A long-term safety evaluation method and system for a complex and dangerous mountainous tunnel group structure

Abstract

This invention provides a method and system for long-term safety evaluation of tunnel groups in complex and dangerous mountainous areas, relating to the field of tunnel safety evaluation technology. The method includes: obtaining risk factors and risk weights for tunnel groups in complex and dangerous mountainous areas; dividing the tunnel groups into risk zones based on the risk factors and risk weights to obtain a tunnel group risk zoning map, which includes multiple risk zoning levels for different sections; based on the tunnel group risk zoning map, and combining structural stability degradation and non-structural risk evolution modeling under risk zoning constraints, obtaining evaluation vectors and safety evaluation level labels for multiple typical sections; and based on the evaluation vectors and safety evaluation level labels of the typical sections, combined with the risk zoning levels of the sections to be evaluated, performing comparative learning under risk zoning constraints to obtain the safety evaluation results for the sections to be evaluated. This invention solves the problem of not considering the differences in risk between sections when conducting safety evaluations of tunnel groups in complex and dangerous mountainous areas.

Description

Technical Field

[0001] This invention relates to the field of tunnel safety evaluation technology, and more specifically, to a method and system for long-term safety evaluation of tunnel groups in complex and dangerous mountainous areas. Background Technology

[0002] In the field of long-term structural safety assessment of tunnel groups in complex and challenging mountainous areas, traditional assessment methods generally rely on field monitoring data or single structural indicators such as lining deformation, crack width, and geological survey parameters, and conduct risk assessments through empirical formulas or simple weighted models. These methods have significant limitations: First, they treat the tunnel group as a whole for unified evaluation, ignoring the heterogeneity of risks in different sections due to geological conditions, environmental effects, and operational loads, and failing to fully consider the interrelationships such as risk diffusion between sections, making it difficult to accurately reflect the true safety status of the tunnel group. Second, although some multi-indicator comprehensive evaluation, fuzzy hierarchical analysis, and multi-attribute decision-making methods have been gradually applied in recent years, there is still a lack of hierarchical modeling approaches for structural and non-structural indicators. This makes it impossible to effectively characterize the long-term degradation trend of tunnel structures and the evolution potential of non-structural risks. Furthermore, they fail to design targeted evaluation mechanisms for the differences in risk between sections, resulting in insufficient mining of constraint information for different risk sections. This leads to low sensitivity in identifying high-risk sections and discontinuous risk level classification, seriously affecting the accuracy and reliability of safety decisions.

[0003] Therefore, there is an urgent need for a typical section safety evaluation method that can combine tunnel group risk zoning information, structural degradation and non-structural risk evolution characteristics, to achieve robust safety evaluation of unmonitored sections or newly built sections, and to meet the practical application needs of long-term safety management and refined monitoring of tunnel groups in complex mountainous areas. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for long-term safety evaluation of tunnel groups in complex and challenging mountainous areas, in order to improve the aforementioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0005] Firstly, this application provides a long-term safety evaluation method for tunnel groups in complex and challenging mountainous areas, including:

[0006] To obtain the risk factors and risk weights of tunnel groups in complex and dangerous mountainous areas;

[0007] Based on risk factors and risk weights, the tunnel group in the complex and dangerous mountainous area is divided into risk zones to obtain a tunnel group risk zoning map. The tunnel group risk zoning map includes multiple risk zoning levels for different sections.

[0008] Based on the risk zoning map of the tunnel group, and combined with the structural stability degradation and non-structural risk evolution modeling under the risk zoning constraints, the evaluation vectors and safety evaluation level labels of multiple typical sections are obtained.

[0009] Based on the evaluation vector and safety evaluation level label of typical cross sections, and combined with the risk zoning level of the cross section to be evaluated, comparative learning under risk zoning constraints is carried out to obtain the safety evaluation result of the cross section to be evaluated.

[0010] Secondly, this application also provides a long-term safety evaluation system for complex and challenging mountain tunnel group structures, including:

[0011] The first acquisition module is used to acquire the risk factors and risk weights of complex and dangerous mountain tunnel groups.

[0012] The risk zoning module is used to zon the complex and dangerous mountain tunnel group according to risk factors and risk weights, and obtain the tunnel group risk zoning map. The tunnel group risk zoning map includes multiple risk zoning levels for the divided sections.

[0013] The second acquisition module is used to obtain evaluation vectors and safety evaluation level labels for multiple typical sections based on the tunnel group risk zoning map and combined with structural stability degradation and non-structural risk evolution modeling under risk zoning constraints.

[0014] The evaluation module is used to perform comparative learning under risk zoning constraints based on the evaluation vector and safety evaluation level label of a typical section, combined with the risk zoning level of the section to be evaluated, to obtain the safety evaluation result of the section to be evaluated.

[0015] The beneficial effects of this invention are as follows:

[0016] (1) This invention uses XGBoost feature gain analysis and SHAP contribution correction to screen key risk factors, combines subjective weights calculated by the fuzzy DEMATEL-ANP model with objective weights calculated by the interval T-spherical fuzzy entropy model, and then obtains risk weights based on the goodness of fit between the evaluation value distribution and the normal distribution. By combining subjective and objective weighting, the invention incorporates experts' engineering understanding of the mechanism of action of risk factors, while also mining objective information from historical monitoring data. At the same time, the goodness of fit calibration ensures the statistical stability of the weight distribution, effectively avoiding the problem of weakening key risks or amplifying secondary risks caused by a single weighting method.

[0017] (2) This invention characterizes the risk differences of different division sections through indicators such as relative advantage degree and comprehensive advantage degree, and adjusts the ranking results by combining the difference in benefit ratio value and group-individual consistency, thereby achieving a robust division of risk levels and solving the pain points of traditional methods such as insensitivity to high-risk section identification and discontinuous partitioning results. At the same time, a stability degradation model is constructed for structural indicators, and a risk evolution and section correction mechanism is designed for non-structural indicators. It takes into account the section risk differences, long-term structural degradation characteristics and non-structural risk evolution potential of complex and dangerous mountain tunnel groups, so that the evaluation vector can accurately reflect the safety status of the section in the real service environment;

[0018] (3) The present invention also introduces comparative learning under risk partition constraints. By constructing comparative sample pairs through the evaluation vector of typical cross sections, safety evaluation level labels and risk partition levels, the risk partition differences are strengthened in the loss function to constrain the embedding space distance, so that the mapping network learns more discriminative safety evaluation features.

[0019] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the long-term safety evaluation method for complex and dangerous mountain tunnel groups as described in this embodiment of the invention.

[0022] Figure 2 This is a schematic diagram of the long-term safety evaluation system for complex and dangerous mountain tunnel groups as described in this embodiment of the invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0024] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] Example 1:

[0026] This embodiment provides a method for long-term safety evaluation of tunnel groups in complex and dangerous mountainous areas.

[0027] It should be noted that different sections of the complex and dangerous mountain tunnel group have different geological conditions, environmental effects, and operational loads. Therefore, the impact of the cross-section on different sections of the complex and dangerous mountain tunnel group must be considered when conducting safety assessments.

[0028] Specifically, within the same tunnel group, the stability of the surrounding rock varies greatly in different sections. Some sections are in a stable surrounding rock environment with good integrity and high mechanical strength, where the structure is subjected to uniform stress and the deformation rate is slow. However, some sections need to pass through fault fracture zones, weak interlayers, or karst development areas, where the surrounding rock is prone to geological disasters such as collapse and water inrush, posing a continuous threat to the bearing capacity of the tunnel lining structure.

[0029] Secondly, there are significant differences in the environmental impact factors faced by different sections. Some sections are located in dry, high-altitude areas, and are only slightly affected by environmental factors such as precipitation and earthquakes, resulting in a slower rate of structural corrosion and damage evolution. Other sections are located in rainy and humid areas, and are subject to long-term erosion by heavy rainfall and soaking by groundwater, making the lining structure prone to defects such as steel corrosion and concrete carbonization. They are also located in earthquake-prone areas, where high-frequency vibrations can easily cause structural cracks to expand and accelerate the process of structural degradation.

[0030] At the same time, different sections of the tunnel group bear different traffic loads. Some sections connecting to branch lines have low traffic volume and are mainly used by light vehicles, resulting in a lower operating load and a slower accumulation of structural fatigue damage. Some sections located on the main line bear heavy traffic, with high vehicle frequency and large axle load. Long-term repeated loading can easily lead to cracking, spalling and other damage to the lining structure, making the structural safety risk significantly higher than other sections.

[0031] Therefore, this embodiment constructs risk factors using various data and combines risk zoning information with the characteristics of structural degradation and non-structural risk evolution to achieve a safety assessment of complex and dangerous mountain tunnel groups.

[0032] See Figure 1 The figure shows that the method includes steps S1, S2, S3 and S4.

[0033] Step S1: Obtain the risk factors and risk weights of the complex and dangerous mountain tunnel group;

[0034] Step S1 includes:

[0035] Step S11: Based on XGBoost feature gain analysis and SHAP contribution correction, obtain the risk factors affecting the safety status of the tunnel structure;

[0036] Step S11 includes:

[0037] Step S111: Collect candidate risk factors affecting the safety status of the tunnel structure;

[0038] It should be noted that the risk factor is used for risk zoning in the segmentation of tunnel groups in complex and dangerous mountainous areas. Therefore, the selection of the risk factor is based on the principles of segment-scale representativeness, data availability, long-term stability, and safety-related significance. Furthermore, in this embodiment, a single tunnel can be considered as a segment.

[0039] The candidate risk factors include, but are not limited to:

[0040] Geological factors are used to characterize the stability of the geological conditions in which the tunnel is located, including fault density, surrounding rock grade, groundwater pressure, and rock mass integrity indicators.

[0041] Structural factors are used to characterize the load-bearing and protective capacity of tunnel structures, including initial support strength, design lining thickness, cross-sectional height-to-span ratio, and support level.

[0042] Environmental factors are used to characterize the impact of external natural environmental factors, including regional seismic intensity, precipitation intensity, land cover conditions, temperature variation range, and changes in land surface load.

[0043] Operational factors are used to characterize the tunnel's service status and operating conditions, including service life, traffic flow, vehicle load levels, and historical safety accident records.

[0044] The aforementioned candidate risk factors were obtained through geological survey reports, design parameters, and historical monitoring statistics.

[0045] Step S112: Calculate the gain value of each candidate risk factor using the XGBoost model, and sort the candidate risk factors by their gain values ​​to obtain the candidate risk factors ranked by importance.

[0046] ;

[0047] In the formula, This represents the gain value when splitting at the current node. and These represent the error information of the left and right child nodes, respectively. and Let represent the sum of the second derivatives of all samples in the left and right child nodes, respectively. and Both represent regularization parameters. Used to control the magnitude of node weights and prevent overfitting. Used to control the cost of splitting operations.

[0048] Step S113: Calculate the SHAP value of each candidate risk factor, and adjust the importance ranking of the candidate risk factors using the SHAP value to obtain the adjusted candidate risk factor ranking.

[0049] In this step, the SHAP value of each candidate risk factor is calculated using the SHAP interpreter, and the SHAP values ​​are used to rank them to obtain the SHAP importance ranking. When the SHAP importance ranking conflicts with the candidate risk factors in the importance ranking, weights are assigned according to the candidate risk factors in the importance ranking and the SHAP importance ranking, a comprehensive ranking score is calculated for each candidate risk factor, and the rank is re-ranked according to the comprehensive ranking score to obtain the corrected candidate risk factor ranking.

[0050] Step S114: Based on the revised ranking of candidate risk factors, perform screening to obtain multiple risk factors;

[0051] In this step, 10 to 12 risk factors are selected for subsequent risk zoning.

[0052] Simultaneously, a risk threshold is set for each risk factor to obtain a risk factor partition threshold table. Using the risk factor partition threshold table and historical monitoring data of the partitioned segments, the factor risk partition level of the corresponding risk factor in the partitioned segment is obtained. This is used to determine the membership interval, non-membership interval, and hesitation interval parameters of the interval T-spherical fuzzy number, providing input for subsequent risk weight calculation based on interval T-spherical fuzzy entropy.

[0053] It should be noted that the factor risk zoning is divided into high-risk, medium-risk, and low-risk zones, with risk thresholds determined based on historical monitoring data, regulatory limits, or expert experience. For example, taking groundwater pressure as an example, when the groundwater pressure is less than 0.3 MPa, it corresponds to a low-risk zone; when the groundwater pressure is between 0.3 and 0.6 MPa, it corresponds to a medium-risk zone; and when the groundwater pressure is greater than 0.6 MPa, it corresponds to a high-risk zone.

[0054] It is understandable that in actual engineering projects, the intensity of the impact of geological conditions, structural state, environmental effects, and operational conditions on tunnel safety varies. For example, fault density and groundwater pressure are generally more sensitive to structural safety, while the impact of some operational factors is indirect and cumulative. Simply applying equal weights to each risk factor can easily amplify secondary risks, weaken critical risks, and reduce the accuracy of risk zoning and safety assessment. Therefore, this embodiment will employ a combination of subjective weighting, objective weighting, and combined weighting methods for calculation.

[0055] Step S12: Obtain the subjective weights of each risk factor using the fuzzy DEMATEL-ANP model constructed based on expert cognition;

[0056] In this step, the causal relationship and feedback structure between risk factors are characterized by the fuzzy DEMATEL-ANP model. The fuzzy representation reduces the impact of uncertainty in expert judgment, and the ANP network structure reflects the structural importance of risk factors in the overall security system.

[0057] Step S12 includes:

[0058] Step S121: Generate a scoring matrix for risk factors based on expert opinions, and construct a fuzzy matrix using the scoring matrix to obtain the average fuzzy matrix;

[0059] In this step, a questionnaire survey was used to collect opinions from various experts. The influence relationships between various risk factors were quantified using a 0-5 scale (0 indicates no influence, 1 indicates very small influence, 2 indicates small influence, 3 indicates moderate influence, 4 indicates large influence, and 5 indicates very large influence). This yielded results from the [unclear - likely a missing section or data point]. The scoring matrix of experts .

[0060] Calculate the interval boundaries and convert the values ​​in the scoring matrix into grayscale intervals to obtain the first interval. Fuzzy Matrix of Experts , , and They represent the first The lower and upper bound matrices corresponding to the scores given by the experts.

[0061] If each expert scores the same risk factor pair multiple times (reassessment), the average of the upper and lower bound matrices is calculated to obtain the average fuzzy matrix. :

[0062] ;

[0063] In the formula, Indicates the first The average fuzzy matrix of the experts and They represent the first The lower and upper bound matrices corresponding to the scores given by the experts.

[0064] Step S122: Sharpen and normalize the average fuzzy matrix to obtain the normalized influence matrix, and then power-sum the normalized influence matrix to obtain the comprehensive influence matrix;

[0065] In this step, the average fuzzy matrix is ​​whitened according to the whitening formula to obtain the sharp value. , Represents element The clarity value, i.e., the first The risk factor affects the first The clear values ​​of each risk factor, in this embodiment, are elements. Actual representation of the first The risk factor affects the first One risk factor.

[0066] The sharpness value matrix can be obtained from the sharpness value. , , , , This represents the total number of risk factors. Then, the clarity value matrix... By performing maximum value normalization, the normalized influence matrix is ​​obtained. ,in, , This indicates taking the maximum value. Finally, the normative influence matrix... conduct By summing the powers, we obtain the comprehensive influence matrix. ,in, .

[0067] Step S123: Calculate the limiting supermatrix by integrating the influence matrix and the ANP network structure, and calculate the subjective weight of each risk factor using the limiting supermatrix.

[0068] In this step, a threshold is set for the influence relationship between various risk factors, and the comprehensive influence matrix is ​​eliminated through the ANP network structure. The elements smaller than a threshold are used to obtain the judgment matrix. A consistency check is performed on the judgment matrix to obtain the unweighted hypermatrix. Column normalization is then performed on the unweighted hypermatrix to obtain the weighted hypermatrix. .

[0069] The weighted hypermatrix is ​​calculated according to the following formula. By performing a limit solution, the limit hypermatrix is ​​obtained. :

[0070] ;

[0071] In the formula, Denotes a weighted hypermatrix. This represents the power of a matrix.

[0072] and the ultimate hypermatrix Any column in the table represents the subjective weight of the corresponding risk factor.

[0073] Step S13: Calculate the objective weights of each risk factor using the interval T-spherical fuzzy entropy model constructed based on historical monitoring data;

[0074] In this step, the interval T-spherical fuzzy entropy model is used to characterize the uncertainty and fuzziness of historical monitoring data and statistical indicators through intervalization and spherical fuzzy expression. The entropy value is used to reflect the ability of risk factors to distinguish between different segments and the degree of information contribution. This provides an objective basis for risk weights from the data level and avoids the oversensitivity of a single deterministic statistical method to outliers and sample fluctuations.

[0075] Step S13 includes:

[0076] Step S131: Construct an interval T-spherical fuzzy decision matrix using historical monitoring data;

[0077] In this step, a finite universe of discourse is defined by constructing historical monitoring data. , ,in, Indicates the first A segmented sample, , To divide the sample into segments, each sample includes One risk factor. (This is referred to as...) For a finite domain If a T-spherical fuzzy set is defined on an interval, then the following is determined: Membership interval Hesitation range Non-membership interval ,in, , Describing a finite domain One of the segmented samples, specifically:

[0078] ;

[0079] ;

[0080] ;

[0081] ;

[0082] In the formula, and They represent The lower and upper limits, and They represent The lower and upper limits, and They represent The lower and upper limits, Represents the norm constraint value.

[0083] By obtaining the fuzzy number of the segmented samples, an interval T-spherical fuzzy decision matrix, i.e., a standardized decision matrix, is constructed. Each element in the matrix contains a membership interval, a hesitation interval, and a non-membership interval.

[0084] ;

[0085] In the formula, Represents the elements in the interval T-spherical fuzzy decision matrix Fuzzy numbers, Represents the elements in the interval T-spherical fuzzy decision matrix The membership interval, Represents the elements in the interval T-spherical fuzzy decision matrix The range of hesitation, Represents the elements in the interval T-spherical fuzzy decision matrix The non-membership interval, Indicates the first One risk factor, This represents the set of benefit-oriented risk factors. This represents the set of cost-based risk factors.

[0086] Elements in the interval T-spherical fuzzy decision matrix Actual representation of the first Risk factors for each segmented sample The position of the T-spherical fuzzy decision matrix in the interval.

[0087] It should be noted that risk factors are divided into benefit-type indicators and cost-type indicators. When constructing an interval T-spherical fuzzy decision matrix, it is necessary to first determine the type of risk factor to which the risk factor belongs. Among them, for benefit-type risk factors, the larger the value, the safer it is, such as lining thickness; for cost-type risk factors, the smaller the value, the safer it is. For benefit-type risk factors, the original fuzzy numbers are directly retained; for cost-type risk factors, standardization is achieved by interchangeing membership and non-membership degrees.

[0088] Step S132: Calculate the comprehensive fuzzy entropy of each risk factor using the interval T-spherical fuzzy decision matrix, and calculate the average fuzzy entropy using the comprehensive fuzzy entropy;

[0089] In this step, for each fuzzy number in the interval T-spherical fuzzy decision matrix Define score intervals Precise interval and missing intervals :

[0090] ;

[0091] ;

[0092] ;

[0093] , ,

[0094] In the formula, and They represent The lower and upper limits, and They represent The lower and upper limits, and They represent The lower and upper limits, and Representing fuzzy numbers respectively The lower and upper limits of the corresponding membership interval, and Representing fuzzy numbers respectively The lower and upper limits of the corresponding hesitation interval, and Representing fuzzy numbers respectively The lower and upper limits of the corresponding non-membership interval. Represents the norm constraint value.

[0095] Treating the score interval, precise interval, and missing interval as three continuous random variables, we define the interval T-spherical fuzzy set. Comprehensive fuzzy entropy:

[0096] ;

[0097] In the formula, The comprehensive fuzzy entropy, Describing a finite domain, This indicates taking the absolute value. Describing a finite domain One of the segmented samples, , and They represent , and Continuous random variables, , and They represent about The score interval, exact interval, and missing interval. Represents the interval T-spherical fuzzy set China regarding Fuzzy numbers.

[0098] Calculate the average fuzzy entropy using all segmented samples:

[0099]

[0100] In the formula, express Average fuzzy entropy, To divide the sample size into segments, express The comprehensive fuzzy entropy, Represents the elements in the interval T-spherical fuzzy decision matrix Fuzzy numbers, Indicates the first One risk factor.

[0101] Step S133: Reverse map the average fuzzy entropy to obtain the objective weight of each risk factor;

[0102] In this step, the objective weights are calculated as follows:

[0103] ;

[0104] In the formula, express Objective weight, Indicates the first One risk factor, express Average fuzzy entropy, This represents the total number of risk factors. express Average fuzzy entropy, Indicates the first One risk factor.

[0105] Step S14: Using the multiplicative addition method and the linear weighted synthesis method, subjective weights and objective weights are combined under different weight calculation coefficients to obtain multiple sets of combined weights and their evaluation values ​​for each risk factor;

[0106] Step S14 includes:

[0107] Step S141: Based on subjective weights and objective weights, use the multiplicative addition method to generate the first combined weight and first evaluation value for each risk factor under different first weight calculation coefficients;

[0108] In this step, the coefficients for the first weighting are calculated. Variation conditions (with a value interval of 0.001, 0≤) ≤1), for each The weights of the first combination are calculated using the multiplication-addition method:

[0109] ;

[0110] In the formula, express The first combination weight, express Subjective weighting, This indicates the coefficient for calculating the first weight. express Objective weight, Indicates the first One risk factor.

[0111] Based on this, utilize Calculate the first evaluation value:

[0112] ;

[0113] In the formula, express The first evaluation value, express The normalized risk factor value.

[0114] Step S142: Based on subjective and objective weights, use a linear weighted synthesis method to generate the second combined weight and second evaluation value for each risk factor under different second weight calculation coefficients;

[0115] In this step, the coefficients for the second weighting are calculated. Variation conditions (with a value interval of 0.001, 0≤) ≤1), for each The weights of the second combination are calculated using the linear weighted composition method:

[0116] ;

[0117] In the formula, express The second combination weight, This indicates the coefficient for calculating the second weight. express Subjective weighting, express Objective weight, Indicates the first One risk factor.

[0118] Based on this, utilize Calculate the second evaluation value:

[0119] ;

[0120] In the formula, express The second evaluation value, express The normalized risk factor value.

[0121] Step S15: Calculate the risk weight of each risk factor using the goodness of fit between the evaluation value distribution and the normal distribution as the optimization criterion.

[0122] Step S15 includes:

[0123] Step S151: Calculate the first goodness of fit of each first weighted coefficient using the first evaluation value, and calculate the second goodness of fit of each second weighted coefficient using the second evaluation value;

[0124] In this step, the comprehensive evaluation value of each segment sample is calculated using all the first combination weights and corresponding first evaluation values ​​for each risk factor. A comprehensive evaluation value sequence is generated from the comprehensive evaluation values ​​of all segment samples. Statistical parameters (sample mean and sample standard deviation) are calculated from the comprehensive evaluation value sequence, and a normal fit is performed using these statistical parameters to obtain the theoretical probability distribution.

[0125] Calculate the cumulative distribution function of the comprehensive evaluation value sequence and the cumulative distribution function of the theoretical probability distribution. Use the Kolmogorov-Smirnov test (KS test) to calculate the KS statistic for different first-weighted coefficients using the cumulative distribution functions of the comprehensive evaluation value sequence and the theoretical probability distribution. Define the first goodness of fit as "1 - standardized KS statistic". Iterate through all first-weighted coefficients to obtain the first goodness of fit sequence.

[0126] Similarly, the second goodness-of-fit sequence is obtained.

[0127] Step S152: The first weight calculation coefficient that maximizes the first goodness of fit is used as the first parameter, and the second weight calculation coefficient that maximizes the second goodness of fit is used as the second parameter.

[0128] Step S153: Determine the risk weight of each risk factor based on the first combined weight corresponding to the first parameter and the second combined weight corresponding to the second parameter.

[0129] In this step, the first parameter is obtained. Second parameter Afterwards, if The corresponding first goodness of fit is greater than If the corresponding second goodness of fit is found, then the first combination weight corresponding to the first parameter is selected to calculate the risk weight; otherwise, the second combination weight corresponding to the second parameter is selected to calculate the risk weight.

[0130] In this embodiment, the fuzzy DEMATEL-ANP model, the interval T-spherical fuzzy entropy model, and the combined weight optimization criterion based on the goodness of fit of the evaluation value distribution are combined to collaboratively determine the risk weight from three levels: factor action mechanism modeling, data uncertainty characterization, and weight fusion rationality calibration. This results in risk weights that simultaneously possess engineering cognitive consistency, data-driven objectivity, and statistical distribution stability.

[0131] Step S2: Divide the complex and dangerous mountain tunnel group into risk zones according to risk factors and risk weights to obtain a tunnel group risk zoning map. The tunnel group risk zoning map includes multiple risk zoning levels for different sections.

[0132] In this step, under the risk zoning scenario of tunnel groups in complex and dangerous mountainous areas, the TODIM-VIKOR method is designed to simultaneously characterize the relative advantages and disadvantages of different sections under multiple risk factors. Based on the trade-off between the optimal compromise solution and local extreme risks, a robust classification of the risk level of the tunnel group is achieved, thereby avoiding the problem that a single ranking method is insensitive to the identification of high-risk sections or the zoning results are discontinuous.

[0133] Step S2 includes:

[0134] Step S21: Calculate the relative dominance of any two sections in the complex and dangerous mountain tunnel group based on the risk factors and risk weights, and calculate the comprehensive dominance between any two sections based on the relative dominance.

[0135] In this step, the formula for calculating relative dominance is:

[0136] ;

[0137] ;

[0138] In the formula, express Down Compared to The relative advantage degree Indicates the first One risk factor, and They represent the first The and the first Each section is divided into segments. express The relative combination weights, Indicates the interval T-spherical fuzzy distance. and These represent the elements in the interval T-spherical fuzzy decision matrix. and Fuzzy numbers, This represents the total number of risk factors. This represents the risk preference coefficient. express Risk weights, This represents the maximum value among all risk weights.

[0139] Among them, the elements in the interval T-spherical fuzzy decision matrix Actual representation of the first Risk factors for each segment The position of the element in the interval T-spherical fuzzy decision matrix; elements in the interval T-spherical fuzzy decision matrix. Actual representation of the first Risk factors for each segment The position of the T-spherical fuzzy decision matrix in the interval.

[0140] Overall advantage is:

[0141] ;

[0142] In the formula, express Down Compared to The overall advantages To divide the number of sections, express Down Compared to The relative advantage.

[0143] Step S22: Calculate the positive and negative ideal solutions for each risk factor using the comprehensive dominance factor;

[0144] In this step, the positive ideal solution and the negative ideal solution are:

[0145] ;

[0146] ;

[0147] In the formula, express The ideal solution, express The negative ideal solution, This indicates taking the maximum value. This indicates taking the minimum value. express Down Compared to The overall advantages.

[0148] Step S23: Calculate the group benefit value and individual regret value for each partition segment using the positive ideal solution and the negative ideal solution;

[0149] In this step, the group benefit value reflects the overall deviation of the proposed solution from the ideal solution. The smaller the value, the closer the proposed solution is to the ideal state, reflecting the comprehensive advantages of the proposed solution under the group's preference. The individual regret value reflects the maximum local loss that the proposed solution may suffer under various evaluation criteria. The smaller the value, the better the proposed solution performs under extreme criteria, characterizing the maximum tolerance limit of individual decision-makers for the disadvantages of the proposed solution.

[0150] Specifically, the group benefit value and the individual regret value are:

[0151] ;

[0152] ;

[0153] ;

[0154] ;

[0155] In the formula, Indicates the first The group benefit value of each segment. This represents the total number of risk factors. express Risk weights, This represents the difference quantization function. Represents the elements in the interval T-spherical fuzzy decision matrix Fuzzy numbers, Indicates the first Individual regret value for each segment.

[0156] Step S24: Calculate the benefit ratio value for each segment using the group benefit value and the individual regret value;

[0157] In this step, the benefit ratio is a weighted composite index of group benefit value and individual regret value, used to balance group benefits and individual regret, and to provide a basis for compromise decision-making.

[0158] Specifically, the profit ratio is:

[0159] ;

[0160] In the formula, Indicates the first The profit ratio value of each divided segment This represents the decision-making mechanism coefficient, which is typically taken as 0.5.

[0161] Step S25: Sort all the divided segments according to the benefit ratio value to obtain the sorting result;

[0162] In this step, the segments are sorted according to their profit ratio values ​​from smallest to largest.

[0163] Step S26: Calculate the benefit ratio gap and group-individual consistency based on the ranking results;

[0164] In this step, the profit ratio difference is the difference between the profit ratio of the later segment and the profit ratio of the earlier segment in adjacent segments.

[0165] Condition ① is constructed based on the difference in the interest ratio: ,in, Indicates the first The profit ratio value of each divided section.

[0166] Group-individual consistency is condition ②: In adjacent partitioned segments, the group benefit value of the preceding partitioned segment is greater than the group benefit value of the following partitioned segment, or the individual regret value of the preceding partitioned segment is less than the individual regret value of the following partitioned segment.

[0167] Step S27: Adjust the ranking results by the difference in benefit ratio values ​​and group-individual consistency, and perform level mapping based on the adjusted ranking results to obtain the risk zoning level of each segment.

[0168] In this step, if the ranking result satisfies both condition ① and condition ②, the ranking result remains unchanged. If only condition ① is satisfied but condition ② is not, then the adjacent partitions that do not meet the condition are all compromise solutions. A secondary ranking is then performed on the samples within the compromise solution group, using the minimum group benefit value or the minimum individual regret value as the core criterion.

[0169] If condition ① is not satisfied, then it is satisfied. The maximum value, then arrive Both are compromise solutions. arrive They represent the first The and the first Each section is divided into segments. It is among all adjacent dividing sections that satisfy less than The maximum value at this time arrive These compromise solutions constitute a compromise solution group. For the division sections within the compromise solution group, a secondary ranking is performed using a dual-index weighted ranking method (ranking is done by calculating a comprehensive performance score based on group benefit value and individual regret value).

[0170] Based on the adjusted sorting results, the risk zones are divided proportionally to obtain the risk zoning level for each zone. The risk zoning levels include high risk, medium risk, and low risk.

[0171] Step S3: Based on the tunnel group risk zoning map, and combined with the structural stability degradation and non-structural risk evolution modeling under risk zoning constraints, obtain the evaluation vectors and safety evaluation level labels for multiple typical sections;

[0172] Step S31: Based on the tunnel group risk zoning map, select multiple typical cross sections within the divided sections corresponding to different risk zoning levels;

[0173] Step S32: Obtain the cross-sectional evaluation index and safety evaluation level label for each typical cross-section, wherein the cross-sectional evaluation index includes structural indexes and non-structural indexes;

[0174] In this step, the cross-sectional evaluation indicators include quantifiable structural and non-structural indicators. For example, structural indicators include lining thickness, cross-sectional height-to-span ratio, cross-sectional span, and arch thickness, while non-structural indicators may include service life, design and construction quality, seismic intensity, traffic flow, and maintenance frequency. Structural indicators reflect the cross-sectional load-bearing capacity and stability, while non-structural indicators reflect environmental and operational impacts. Specific cross-sectional evaluation indicators can be determined through expert evaluation and other methods.

[0175] It should be noted that the cross-sectional evaluation indicators do not include structural stress monitoring data that are difficult to measure. This is because obtaining data such as the contact pressure between the primary support and secondary lining, the stress of the steel arch frame, the axial force of the secondary lining, and the bending moment of the secondary lining presents numerous practical obstacles. In complex and challenging mountainous terrain, the steep slopes make the transportation and installation of specialized monitoring equipment such as stress gauges and pressure sensors extremely difficult, with some high-risk sections even lacking suitable working space. Furthermore, this type of data requires continuous long-term collection, but the complex operating environment of mountain tunnel groups, including surrounding rock deformation and groundwater erosion, can easily lead to equipment damage or data distortion. Traffic flow interference and maintenance limitations during operation further reduce the rate of obtaining effective data. In addition, large-scale deployment of monitoring equipment is difficult to achieve comprehensive coverage for long-distance tunnel groups.

[0176] Therefore, this embodiment selects structural indicators such as lining thickness and cross-sectional span, which are easy to obtain through design documents or simple on-site measurements, as well as non-structural indicators such as service life and traffic flow, which can be statistically analyzed through operation records. This avoids the problem of difficulty in obtaining structural stress monitoring data, and through these quantifiable and easily obtainable indicators, combined with subsequent degradation and evolution modeling, it can indirectly reflect the structural bearing capacity and long-term safety potential, ensuring the engineering practicality and operability of the evaluation method.

[0177] The safety assessment level is divided into four levels, labeled 1 to 4, corresponding to Level I, Level II, Level III, and Level IV. Furthermore, the safety assessment level of a typical cross-section can be determined through historical monitoring, structural analysis, and expert consultation.

[0178] Step S33: Construct a risk scenario identification vector for each typical section using risk zoning level and section risk intensity factor;

[0179] In this step, the first... A typical cross-section is And obtain the corresponding risk zoning level. , ,in, express This is a low-risk zone. express The area is classified as medium-risk. express This is a high-risk zone.

[0180] Risk scenario identification vector is ,in, express The segment risk intensity factor, among which, The risk intensity factor of a segment is determined based on the position of the segment in the overall ranking of all segments.

[0181] Step S34: Model the stability degradation and risk evolution of the cross-section evaluation indicators to obtain the structural stability evaluation vector and the non-structural risk evolution evaluation vector;

[0182] Step S34 includes:

[0183] Step S341: Map each structural index to a relative stability value within the corresponding segment;

[0184] In this step, the maximum and minimum values ​​of each structural index mapped to the corresponding segment are obtained, and then the relative stability value of each structural index for each typical section is calculated:

[0185] ;

[0186] In the formula, Indicates the first The relative stability value of each structural indicator. Indicates the first The values ​​of structural indicators and They represent the first The minimum and maximum values ​​of each structural indicator within its corresponding segment.

[0187] Understandably, the calculation of relative stability values ​​is based on the corresponding segment, rather than the entire tunnel group. A benchmark is constructed by using the maximum and minimum values ​​of the indicators within the segment. For example, within the same high-risk segment, the lining thickness of different typical sections may vary. The relative stability value can accurately reflect the structural advantages of a single section within the segment, avoiding the obscuring of individual structural differences by the overall risk level of the segment.

[0188] Step S342: Introduce a structural degradation function and calculate the structural stability characterization using relative stability values;

[0189] Understandably, different risk zoning levels within complex and challenging mountain tunnel complexes correspond to different structural degradation environments. For example, high-risk sections (with fault fracture zones and areas with strong groundwater influence) are more significantly affected by surrounding rock deformation and environmental erosion, resulting in a faster degradation rate; while structural degradation in low-risk sections (such as stable surrounding rock areas) is negligible. Therefore, the structural degradation function can accurately match the actual degradation patterns of different zoning sections.

[0190] Specifically, this step employs a segment risk-guided structural degradation mechanism to construct structural stability representations for each structural index of each typical section:

[0191] ;

[0192] In the formula, Indicates the first Structural stability characterization of structural indicators Indicates the first The relative stability value of each structural indicator. Represents an exponential function. Indicates the structural risk sensitivity coefficient. Indicates the first The risk zoning level of the section to which a typical cross-section belongs.

[0193] Step S343: Perform structural encoding on the structural stability characterization to obtain the structural stability evaluation vector;

[0194] In this step, a stability vector for each typical section is constructed through multiple structural stability characteristics. Then, the stability vector is encoded through an encoding function to obtain the structural stability evaluation vector for each typical section. The structural encoding function can be a linear mapping function or an MLP.

[0195] Step S344: Calculate the state vector and evolution vector of each typical section based on the non-structural indices;

[0196] In this step, all non-structural indicators for each typical cross-section are normalized to obtain the corresponding state vector. The evolution vector of the corresponding typical cross-section is calculated using the evolution function and all non-structural indicators of the typical cross-section. The evolution function can be set as a linearly increasing function or an exponentially decaying function with respect to time, depending on the characteristics of the indicator, to characterize the risk evolution trend of the indicator with the service life or environmental changes.

[0197] Step S345: Correct the state vector by applying the segment risk consistency constraint using the non-structural risk correction function to obtain the corrected state vector;

[0198] It is understandable that the impact of non-structural indicators (such as service life, traffic flow, and seismic intensity) on tunnel safety is not solely determined by the indicator values ​​themselves, but is also constrained by the risk environment of the section. For example, under the same traffic flow, tunnels in high-risk sections (such as those crossing fault fracture zones) experience more significant structural disturbances caused by vehicle loads, and the contribution of non-structural risks should be higher. Conversely, under the same service life, tunnels in high-risk sections (such as those experiencing heavy rainfall or severe groundwater erosion) experience faster material aging and functional degradation, requiring further strengthening of non-structural risks. Although step S344 significantly reduces the structural stability characterization of high-risk sections through the structural degradation function, reflecting the segmental differences in structural risk, without correction for non-structural risks, structural risks may have already intensified with the segmental level, while non-structural risks remain at a uniform scale. Therefore, correction can enable non-structural risks to synergize with structural risks.

[0199] Specifically, the state vector is corrected for different typical cross-sections:

[0200] ;

[0201] In the formula, Indicates the first The correction state vector of a typical cross section, Indicates the first The state vector of a typical cross-section. This represents the non-structural risk amplification factor. Indicates the first The risk zoning level of the section to which a typical cross-section belongs.

[0202] As can be seen, in the high-risk zones, the risk contribution of the corresponding non-structural factors is amplified.

[0203] Step S346: Construct an unstructured risk evolution evaluation vector using the evolution vector and the correction state vector.

[0204] In this step, a comprehensive evolution vector for each typical section is constructed using the evolution vectors of all non-structural indicators, and a comprehensive correction state vector for each typical section is constructed using the correction state vectors of all non-structural indicators. The combination of the comprehensive evolution vector and the comprehensive correction state vector is then input into an encoding function for encoding to obtain the non-structural risk evolution evaluation vector.

[0205] Step S35: Construct the evaluation vector for each typical section using the risk scenario identification vector, structural stability evaluation vector, and non-structural risk evolution evaluation vector.

[0206] Understandably, by mapping the relative stability and modeling the degradation of structural indicators, the long-term performance changes of the structure can be reflected. Furthermore, through state vectors, evolution vectors, and risk correction functions, the evolutionary trends of non-structural indicators over time, influenced by environmental and usage factors, can be reflected. The risk impact of high-risk sections can be amplified based on their risk levels. Compared to methods that do not consider section differences, this modeling approach can more accurately reflect the risk differences between different cross-sections in a real tunnel group environment.

[0207] Step S4: Based on the evaluation vector and safety evaluation level label of the typical section, and combined with the risk zoning level of the section to be evaluated, comparative learning under the risk zoning constraint is carried out to obtain the safety evaluation result of the section to be evaluated.

[0208] Step S41: Construct comparative sample pairs using the evaluation vectors, safety evaluation level labels, and risk zoning levels of typical cross sections;

[0209] In this step, the comparison sample pairs include positive sample pairs and negative sample pairs. Specifically, for any typical section, a positive sample pair is constructed by using typical sections that are equal to its safety evaluation level label and risk zoning level; otherwise, it is a negative sample pair.

[0210] Step S42: Map the evaluation vectors of typical cross sections to the safety evaluation embedding space using a mapping network;

[0211] In this step, the mapping network uses a multilayer perceptron (MLP).

[0212] Step S43: Define the contrastive loss function for risk partitioning constraints:

[0213] In this step, when the risk partition levels of the two cross sections differ significantly, the network more strictly distinguishes them in the embedding space. Therefore, the contrastive loss function is defined as:

[0214] ;

[0215] ;

[0216] In the formula, This represents the contrastive loss function. Indicates the number of typical cross-sections. Indicates the first A typical cross-section and the first Risk zoning weighting coefficients for a typical cross section Indicates the first A typical cross-section and the first Are the safety evaluation level labels the same for each typical cross-section? and They represent the first A typical cross-section and the first Embedding vectors of a typical cross section, Represents Euclidean distance. This indicates taking the absolute value. Indicates the risk zoning sensitivity coefficient. This indicates taking the absolute value.

[0217] in, This indicates that the first A typical cross-section and the first If the safety evaluation level labels for all typical cross sections are the same, then the label is 0.

[0218] Step S44: Optimize the network parameters of the mapping network by comparing sample pairs and the contrastive loss function;

[0219] In this step, gradient descent is used to optimize the network parameters.

[0220] Step S45: Map the evaluation vector of the section to be evaluated to the same embedding space through the optimized mapping network to obtain the safety evaluation level of the section to be evaluated.

[0221] In this step, the evaluation vector of the section to be evaluated is mapped to the embedding space to obtain the embedding vector of the section to be evaluated. Then, the embedding space distance between the section to be evaluated and the typical section in the embedding space is calculated by using nearest neighbor or weighted KNN, thereby obtaining the safety evaluation level of the section to be evaluated.

[0222] Therefore, this step enhances the distance discrimination of high-risk section sections by integrating risk zoning constraint information from typical sections in the embedding space. This not only enables robust safety evaluation of unmonitored or newly constructed sections based on the embedding representation learned from typical sections, without relying on full-section on-site structural stress monitoring data, but also significantly improves the sensitivity of high-risk section evaluation. This compensates for the shortcomings of traditional methods in utilizing section constraint information, and provides a scientific decision-making basis for the long-term safety management and refined monitoring of tunnel groups in complex and dangerous mountainous areas.

[0223] Example 2:

[0224] As shown in Figure 2, this embodiment provides a long-term safety evaluation system for complex and dangerous mountain tunnel group structures. The system includes:

[0225] The first acquisition module is used to acquire the risk factors and risk weights of complex and dangerous mountain tunnel groups.

[0226] The risk zoning module is used to zon the complex and dangerous mountain tunnel group according to risk factors and risk weights, and obtain the tunnel group risk zoning map. The tunnel group risk zoning map includes multiple risk zoning levels for the divided sections.

[0227] The second acquisition module is used to obtain evaluation vectors and safety evaluation level labels for multiple typical sections based on the tunnel group risk zoning map and combined with structural stability degradation and non-structural risk evolution modeling under risk zoning constraints.

[0228] The evaluation module is used to perform comparative learning under risk zoning constraints based on the evaluation vector and safety evaluation level label of a typical section, combined with the risk zoning level of the section to be evaluated, to obtain the safety evaluation result of the section to be evaluated.

[0229] The first acquisition module includes:

[0230] The first acquisition unit is used to acquire risk factors affecting the safety status of tunnel structures based on XGBoost feature gain analysis and SHAP contribution correction.

[0231] Subjective assignment unit, used to obtain the subjective weights of each risk factor in the fuzzy DEMATEL-ANP model constructed based on expert cognition;

[0232] The objective assignment unit is used to calculate the objective weights of each risk factor based on the interval T-spherical fuzzy entropy model constructed based on historical monitoring data.

[0233] The combination unit is used to combine subjective weights and objective weights under different weight calculation coefficients using the multiplicative addition method and the linear weighted synthesis method to obtain multiple sets of combined weights and their evaluation values ​​for each risk factor.

[0234] The first calculation unit is used to calculate the risk weight of each risk factor based on the goodness of fit between the evaluation value distribution and the normal distribution as the optimization criterion.

[0235] The second acquisition module includes:

[0236] The selection unit is used to select multiple typical cross-sections within the divided sections corresponding to different risk zoning levels based on the tunnel group risk zoning map.

[0237] The second acquisition unit is used to acquire the cross-sectional evaluation index and safety evaluation level label for each typical cross-section. The cross-sectional evaluation index includes structural indexes and non-structural indexes.

[0238] The first construction unit is used to construct a risk scenario identification vector for each typical section by using the risk zoning level and the section risk intensity factor.

[0239] The second building unit is used to model the stability degradation and risk evolution of the cross-section evaluation index, and obtain the structural stability evaluation vector and the non-structural risk evolution evaluation vector.

[0240] The third building unit is used to construct the evaluation vector for each typical section through the risk scenario identification vector, the structural stability evaluation vector, and the non-structural risk evolution evaluation vector.

[0241] The evaluation module includes:

[0242] The construction unit is used to construct comparative sample pairs using evaluation vectors of typical cross sections, safety evaluation level labels, and risk zoning levels.

[0243] The first mapping unit is used to map the evaluation vector of a typical section to the safety evaluation embedding space through a mapping network;

[0244] Define a unit to define the contrastive loss function for risk partitioning constraints:

[0245] The optimization unit is used to optimize the network parameters of the mapping network by comparing sample pairs and the comparison loss function;

[0246] The second mapping unit is used to map the evaluation vector of the section to be evaluated to the same embedding space through the optimized mapping network, so as to obtain the safety evaluation level of the section to be evaluated.

[0247] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0248] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0249] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for long-term safety evaluation of complex and dangerous mountain tunnel group structure, characterized in that, include: To obtain the risk factors and risk weights of tunnel groups in complex and dangerous mountainous areas; Based on risk factors and risk weights, the tunnel group in the complex and dangerous mountainous area is divided into risk zones to obtain a tunnel group risk zoning map. The tunnel group risk zoning map includes multiple risk zoning levels for different sections. Based on the risk zoning map of the tunnel group, and combined with the structural stability degradation and non-structural risk evolution modeling under the risk zoning constraints, the evaluation vectors and safety evaluation level labels of multiple typical sections are obtained. Based on the evaluation vector and safety evaluation level label of typical cross sections, and combined with the risk zoning level of the cross section to be evaluated, comparative learning under risk zoning constraints is carried out to obtain the safety evaluation result of the cross section to be evaluated.

2. The long-term safety evaluation method for complex and dangerous mountain tunnel groups according to claim 1, characterized in that, The acquisition of risk factors and risk weights for complex and challenging mountain tunnel groups includes: Based on XGBoost feature gain analysis and SHAP contribution correction, risk factors affecting the safety status of tunnel structures are obtained. The subjective weights of each risk factor are obtained by using a fuzzy DEMATEL-ANP model constructed based on expert cognition. The objective weights of each risk factor are calculated using an interval T-spherical fuzzy entropy model constructed based on historical monitoring data. By employing the multiplicative addition method and the linear weighted synthesis method, subjective weights and objective weights are combined under different weight calculation coefficients to obtain multiple sets of combined weights and their evaluation values ​​for each risk factor. The risk weight of each risk factor is calculated based on the goodness of fit between the evaluation value distribution and the normal distribution.

3. The long-term safety evaluation method for complex and dangerous mountain tunnel groups according to claim 1, characterized in that, The process of risk zoning of tunnel groups in complex and dangerous mountainous areas based on risk factors and risk weights yields a risk zoning map of the tunnel groups, including: The relative advantage degree of any two sections in a complex and dangerous mountain tunnel group is calculated based on risk factors and risk weights, and the comprehensive advantage degree between any two sections is calculated based on the relative advantage degree. The positive and negative ideal solutions for each risk factor are calculated using the comprehensive dominance factor. The group benefit value and individual regret value of each segment are calculated using the positive ideal solution and the negative ideal solution; The benefit ratio of each segment is calculated using the group benefit value and the individual regret value. All the partitioned sections are sorted by their benefit ratio values ​​to obtain the sorting results; The difference in benefit ratios and group-individual consistency are calculated based on the ranking results. The ranking results are adjusted by the difference in interest ratio and the consistency between the group and the individual, and the risk zoning level of each segment is obtained by mapping the adjusted ranking results.

4. The long-term safety evaluation method for complex and dangerous mountain tunnel group structures according to claim 1, characterized in that, The process of obtaining evaluation vectors and safety evaluation level labels for multiple typical cross-sections includes: Based on the tunnel group risk zoning map, multiple typical cross sections are selected in the divided sections corresponding to different risk zoning levels. Obtain the cross-sectional evaluation indicators and safety evaluation level labels for each typical cross-section. The cross-sectional evaluation indicators include structural indicators and non-structural indicators. A risk scenario identification vector for each typical section is constructed by using risk zoning level and section risk intensity factor. Stability degradation and risk evolution models are performed on the cross-sectional evaluation indicators to obtain structural stability evaluation vectors and non-structural risk evolution evaluation vectors; The evaluation vector for each typical section is constructed by using risk scenario identification vector, structural stability evaluation vector, and non-structural risk evolution evaluation vector.

5. The long-term safety evaluation method for complex and dangerous mountain tunnel groups according to claim 4, characterized in that, The stability degradation and risk evolution modeling of the cross-sectional evaluation indicators yields structural stability evaluation vectors and non-structural risk evolution evaluation vectors, including: Each structural index is mapped to a relative stability value within the corresponding segment; A structural degradation function is introduced, and the structural stability characterization is calculated through relative stability values; The structural stability characterization is encoded to obtain the structural stability evaluation vector; Calculate the state vector and evolution vector of each typical section based on non-structural indicators; The corrected state vector is obtained by applying segment risk consistency constraint correction to the state vector using a non-structural risk correction function. An evolutionary evaluation vector for non-structural risk is constructed using an evolution vector and a correction state vector.

6. The long-term safety evaluation method for complex and dangerous mountain tunnel groups according to claim 1, characterized in that, The evaluation vector and safety evaluation level label based on typical cross sections, combined with the risk zoning level of the cross section to be evaluated, are used for comparative learning under risk zoning constraints to obtain the safety evaluation results of the cross section to be evaluated, including: Comparative sample pairs are constructed using evaluation vectors of typical cross sections, safety evaluation level labels, and risk zoning levels. The evaluation vectors of typical sections are mapped to the safety evaluation embedding space through a mapping network; Define the contrastive loss function for risk partitioning constraints: The network parameters of the mapping network are optimized by comparing sample pairs and the comparison loss function. The optimized mapping network maps the evaluation vector of the section to be evaluated to the same embedding space, thereby obtaining the safety evaluation level of the section to be evaluated.

7. A long-term safety evaluation system for complex and challenging mountain tunnel complexes, characterized in that, include: The first acquisition module is used to acquire the risk factors and risk weights of complex and dangerous mountain tunnel groups. The risk zoning module is used to zon the complex and dangerous mountain tunnel group according to risk factors and risk weights, and obtain the tunnel group risk zoning map. The tunnel group risk zoning map includes multiple risk zoning levels for the divided sections. The second acquisition module is used to obtain evaluation vectors and safety evaluation level labels for multiple typical sections based on the tunnel group risk zoning map and combined with structural stability degradation and non-structural risk evolution modeling under risk zoning constraints. The evaluation module is used to perform comparative learning under risk zoning constraints based on the evaluation vector and safety evaluation level label of a typical section, combined with the risk zoning level of the section to be evaluated, to obtain the safety evaluation result of the section to be evaluated.

8. The long-term safety evaluation system for complex and challenging mountain tunnel groups according to claim 7, characterized in that, The first acquisition module includes: The first acquisition unit is used to acquire risk factors affecting the safety status of tunnel structures based on XGBoost feature gain analysis and SHAP contribution correction. Subjective assignment unit, used to obtain the subjective weights of each risk factor in the fuzzy DEMATEL-ANP model constructed based on expert cognition; The objective assignment unit is used to calculate the objective weights of each risk factor based on the interval T-spherical fuzzy entropy model constructed based on historical monitoring data. The combination unit is used to combine subjective weights and objective weights under different weight calculation coefficients using the multiplicative addition method and the linear weighted synthesis method to obtain multiple sets of combined weights and their evaluation values ​​for each risk factor. The first calculation unit is used to calculate the risk weight of each risk factor based on the goodness of fit between the evaluation value distribution and the normal distribution as the optimization criterion.

9. The long-term safety evaluation system for complex and challenging mountain tunnel groups according to claim 7, characterized in that, The second acquisition module includes: The selection unit is used to select multiple typical cross-sections within the divided sections corresponding to different risk zoning levels based on the tunnel group risk zoning map. The second acquisition unit is used to acquire the cross-sectional evaluation index and safety evaluation level label for each typical cross-section. The cross-sectional evaluation index includes structural indexes and non-structural indexes. The first construction unit is used to construct a risk scenario identification vector for each typical section by using the risk zoning level and the section risk intensity factor. The second building unit is used to model the stability degradation and risk evolution of the cross-section evaluation index, and obtain the structural stability evaluation vector and the non-structural risk evolution evaluation vector. The third building unit is used to construct the evaluation vector for each typical section through the risk scenario identification vector, the structural stability evaluation vector, and the non-structural risk evolution evaluation vector.

10. The long-term safety evaluation system for complex and challenging mountain tunnel groups according to claim 7, characterized in that, The evaluation module includes: The construction unit is used to construct comparative sample pairs using evaluation vectors of typical cross sections, safety evaluation level labels, and risk zoning levels. The first mapping unit is used to map the evaluation vector of a typical section to the safety evaluation embedding space through a mapping network; Define a unit to define the contrastive loss function for risk partitioning constraints: The optimization unit is used to optimize the network parameters of the mapping network by comparing sample pairs and the comparison loss function; The second mapping unit is used to map the evaluation vector of the section to be evaluated to the same embedding space through the optimized mapping network, so as to obtain the safety evaluation level of the section to be evaluated.

Images