pressure include landslide scale data, fracture zone impact data, soil porosity data, soil shear strength data, mountain slope data, mountain slope height data, topographic humidity index data, slope toe scour data, emergency excavation disturbance intensity data, and slope rock and soil type data; the initial assessment indicators for state include cumulative slope displacement data, slope deformation rate data, surface crack development degree data, tunnel distance from the slip zone data, tunnel lining spalling state data, and number of tunnel lining cracks; the initial assessment indicators for impact include damage to the regional road network data, number of affected people data, vegetation damage degree data, traffic interruption duration data, and traffic capacity data; and the initial assessment indicators for response include emergency rescue access time data, response timeliness data, medical rescue support capacity data, flood control facility and equipment input data, and estimated repair cost data.

[0026] Among them, the topographic humidity index The calculation formula is as follows: ; In the formula, The uphill catchment area per unit contour line length; This indicates the terrain slope angle at the target grid point.

[0027] Step S2 specifically includes the following steps: S21. Obtain the importance scores of the initial risk assessment indicators through expert scoring, and then perform dimensionless processing on them: ; In the formula, Indicates the first Experts on the initial evaluation indicators The dimensionless importance score; Indicates the first Experts on the initial evaluation indicators Importance rating; This indicates the total number of experts who participated in the importance rating.

[0028] S22, Based on the results obtained in step S21 Calculate the cumulative contribution rate of the initial evaluation indicators: ; in, ; In the formula, Indicates the preceding The cumulative contribution rate of each initial evaluation indicator; and They represent the first The and the first The coefficient of variation of each initial evaluation indicator; This indicates the total number of indicators in the initial indicator set for risk assessment.

[0029] S23, Filter to meet the requirements The risk assessment indicators were used to obtain the indicator set after the first round of screening. This represents the cumulative contribution rate threshold for the first round of indicator selection; This indicates that after sorting by coefficient of variation in descending order, the first... The cumulative contribution rate of each initial evaluation indicator.

[0030] S24. Based on the set of indicators filtered in step S23, calculate the independence value: ; in, ; In the formula, This indicates the index set after the first round of screening. The independence value of each indicator; This indicates the index set after the first round of screening. The first indicator and the first The correlation coefficient of each indicator; This represents the total number of indicators in the indicator set after the first round of screening. Indicates the first The expert commented on the first The dimensionless importance score of each indicator; and They represent the first The and the first The average of the dimensionless importance scores for each indicator.

[0031] Simultaneously, calculate the sensitivity value: ; in, ; ; ; In the formula, Indicates the first The experts reviewed the first round of screening of the selected indicators. The characteristic weight of each indicator's importance score; Indicates the first The entropy value of each indicator; Indicates the first The entropy weights of each indicator; Indicates the first Sensitivity values ​​of each indicator.

[0032] S25. Calculate the comprehensive value of the index based on the independence value and the sensitivity value: ; In the formula, Indicates the first The composite value of the indicators; and All represent linear coefficients; Indicates the first The independence value of each indicator.

[0033] S26. The secondary cumulative contribution rate of each indicator can be calculated based on the comprehensive value of the evaluation indicators: ; In the formula, Indicates the preceding The cumulative contribution rate of each indicator after the first round of screening.

[0034] S27. Descend the cumulative contribution rate of the second phase. After sorting, the top 95% of the indicators are selected as effective evaluation indicators.

[0035] In step S2, after two rounds of index screening, the effective evaluation indicators for driving forces include rainfall data, rainfall duration data, and river and ditch water level rise data. Effective evaluation indicators for pressure include landslide scale data in the hazard area, impact data of fracture zones, slope data, disturbance intensity data from emergency excavation, and slope soil and rock type data; effective evaluation indicators for state include cumulative slope displacement data, slope deformation rate data, surface crack development data, distance between the tunnel and the slip zone data, and tunnel lining spalling status data; effective evaluation indicators for impact include damage data to the regional road network, number of affected people, duration of traffic interruption data, and traffic capacity data; effective evaluation indicators for response include response timeliness data, medical rescue support capacity data, emergency rescue access time data, and estimated repair cost data.

[0036] Step S3 specifically includes the following steps: S31. Based on the standardized risk assessment indicator system constructed in step S2, determine the indicator set as follows: ,in, This indicates the total number of valid evaluation indicators included in the standardized indicator system. This indicates the first in the standardized risk assessment indicator system. 1. Effective evaluation indicators; 2. Simultaneously, the frequency of correlations between these indicators in landslide cases at mountain railway tunnel sites was statistically analyzed to construct an original correlation data matrix. .

[0037] S32. The original association data matrix constructed in step S31 After performing maximum value normalization, the fuzzy correlation matrix is ​​obtained. .

[0038] S33, Set the correlation strength threshold The fuzzy correlation matrix obtained in step S32 Perform threshold transformation to obtain the adjacency matrix. .

[0039] S34. Adjacency matrix obtained in step S33 The reachability matrix is ​​obtained by computation through transitive closure. Each element in this matrix represents whether a factor can influence another factor through a direct or indirect path.

[0040] S35. The reachability matrix obtained in step S34 Calculate each indicator reachable set Precedent Collection and intersection The indicator hierarchy was then divided according to the following hierarchical rules: like Then the indicator Assign indicators to the current level to obtain the current level set, then remove the indicators of the assigned levels and their corresponding values ​​in the reachability matrix. The corresponding rows and columns are repeated. The discrimination is carried out until all indicators are hierarchically divided; based on the hierarchical division results, the FISM multi-level hierarchical causal structure is obtained.

[0041] Step S4 specifically includes the following steps: S41. Extract all hierarchical elements, inter-hierarchical transmission relationships, and directed causal orientations of nodes in the FISM multi-level hierarchical causal structure constructed in step S3, and clarify the hierarchical affiliation of each indicator node and the correspondence between parent and child nodes.

[0042] S42. Map each effective assessment indicator in the standardized risk assessment indicator system constructed in step S2 to a basic node of the Bayesian network; and set the post-landslide emergency risk level as the target total node of the Bayesian network, thus completing the definition of all nodes of the Bayesian network.

[0043] S43. Based on the directed causal propagation rules of the FISM multilevel hierarchical causal structure, configure directed connection edges for the Bayesian network nodes: from the upper-level parent node to the lower-level child node, following the causal propagation direction in the FISM multilevel hierarchical causal structure, and construct the initial directed acyclic topology of the Bayesian network.

[0044] S44. Perform compliance verification on the initial topology of the Bayesian network: delete invalid and redundant associated edges, correct logically erroneous associations, and retain valid associations consistent with the risk transmission mechanism after landslide disasters.

[0045] S45. Based on the optimized topology, identify the types of Bayesian network nodes: classify basic nodes without parent nodes as root nodes, basic nodes with parent nodes and child nodes as intermediate nodes, and target nodes that are only risk level child nodes as leaf nodes. Clarify the causal attributes of the directed associations between all nodes, and finally form a complete Bayesian network topology.

[0046] In step S5, the risk status classification expression is as follows: ; In the formula, This indicates the Bayesian network's first... The risk status of each node; This indicates the Bayesian network's first... The measured index values ​​of each node; , , and These represent low-risk, moderate-risk, relatively high-risk, and major-risk states, respectively.

[0047] In step S6, the specific steps for calculating the conditional probability table of the Bayesian network child nodes using triangular fuzzy numbers combined with the Noisy-Max model are as follows: The first step is to construct a probabilistic language scoring system for child nodes under the influence of a single parent node, divide the expert evaluation language into 7 levels, match the corresponding triangular fuzzy numbers, and complete the standard definition of fuzzy numbers; the rules for triangular fuzzy numbers are as follows: Very low: (0,0,0.1); Low: (0, 0.1, 0.3); Low: (0.1, 0.3, 0.5); Medium: (0.3, 0.5, 0.7); Slightly high: (0.5, 0.7, 0.9); High: (0.7, 0.9, 1.0); Very high: (0.9, 1.0, 1.0).

[0048] The second step involves inviting experts to conduct linguistic scoring on the probability of each state of the child node under the influence of a single parent node. The linguistic variables are then converted into triangular fuzzy numbers, and a weighted fusion is performed to obtain a comprehensive triangular fuzzy number. The formula for calculating the comprehensive triangular fuzzy number is as follows: ; and ; In the formula, Represents a single parent node Under the influence of this action, the child node's... A comprehensive triangular fuzzy number for each risk state. (These correspond to four different risk levels). Indicates the first The weight of each expert; This indicates the total number of experts involved in the language assessment; Indicates the first An expert on single parent node Under the influence of this action, the child node's... The triangular fuzzy number corresponding to the language score given for each risk status; Indicates the first The triangular fuzzy number given by the experts.

[0049] The third step is to defuzzify and normalize the comprehensive triangular fuzzy number to obtain the conditional probability of the child node under the influence of a single parent node: ; ; In the formula, This represents the probability value after defuzzification; This represents the normalized conditional probability of a single parent node. , , Both represent the weighting coefficients for defuzzification using triangular fuzzy numbers.

[0050] Step 4: Based on the Noisy-Max model, assuming that all parent nodes have an enhancing effect on child nodes, the max operator is used to synthesize the conditional probabilities of child nodes under the combined effect of multiple parent nodes; the formula for synthesizing multiple parent nodes is as follows: ; In the formula, Indicates a child node in a Bayesian network Two parent nodes In state Parent node In state At that time, child nodes It is in a state of risk. The conditional probability, Represents a child node in a Bayesian network. and These are any two parent nodes of the child node. and The risk status corresponding to the parent node; This represents the total number of risk status levels of the nodes. ; , This represents the intermediate state level variable of a child node under the influence of a single parent node. Indicates at the parent node In state At that time, child nodes In a state of intermediate risk The conditional probability; Indicates at the parent node In state At that time, child nodes In a state of intermediate risk The conditional probability; This indicates that the final state of a child node is the higher of the two intermediate states.

[0051] Step 5: Traverse all child nodes and their parent node combinations, complete all conditional probability calculations, and summarize to generate a Bayesian network child node conditional probability table.

[0052] In this embodiment, the final risk level determined by S9 and the sensitivity indicators identified by S10 can be combined to generate a risk prevention and emergency response plan for post-landslide emergency rescue based on risk acceptance criteria.

[0053] Engineering Verification Examples This embodiment selects the Daping landslide emergency rescue hazard area at the exit section of the Sunjiaya Tunnel in a mountainous area of ​​Southwest China as the verification object. The Sunjiaya Tunnel is located in Fengjie County, near the left bank of the Meixi River, about 2 km from the confluence of the Yangtze River and its slope, which is part of the Daping landslide hazard area within the study area. The Daping landslide is a complex old landslide, composed of a preceding shallow deposit landslide and a subsequent fractured rock and soil landslide. Under the combined effects of regional geological structure, heavy rainfall, river erosion, and engineering disturbances, the slope stability is poor, and it is still in a state of continuous deformation, exhibiting obvious reactivation characteristics. Because this landslide has undergone sliding deformation and continues to be active under rainfall and external disturbances, it is not a typical static hazard point, but an active hazard area with long-term creep and the risk of re-instability. The continuous activity of the landslide has resulted in significant deformation responses on the surface and in buildings, with cracks in the provincial highway above the slope and cracks in the walls of residential buildings being particularly typical examples.

[0054] From a topographical perspective, the landslide area is characterized by a river valley slope, exhibiting an alternating pattern of steep-gentle-steep slopes, with elevations ranging from approximately 164.0 to 425.5 meters and a relative elevation difference of about 261.5 meters. Locally, the bedrock scarp at the rear edge can reach approximately 55°, resulting in significant slope undulation and complex topographic conditions. The landslide is controlled by natural gullies on both sides, and the bedrock scarp at the rear edge and the riverbank slope at the front edge jointly define the landslide boundary, indicating that the topography within the area has a strong controlling effect on landslide development and deformation. According to the nodal grading standard, the slope conditions in the study area are generally at a relatively unfavorable level, with a significant development of the rear scarp and exposed surface, which is detrimental to the long-term stability of the slope.

[0055] Based on stratigraphic lithology and slope structure characteristics, the landslide area mainly consists of Quaternary landslide deposits, colluvial gravelly soil, and Triassic Badong Formation marl interbedded with mudstone. The slope's rock and soil types are predominantly soil-rock mixtures and fractured rock masses, with intense local weathering, poor rock mass integrity, well-developed slip zones, and high water content, possessing strong softening and deformation conditions. The landslide is large in scale, with a volume far exceeding 500,000 m³. 3 This landslide is classified as a large-scale potential landslide. Based on the on-site investigation and existing survey results, it is known that the landslide exhibits significant differentiation between its preceding and subsequent stages, has a relatively complex structure, poor internal deformation coordination, and is prone to localized accelerated deformation under the influence of external triggering factors.

[0056] From a meteorological and hydrological perspective, the landslide area belongs to the mid-latitude subtropical warm and humid monsoon climate zone, characterized by a mild and humid climate with abundant rainfall. The average annual precipitation is approximately 1179 mm, with over 68% of the annual rainfall occurring from March to August. The maximum 3-day rainfall can reach 200-350 mm, and the maximum 24-hour rainfall can reach 80-120 mm, indicating that both short-duration heavy rainfall and continuous rainfall events occur in the area. The Meixi River, a first-level tributary of the Yangtze River, flows into the area in front of the landslide. The river's flow is mainly controlled by rainfall, with the high-water season concentrated from May to September, and the average annual flow rate is approximately 45.9 m³. 3 / s.

[0057] In addition, two gullies, roughly aligned with the main sliding direction, have developed within the landslide area. During the rainy season, runoff can migrate along the interior and surface of the landslide body. The upper and middle parts of the landslide are mainly dry land with a few ponds, where surface runoff and shallow seepage are relatively well-developed. Groundwater in the area mainly consists of pore water from Quaternary loose rocks and fissure water from bedrock, with karst fissure water being the predominant type. Well-developed bedrock structural fissures and karst fissures provide favorable conditions for the formation, accumulation, and migration of groundwater. Rainfall infiltration, groundwater uplift, and river erosion collectively weaken the anti-sliding capacity of the slope front, creating unfavorable hydrogeological conditions for the continued landslide activity.

[0058] It should be noted that the Daping landslide is closely related to the Sunjiaya Tunnel project, and is a typical landslide hazard in the tunnel site area. The Sunjiaya Tunnel and related engineering works pass through the front and middle parts of the landslide body, and there is a clear spatial coupling relationship between the tunnel's location and the landslide deformation zone. Affected by the long-term creep deformation of the landslide and the disturbance of construction excavation, the stress conditions and structural stability of the surrounding rock in the tunnel site area have changed. During tunnel construction, phenomena such as initial support cracking, steel arch deformation, increased surrounding rock pressure, and abnormal stress on the lining were observed, indicating that the landslide activity has directly affected the tunnel's structural safety and construction and operation conditions.

[0059] Specifically, the landslide activity has already caused some damage to road traffic in the area, with deformation and traffic restrictions occurring in some sections, reflecting its actual impact on the regional road network operation. In summary, the Daping landslide hazard area is characterized by its large scale, steep slope, fractured soil and rock, significant hydrological disturbance, obvious engineering disturbance, structural impact on the tunnel site, and high risk of road network damage. Furthermore, the landslide has already occurred and is still undergoing deformation activity, meeting the basic characteristics of a landslide disaster emergency rescue hazard area in a mountainous railway tunnel site. Therefore, this landslide hazard area was selected as an engineering case study to conduct a risk level assessment of the hazard area, in order to verify the applicability of the model.

[0060] Evaluation of the implementation process 1. Basic data acquisition and processing; like Figure 5As shown, to assess the risk level of the Daping landslide hazard area at the Sunjiaya Tunnel site and provide a basis for subsequent Bayesian network parent node state classification, this embodiment collects and organizes basic information corresponding to parent nodes D2, D3, P1, P2, P3, P4, P5, I1, and R1. Data sources mainly include engineering geological survey reports, construction and monitoring data, meteorological and hydrological data, field investigation records, and relevant engineering design documents. During data processing, quantitative indicators preferentially use measured values, statistical values, or survey calculation results as basic data, and are classified according to node state classification standards; qualitative indicators are classified based on field investigation phenomena, engineering records, and standard criteria. After obtaining the basic data of the parent nodes, the node state classification is performed according to Table 1, and the results are shown in Table 2.

[0061] Table 1. Bayesian Network Node State Classification Table

[0062] Table 2 State classification results

[0063] 2. Determine the probability of BN nodes in the risk level assessment of the hazard area. In the established Bayesian network for risk level assessment of landslide disaster emergency rescue hazard areas in mountainous railway tunnel sites, there are two different types of nodes. The first type of node is the root node, such as rainfall duration, mountain slope, and landslide size in the hazard area; the second type is the child nodes, such as traffic capacity, deformation rate, and medical rescue support capacity. To analyze and assess the risk level of this hazard area, it is necessary to calculate the prior probability table and conditional probability table of the nodes.

[0064] This embodiment employs the Bayesian network modeling tool Netica to achieve efficient model construction and simulation. The study collected data from 75 tunnel engineering cases and used the Expectation-Maximization (EM) algorithm to autonomously learn the node parameters. The probability distribution of each node was obtained through iterative software calculation. The results are shown in Table 3.

[0065] Table 3. Prior probability table for parent nodes calculated using the EM algorithm.

[0066] After calculating the prior probabilities of the nine parent nodes using the EM algorithm, it is necessary to further determine the conditional probability tables of the child nodes to complete the Bayesian network model and ultimately obtain the system's probability distribution. For the Bayesian network used in this embodiment for risk assessment of landslide hazard areas, the risk level node T is not determined by a single factor but is influenced by the combined effects of multiple consequence-type nodes. Directly assigning values ​​to each combination of parent node states would not only require a large number of parameters but also make it difficult to guarantee the consistency of the conditional probability tables. Therefore, this embodiment uses a combination of triangular fuzzy numbers and the Noisy-Max model to calculate the conditional probability tables of the child nodes. Since there are many child nodes and their combinations, this embodiment takes the risk level node T as an example to solve its conditional probability table, obtaining the corresponding conditional probability table. The other child nodes are not discussed further here.

[0067] (1) Calculation of the conditional probability of node risk level T under the influence of a single factor event. In the process of solving the conditional probability table of multiple parent nodes, it is first necessary to determine the probability of the influence of each parent node on the child node under its individual effect. This embodiment introduces the expert linguistic variable scoring method and uses triangular fuzzy numbers to describe the expert judgment results, thereby obtaining the conditional probability distribution of node T under the effect of a single factor. After multiple rounds of expert scoring and calculation, this embodiment obtains the conditional probability distribution of node T under the individual effects of parent nodes I2 and I3 as shown in Table 4.

[0068] Table 4. Conditional Probability of Node Risk Level T under the Influence of a Single Factor Event

[0069] As shown in Table 4, node T is in a high-risk state when the state level of I2 or I3 gradually increases. or The probability of [something] gradually increases, while [something] is in a low-risk state. or The probability gradually decreases. This is consistent with the risk evolution pattern of landslide hazard zones, indicating that the determination result of single-factor conditional probability has good rationality.

[0070] (2) Calculation of the conditional probability of node risk level T under the combined condition of parent nodes. After obtaining the conditional probability of a single factor, the conditional probability of node T under the combined effect of I2 and I3 is further calculated. Since the influence of the two parent nodes on the node is a reinforcing relationship, that is, when the state values ​​of both increase, they will both cause the final risk level to evolve to a higher state, the Noisy-Max model is used for combined calculation. The conditional probability of node T calculated using this invention is shown in Table 5.

[0071] Table 5 Conditional Probabilities of Node Risk Level T

[0072] Finally, the prior probability table of the parent node and the conditional probability table of the child node are imported into Netica software to complete the risk level assessment and index sensitivity analysis of the Daping landslide hazard area.

[0073] 3. Risk level assessment and analysis of the landslide hazard area in the tunnel site area; After calculating the prior probability of the root node and the conditional probability of the child nodes, the risk level of the Daping landslide hazard area is evaluated and analyzed in Netica software using a Bayesian network model. Then, the sensitivity of the risk level indicators is analyzed by setting the target values ​​of the child nodes.

[0074] (1) Risk level assessment results; In the Bayesian network structure for risk level assessment of the landslide hazard area in the mountain railway tunnel site, after importing the prior probability of the parent node and the conditional probability of the child node, the risk level of the Daping landslide hazard area is finally obtained, such as... Figure 6 As shown in the figure, node T indicates that the risk level of the Daping landslide hazard area is Level III, with a probability of 34.9%, which is a relatively high risk. The acceptance principle is not to expect it, and the risk control principle is to pay attention to it and take effective measures to deal with it, and strengthen risk monitoring.

[0075] (2) Sensitivity analysis of indicators; When conducting sensitivity analysis on various indicators affecting the risk level of landslide disaster emergency rescue hazard areas in mountainous railway tunnel sites, the state probability values ​​of target nodes in the Bayesian network model can be used for reverse reasoning to calculate the state probability values ​​of all parent nodes affecting that child node. By comparing the changes in the state probability values ​​of the parent nodes before and after the landslide, the sensitivity of the indicators affecting the child node can be determined. For example, the risk level assessment results of the Daping landslide hazard area show that, after analysis and calculation using the Bayesian network model, the state probability values ​​of node risk level T for levels I, II, III, and IV are 11.3%, 27.8%, 34.9%, and 26.0%, respectively, with each level's state probability value accounting for a significant percentage. Figure 7 As shown, the probability values ​​of the number of affected people (I2) of its parent node are 38.4%, 25.4%, 21.4%, and 14.8%, respectively; the probability values ​​of the interruption duration (I3) are 10.9%, 14.3%, 21.5%, and 53.2%, respectively.

[0076] The risk levels I, II, III, and IV of the Daping landslide hazard area are set to 80%, 20%, 0%, and 0% respectively for reverse analysis. The comparison diagram of the probability distribution of each parent node is shown below. Figure 8 As shown, we can see I3 ( The percentage needs to be reduced from 53.2% to 38.2%, and I2 ( It needs to be increased from 38.4% to 55.4%.

[0077] The inverse reasoning results of the sensitivity analysis show that, while keeping the root node state and its prior distribution unchanged, the key influencing factors on the risk level T of the target node exhibit significant differences. When the risk level of the Daping landslide hazard area is gradually reduced from Level III to Level I, the positive improvement of the traffic interruption duration I3 in the parent node contributes significantly more to T than the number of affected people I2. This indicates that, under the current network structure and parameter system, the marginal impact of the traffic interruption duration I3 on ​​the risk level node is more prominent, making it a major sensitive factor for changes in risk level. Furthermore, considering the initial state distribution of variables and the engineering background, the number of affected people I2 is already at a relatively low level under this engineering background. Therefore, even with positive regulation of I2, its potential for improvement is relatively limited, resulting in a less significant reduction effect on the risk level T. In contrast, the state change of the traffic interruption duration I3 not only directly reflects the intensity of the disturbance to the transportation system caused by the landslide disaster but also more sensitively characterizes the degree of limitation on the operational function of the line and the difference in recovery capacity. Especially in mountainous tunnel areas where railways serve as vital regional transportation arteries, a landslide causing traffic disruption can trigger a chain reaction, even within a short period, leading to decreased traffic efficiency, limited emergency resource allocation, and disruptions to regional logistics and personnel movement. This can result in significant economic losses and social impacts. Therefore, the dominant role of I3 in the risk level analysis is consistent with actual engineering practices.

[0078] In summary, based on the Bayesian network sensitivity analysis, the duration of traffic interruption (I3) is a key sensitive factor affecting the risk level changes in the Daping landslide hazard area. In formulating risk level control and disaster reduction strategies, priority should be given to management and engineering measures related to I3, such as improving emergency road clearing efficiency, optimizing emergency traffic organization, improving detour and control plans, and strengthening post-disaster recovery resource allocation, thereby achieving more significant results in reducing the risk level.

[0079] The duration of traffic interruption I3 is mainly affected by the traffic capacity I4 and the estimated repair cost R4. The traffic capacity I4 has many parent nodes. Taking the traffic capacity I4 as an example, the traffic capacity I4 ( ) set to 75%, I4 ( The spalling rate of the tunnel lining is set at 25%, S1 ( The percentage needs to be increased from 16.6% to 34.5%; the cumulative slope displacement value S4 ( The deformation rate (S5) needs to be increased from 18.0% to 27.9%. The probability needs to be increased from 5.9% to 15.3%. A comparison of the probability distribution of each parent node's state is shown in the diagram below. Figure 9As shown, the spalling index of tunnel lining is more sensitive. This is mainly because lining spalling is a structural indicator that directly constrains traffic safety. Its deterioration can lead to management decisions at the operational level, ranging from speed limits and traffic restrictions to closures. Once there is a risk of continuous rockfall or exposed rebar, even if the slope deformation is not large, traffic capacity may still be significantly reduced due to safety risks. Conversely, when the lining condition improves to the point of having no impact ( ) or only localized chipping ( When [the traffic flow is restored], the effect of restoring traffic capacity is more obvious.

[0080] Therefore, in the causal chain of the Bayesian network, the influence path of S1 on I4 is more direct, resulting in a higher impact in the sensitivity analysis. Thus, in practical risk management, improving structural safety should be prioritized, while simultaneously implementing deformation control and monitoring / early warning systems to effectively enhance traffic capacity and reduce secondary operational risks, validating the model's accuracy and practicality.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for evaluating the post-disaster emergency risk level of a landslide in a tunnel address area based on a DPSIR-FISM-BN coupling, characterized in that: Includes the following steps: S1. Based on the DPSIR framework, the initial indicators for risk level assessment of landslide disaster emergency rescue hazard areas in mountainous railway tunnel sites are initially identified from five dimensions: driving force, pressure, state, impact, and response, forming an initial indicator set for risk assessment; S2. Calculate the coefficient of variation, independence value, and sensitivity value of the initial risk assessment indicator set obtained in S1. Based on the cumulative contribution rate threshold, complete two rounds of indicator screening, eliminate redundant indicators, and construct a standardized risk assessment indicator system. S3. Based on the standardized risk assessment indicator system constructed by S2, the correlation data matrix is ​​constructed by statistically analyzing the relationship between the indicators. After normalization, the fuzzy correlation matrix is ​​obtained. Then, through threshold transformation and reachability matrix calculation, a multi-level hierarchical causal structure of FISM is constructed. S4. Combining the topological characteristics of Bayesian networks, the FISM multi-level hierarchical causal structure constructed in S3 is transformed into a Bayesian network topology structure, clarifying the network nodes and the directed associations between nodes. S5. Based on the railway tunnel engineering risk management specifications, classify the risk status of all nodes in the Bayesian network topology obtained in S4, and determine a unified node status classification standard. S6. The EM algorithm is used to learn the parameters of the root node of the Bayesian network topology obtained in S4, and the prior probability table of the root node is calculated. At the same time, the triangular fuzzy number combined with the Noisy-Max model is used to calculate the conditional probability table of the Bayesian network child nodes. S7 integrates the Bayesian network topology of S4, the node state classification standard of S5, and the root node prior probability table and child node conditional probability table of S6 to construct a FISM-BN coupled landslide post-disaster emergency risk level assessment model. S8. Assign status values ​​to the measured data of the landslide hazard area in the target tunnel site according to the node status classification standard determined in S5. Input the assigned data into the landslide post-disaster emergency risk level assessment model constructed in S7. Calculate the risk level probability distribution through Bayesian network forward inference to determine the final risk level of the target area. S9. Based on the final risk level determined in S8, set the target risk level state in the landslide post-disaster emergency risk level assessment model constructed in S7, and calculate the probability change range of each indicator node through Bayesian network inverse reasoning to identify sensitive indicators affecting the risk level.

2. The method according to claim 1, wherein the method is characterized in that: In step S1, the initial evaluation indicators for driving forces include rainfall data, rainfall duration data, rainfall intensity data, and river and ditch water level rise data; The initial assessment indicators for stress include landslide size data in the hazard area, impact data of fracture zones, soil porosity data, soil shear strength data, slope data, slope height data, topographic humidity index data, slope toe scour data, emergency excavation disturbance intensity data, and slope rock and soil type data. The initial assessment indicators for the condition category include cumulative slope displacement data, slope deformation rate data, surface crack development degree data, distance between the tunnel and the slip zone data, tunnel lining spalling status data, and number of tunnel lining cracks; The initial assessment indicators for impact include data on road network damage within the region, number of people affected, degree of vegetation damage, duration of traffic interruption, and traffic capacity. Initial assessment indicators for response include data on emergency rescue availability, response timeliness, medical rescue support capabilities, investment in flood control facilities and equipment, and estimated repair costs.

3. The method according to claim 2, wherein the method is characterized in that: Step S2 specifically includes the following steps: S21. Obtain the importance scores of the initial risk assessment indicators through expert scoring, and then perform dimensionless processing on them: ； In the formula, Indicates the first Experts on the initial evaluation indicators The dimensionless importance score; Indicates the first Experts on the initial evaluation indicators Importance rating; This indicates the total number of experts who participated in the importance rating; S22, Based on the results obtained in step S21 Calculate the cumulative contribution rate of the initial evaluation indicators: ； in, ； In the formula, Indicates the preceding The cumulative contribution rate of each initial evaluation indicator; and They represent the first The and the first The coefficient of variation of each initial evaluation indicator; This indicates the total number of indicators in the initial set of risk assessment indicators; S23, Filter to meet the requirements The risk assessment indicators were used to obtain the indicator set after the first round of screening. This represents the cumulative contribution rate threshold for the first round of indicator selection; This indicates that after sorting by coefficient of variation in descending order, the first... The cumulative contribution rate of each initial evaluation indicator; S24. Based on the set of indicators filtered in step S23, calculate the independence value: ； in, ； In the formula, This indicates the index set after the first round of screening. The independence value of each indicator; This indicates the index set after the first round of screening. The first indicator and the first The correlation coefficient of each indicator; This represents the total number of indicators in the indicator set after the first round of screening. Indicates the first The expert commented on the first The dimensionless importance score of each indicator; and They represent the first The and the first The mean of the dimensionless importance scores of each indicator; Simultaneously, calculate the sensitivity value: ； in, ； ； ； In the formula, Indicates the first The experts reviewed the first round of screening of the selected indicators. The characteristic weight of each indicator's importance score; Indicates the first The entropy value of each indicator; Indicates the first The entropy weights of each indicator; Indicates the first Sensitivity values ​​of each indicator; S25. Calculate the comprehensive value of the index based on the independence value and the sensitivity value: ； In the formula, Indicates the first The composite value of the indicators; and All represent linear coefficients; Indicates the first The independence value of each indicator; S26. The secondary cumulative contribution rate of each indicator can be calculated based on the comprehensive value of the evaluation indicators: ； In the formula, Indicates the preceding The cumulative contribution rate of each indicator after the first round of screening; S27. Descend the cumulative contribution rate of the second phase. After sorting, the top 95% of the indicators are selected as effective evaluation indicators.

4. The method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling as described in claim 3, characterized in that: In step S2, after two rounds of index screening, the effective evaluation indicators for driving forces include rainfall data, rainfall duration data, and river and ditch water level rise data; Effective stress assessment indicators include landslide size data in the hazard area, impact data of fracture zones, slope data, disturbance intensity data of emergency excavation, and soil and rock type data of the slope. Effective assessment indicators for condition-related data include cumulative slope displacement data, slope deformation rate data, surface crack development data, tunnel distance from slip zone data, and tunnel lining spalling status data. Effective impact assessment indicators include data on road network damage within the region, number of affected people, duration of traffic interruption, and traffic capacity. Effective evaluation indicators for response include data on response timeliness, medical rescue and support capabilities, emergency rescue availability time, and estimated repair costs.

5. The method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling as described in claim 4, characterized in that: Step S3 specifically includes the following steps: S31. Based on the standardized risk assessment indicator system constructed in step S2, determine the indicator set as follows: ,in, This indicates the total number of valid evaluation indicators included in the standardized indicator system. This indicates the first in the standardized risk assessment indicator system.

1. Effective evaluation indicators; 2. Simultaneously, the frequency of correlations between these indicators in landslide cases at mountain railway tunnel sites was statistically analyzed to construct an original correlation data matrix. ; S32. The original association data matrix constructed in step S31 After performing maximum value normalization, the fuzzy correlation matrix is ​​obtained. ; S33, Set the correlation strength threshold The fuzzy correlation matrix obtained in step S32 Perform threshold transformation to obtain the adjacency matrix. ; S34. Adjacency matrix obtained in step S33 The reachability matrix is ​​obtained by computation through transitive closure. ; S35. The reachability matrix obtained in step S34 Calculate each indicator reachable set Precedent Collection and intersection The indicator hierarchy was then divided according to the following hierarchical rules: like Then the indicator Assign indicators to the current level to obtain the current level set, then remove the indicators of the assigned levels and their corresponding values ​​in the reachability matrix. The corresponding rows and columns are repeated. The discrimination is carried out until all indicators are hierarchically divided; based on the hierarchical division results, the FISM multi-level hierarchical causal structure is obtained.

6. The method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling as described in claim 5, characterized in that: Step S4 Specifically, the following steps are included: S41. Extract all hierarchical elements, inter-hierarchical transmission relationships, and directed causal orientations of nodes in the FISM multi-level hierarchical causal structure constructed in step S3, and clarify the hierarchical affiliation of each indicator node and the correspondence between parent and child nodes. S42. Map each effective assessment indicator in the standardized risk assessment indicator system constructed in step S2 to a basic node of the Bayesian network; and set the post-landslide emergency risk level as the target total node of the Bayesian network, thus completing the definition of all nodes of the Bayesian network. S43. Based on the directed causal propagation rules of the FISM multilevel hierarchical causal structure, configure directed connection edges for the Bayesian network nodes: from the upper-level parent node to the lower-level child node, following the causal propagation direction in the FISM multilevel hierarchical causal structure, and construct the initial directed acyclic topology of the Bayesian network. S44. Perform compliance verification on the initial topology of the Bayesian network: delete invalid and redundant associated edges, correct logically erroneous associations, and retain valid associations that are consistent with the risk transmission mechanism after landslide disasters; S45. Based on the optimized topology, identify the types of Bayesian network nodes: classify basic nodes without parent nodes as root nodes, basic nodes with parent nodes and child nodes as intermediate nodes, and target nodes that are only risk level child nodes as leaf nodes. Clarify the causal attributes of the directed associations between all nodes, and finally form a complete Bayesian network topology.

7. The method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling as described in claim 6, characterized in that: In step S5, the risk status classification expression is as follows: ； In the formula, This indicates the Bayesian network's first... The risk status of each node; This indicates the Bayesian network's first... The measured index values ​​of each node; , , and These represent low-risk, moderate-risk, relatively high-risk, and major-risk states, respectively.

8. The method for assessing the emergency risk level of landslides in tunnel sites based on DPSIR-FISM-BN coupling as described in claim 7, characterized in that: In step S6, the specific steps for calculating the conditional probability table of the Bayesian network child nodes using triangular fuzzy numbers combined with the Noisy-Max model are as follows: The first step is to construct a probabilistic language scoring system for child nodes under the influence of a single parent node, divide the expert evaluation language into 7 levels, match the corresponding triangular fuzzy numbers, and complete the standard definition of fuzzy numbers; the rules for triangular fuzzy numbers are as follows: Very low: (0,0,0.1); Low: (0, 0.1, 0.3); Low: (0.1, 0.3, 0.5); Medium: (0.3, 0.5, 0.7); Slightly high: (0.5, 0.7, 0.9); High: (0.7, 0.9, 1.0); Very high: (0.9, 1.0, 1.0); The second step involves inviting experts to conduct linguistic scoring on the probability of each state of the child node under the influence of a single parent node. The linguistic variables are then converted into triangular fuzzy numbers, and a weighted fusion is performed to obtain a comprehensive triangular fuzzy number. The formula for calculating the comprehensive triangular fuzzy number is as follows: ； and ; In the formula, Represents a single parent node Under the influence of this action, the child node's... A comprehensive triangular fuzzy number for each risk state. ; Indicates the first The weight of each expert; This indicates the total number of experts involved in the language assessment. Indicates the first An expert on single parent node Under the influence of this action, the child node's... The triangular fuzzy number corresponding to the language score given for each risk status; Indicates the first The triangular fuzzy number given by an expert; The third step is to defuzzify and normalize the comprehensive triangular fuzzy number to obtain the conditional probability of the child node under the influence of a single parent node: ； ； In the formula, This represents the probability value after defuzzification; This represents the normalized conditional probability of a single parent node. , , All represent the weighting coefficients for defuzzification using triangular fuzzy numbers; Step 4: Based on the Noisy-Max model, assuming that all parent nodes have an enhancing effect on child nodes, the max operator is used to synthesize the conditional probabilities of child nodes under the combined effect of multiple parent nodes; the formula for synthesizing multiple parent nodes is as follows: ； In the formula, Indicates a child node in a Bayesian network Two parent nodes In state Parent node In state At that time, child nodes It is in a state of risk. The conditional probability, Represents a child node in a Bayesian network. and These are any two parent nodes of the child node. and The risk status corresponding to the parent node; This represents the total number of risk status levels of the nodes. ; , This represents the intermediate state level variable of a child node under the influence of a single parent node. Indicates at the parent node In state At that time, child nodes In a state of intermediate risk The conditional probability; Indicates at the parent node In state At that time, child nodes In a state of intermediate risk The conditional probability; This indicates that the final state of a child node is the higher of the two intermediate states. Step 5: Traverse all child nodes and their parent node combinations, complete all conditional probability calculations, and summarize to generate a Bayesian network child node conditional probability table.

Images