Sand dolomite tunnel disaster risk intelligent evaluation method and system

Abstract

The present application belongs to the technical field of tunnel engineering disaster assessment, and provides a sanding dolomite tunnel disaster risk intelligent assessment method and system, which comprises the following steps: performing risk grade jump analysis through historical disaster risk assessment data, identifying risk grade jump assessment, and statistically judging whether the current model has a jump phenomenon; if so, performing nonlinear coupling analysis on the variables, determining whether the risk jump is caused by nonlinear coupling of the variables, and constructing a nonlinear variable set; and judging the nonlinear relationship identification ability of the current model through a nonlinear identification index; if the identification ability is weak, a coupling-superposition double module model is constructed with coupling increment as the link; and finally, according to the comparison result of the current variables and the nonlinear variable set, the corresponding model is dynamically called to complete real-time risk assessment, solving the problems of ignoring variable nonlinear coupling, frequent risk grade jumps and low assessment accuracy of the existing model.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of tunnel engineering disaster assessment, and in particular relates to a method and system for intelligently assessing disaster risks of a dolomitic sandstone tunnel. BACKGROUND

[0002] Dolomitic sandstone is a common complex geological body in tunnel engineering, which has loose rock structure and poor mechanical properties, and is easily affected by groundwater, ground stress and construction disturbance, thereby inducing disasters such as collapse, water gushing and piping, which seriously threatens the safety of engineering construction and the stability of operation. Therefore, accurate disaster risk assessment of a dolomitic sandstone tunnel is a core link to ensure engineering safety.

[0003] Existing methods for assessing disaster risks of a dolomitic sandstone tunnel are mostly based on linear models, and only consider the independent contribution of a single variable (such as sanding grade, groundwater pressure, and excavation footage) to the risk, ignoring the nonlinear coupling effect between variables, for example, the risk caused by the synergistic effect of high sanding grade and high groundwater pressure is much greater than the linear superposition risk of the two alone. Such linear evaluation ideas have the following key defects:

[0004] Frequent risk level jumps: Since the nonlinear coupling mutation effect of the variables cannot be captured, the current model is prone to cross-level jumps in risk level evaluation, and cannot distinguish whether the jump is a normal occasional fluctuation of the project or a defect of the model itself, resulting in low credibility of the evaluation results;

[0005] Insufficient evaluation accuracy: the linear model has a large deviation in quantifying the nonlinear coupling risk, which is prone to underestimating the risk of high synergistic coupling scenarios or overestimating the risk of low synergistic scenarios, making it difficult to meet the precise needs of engineering safety control;

[0006] Poor real-time adaptability: existing models are mostly fixed structures and cannot dynamically adjust the evaluation logic according to the coupling state of the current variables, and have weak real-time evaluation capability in the face of complex and variable geological-construction conditions of dolomitic sandstone tunnels.

[0007] Therefore, the application provides a method and system for intelligently assessing disaster risks of a dolomitic sandstone tunnel. SUMMARY

[0008] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background art.

[0009] The technical scheme adopted by the application to solve its technical problems is: a method for intelligently assessing disaster risks of a dolomitic sandstone tunnel, characterized by comprising:

[0010] Step one: through the risk grade jump analysis of the risk assessment of the sanding dolomite tunnel historical disasters, identify the risk grade jump assessment, and statistically analyze the risk grade jump assessment to determine whether there is a risk grade jump phenomenon when the current model assesses the disaster risk;

[0011] Step two: if there is, perform nonlinear coupling analysis on the variables of the risk grade jump assessment, identify the nonlinear variable pair, and determine whether the risk grade jump phenomenon is caused by variable nonlinear coupling. If so, mark the risk grade jump assessment as nonlinear jump assessment, and integrate the nonlinear variable pair into a nonlinear variable set;

[0012] Step three: statistically analyze the nonlinear jump assessment and determine whether the current model has weak nonlinear relationship identification ability by combining the risk grade jump degree of the nonlinear jump assessment;

[0013] Step four: if so, construct a coupling-superposition double module model based on the nonlinear coupling relationship of the nonlinear variable pair;

[0014] Step five: for the current disaster risk assessment, compare and analyze the variables and the nonlinear variable set to determine whether there is a risk grade jump risk. If there is, call the coupling-superposition double module model to assess the current disaster risk.

[0015] Further, the identification method of the risk grade jump assessment is:

[0016] For any historical disaster risk assessment:

[0017] Determine the risk grade jump amplitude by subtracting the risk grade of the historical disaster risk assessment from the risk grade of the previous historical disaster risk assessment. If the risk grade jump amplitude is greater than or equal to 2 levels, mark the historical disaster risk assessment as a risk grade jump assessment;

[0018] Statistically analyze the proportion of risk grade jump assessments in historical disaster risk assessments to obtain the risk grade jump proportion;

[0019] Compare the risk grade jump proportion with the preset proportion. If the risk grade jump proportion is greater than the preset proportion, there is a risk grade jump phenomenon when the current model assesses the disaster risk.

[0020] Further, the identification method of the nonlinear variable pair is:

[0021] For any pair of variables, respectively take the two variables as independent variables and the risk assessment grade as the dependent variable, and standardize the independent variables and the dependent variable;

[0022] According to the standardized independent variables and dependent variables, a linear model and a nonlinear coupling model are constructed respectively;

[0023] The goodness of fit and the residual sum of squares of the linear model and the nonlinear model are calculated respectively;

[0024] The difference between the goodness of fit of the nonlinear coupling model and the goodness of fit of the linear model is calculated to obtain the goodness of fit improvement of the nonlinear coupling model;

[0025] The deviation between the residual sum of squares of the linear model and the residual sum of squares of the nonlinear coupling model is calculated to obtain the residual sum of squares reduction of the nonlinear coupling model;

[0026] The statistical significance of the coupling coefficient is judged by t-test, wherein the coupling coefficient is the coefficient of the interaction term of two variables in the nonlinear coupling model;

[0027] If the goodness of fit improvement of the nonlinear coupling model and the residual sum of squares reduction both meet the requirements, and the coupling coefficient is statistically significant, then the variable pair has nonlinear coupling.

[0028] Further, the manner of judging whether the disaster risk level jump phenomenon is caused by nonlinear coupling of variables is:

[0029] The coupling increment of each nonlinear variable pair is calculated based on the nonlinear coupling model;

[0030] The coupling increment contribution ratio is obtained by proportionally calculating the coupling increment and the nonlinear coupling effect;

[0031] For all variables of the nonlinear variable pairs, the change amount of the standardized value between the current evaluation and the previous evaluation is calculated: variable change amount = current variable standardized value - previous variable standardized value;

[0032] If the change amount of no single variable is greater than or equal to the preset change amount, then the risk level jump caused by the dramatic change of a single variable is excluded;

[0033] For all identified nonlinear variable pairs, the coupling increment change amount between adjacent two risk jump evaluations is calculated;

[0034] If the coupling increment change amount is greater than or equal to the threshold value, and the coupling increment contribution ratio is greater than or equal to the preset ratio, then the risk level jump phenomenon is caused by nonlinear coupling of variables, and the risk level jump evaluation is marked as nonlinear jump evaluation.

[0035] Further, the calculation manner of the coupling increment is:

[0036] Based on the nonlinear coupling model, the risk evaluation level when one variable is 0 and the other variable is 1 is calculated respectively, and the linear superposition expectation is obtained by adding them;

[0037] Set both variables to 1, substitute them into the nonlinear coupling model, and obtain the risk assessment level under the nonlinear coupling effect;

[0038] The coupling increment is obtained by subtracting the risk assessment level under nonlinear coupling from the linear superposition expectation.

[0039] Furthermore, the method for determining whether the current model has a weak ability to identify nonlinear relationships is as follows:

[0040] The proportion of nonlinear jump assessments in historical disaster risk assessments is statistically analyzed to obtain the jump ratio.

[0041] Calculate the average risk level jump range of all jump disaster risk assessments, and then normalize it to obtain the average risk level jump rate;

[0042] The non-linear identification index is obtained by adding the percentage of jumps to the average risk level jump rate.

[0043] If the nonlinear recognition index is greater than the preset recognition index, then the current model has a weak ability to recognize nonlinear relationships.

[0044] Furthermore, the construction method of the coupled-superimposed dual-module model is as follows:

[0045] The basic risk output of the model is obtained by linearly superimposing the standardized values ​​of each variable with the corresponding linear risk coefficients.

[0046] The linear risk coefficient is derived by standardizing the regression coefficient obtained through fitting, with the historical actual risk assessment level as the dependent variable and the standardized value of the variable as the independent variable.

[0047] For any pair of nonlinear variables;

[0048] The coupling increment of a single pair of nonlinear variables is the product of the coupling coefficient, the coupling state coefficient, and the theoretical maximum coupling increment of the nonlinear variable pair.

[0049] Wherein, the coupling state coefficient S is the product of the standardized values ​​of the two nonlinear variables in the current nonlinear variable pair, and the theoretical maximum coupling increment is the coupling increment output by the nonlinear coupling model when both nonlinear variable pairs are 1.

[0050] The coupling weights of nonlinear variable pairs are calculated by combining coupling strength and the importance of risk dimensions;

[0051] The total coupling increment correction value output by the nonlinear risk model is the sum of the product of the coupling increment and the coupling weight of each nonlinear variable pair;

[0052] By superimposing the basic risk layer and the nonlinear risk layer, a coupled-superimposed dual-module model is obtained, which outputs a comprehensive risk index.

[0053] Furthermore, the coupling weights of the nonlinear variable pairs are calculated as follows:

[0054] The coupling strength weight is the proportion of the coupling coefficient of a nonlinear variable pair in the sum of the coupling coefficients of all nonlinear variable pairs;

[0055] The weights of the nonlinear variables in the risk dimension are calculated as follows: statistically analyze the historical disaster data of the current sandy dolomite tunnel, calculate the loss ratio of different disaster types, and allocate weights according to the loss ratio.

[0056] The final coupling weight of a nonlinear variable pair is the percentage of the sum of the products of coupling strength weight and risk dimension weight for all nonlinear variable pairs.

[0057] Furthermore, the method for determining whether the current disaster risk assessment has a risk level jump risk is as follows:

[0058] List all variable combinations involved in the current disaster risk assessment. Compare with the set of nonlinear variables. If a pair of nonlinear variables exists, calculate the coupling state coefficient of the nonlinear variable pair, which is the product of the standardized values ​​of the two nonlinear variables in the nonlinear variable pair.

[0059] The coupling state coefficient of the nonlinear variable pair is compared with the coordination threshold. If the coupling state coefficient is greater than or equal to the coordination threshold, the current disaster risk assessment has the risk of a risk level jump.

[0060] The coupled-overlay dual-module model is invoked to output a comprehensive risk index.

[0061] A smart risk assessment system for sand-formed dolomite tunnel hazards includes the following modules:

[0062] Jump identification module: By analyzing the risk level jumps in multiple historical disaster risk assessments of sandy dolomite tunnels, the module identifies risk level jump assessments and performs statistical analysis on the risk level jump assessments to determine whether there are risk level jump phenomena when the current model assesses disaster risks.

[0063] Coupling Analysis Module: If it exists, perform nonlinear coupling analysis on the variables of risk level jump assessment, identify nonlinear variable pairs, and determine whether the risk level jump phenomenon is caused by nonlinear coupling of variables. If so, mark the risk level jump assessment as a nonlinear jump assessment and integrate the nonlinear variable pairs into a nonlinear variable set.

[0064] Capability assessment module: By performing statistical analysis on the nonlinear jump assessment and combining it with the degree of risk level jump in the nonlinear jump assessment, it determines whether the current model has a weak ability to identify nonlinear relationships;

[0065] Model building module: If so, construct a coupled-superimposed dual-module model with coupling increment as the link, based on the nonlinear coupling relationship of the nonlinear variable pairs;

[0066] Assessment module: For the current disaster risk assessment, the variables are compared and analyzed with the set of nonlinear variables to determine whether there is a risk level jump. If so, the coupled-superimposed dual-module model is invoked to assess the current disaster risk.

[0067] The beneficial effects of this invention are as follows: By using risk level jump analysis and nonlinear coupling judgment, it is the first time to distinguish between accidental fluctuations in engineering and model defects, solving the problem that existing models cannot explain the causes of risk jumps, thus improving the credibility of the assessment results. It constructs a nonlinear coupling model and a coupling-superposition dual-module model to accurately capture the risk increment of variable synergy, avoiding the underestimation / overestimation of coupling risk by linear models, thereby improving the assessment accuracy. It dynamically calls the model (dual-module model / linear model) according to the current coupling state of variables, adapting to the variability of geological and construction conditions in sandy dolomite tunnels, meeting the real-time control needs of engineering sites. The supporting intelligent assessment system realizes full automation of the data acquisition-analysis-modeling-assessment process, reducing manual intervention and lowering the professional threshold for assessment personnel. At the same time, the storage module can accumulate historical data to provide data support for subsequent model optimization. Attached Figure Description

[0068] The invention will now be further described with reference to the accompanying drawings.

[0069] Figure 1 This is a flowchart illustrating the steps of an intelligent assessment method for disaster risks in sandy dolomite tunnels as described in Embodiment 1 of the present invention.

[0070] Figure 2 This is a logic diagram of an intelligent assessment method for disaster risks in sandy dolomite tunnels, as described in Embodiment 1 of the present invention.

[0071] Figure 3 This is a flowchart of a disaster risk intelligent assessment system for sandy dolomite tunnels, as described in Embodiment 2 of the present invention. Detailed Implementation

[0072] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0073] Example 1: Please refer to Figure 1 As shown in the embodiment of the present invention, a method for intelligent assessment of disaster risk in sandy dolomite tunnels includes the following steps:

[0074] Step 1: By conducting risk level jump analysis on multiple historical disaster risk assessments of the sandy dolomite tunnel, identify risk level jump assessments, and perform statistical analysis on the risk level jump assessments to determine whether there is a risk level jump phenomenon when the current model assesses disaster risk.

[0075] In step one, the identification process for the risk level jump assessment includes:

[0076] Obtain historical disaster risk assessment data for the sandy dolomite tunnel multiple times, including:

[0077] Basic disaster information: time of disaster occurrence, tunnel mileage, disaster type, and changes in actual risk level;

[0078] Time series data of variables: monitoring data of disaster-causing variables from 72 hours before the disaster to the time of the disaster;

[0079] Current model assessment data: the fitting results of variable relationships in the current risk assessment model and the risk level assessment results when historical disasters occurred;

[0080] For any given historical disaster risk assessment:

[0081] The previous historical disaster risk assessment, which is adjacent in time and in the same space as the current historical disaster risk assessment, is selected as the comparison benchmark.

[0082] The risk level jump range is obtained by subtracting the risk level of the previous historical disaster risk assessment from the risk level of the previous historical disaster risk assessment. If the risk level jump range is greater than or equal to 2 levels, the historical disaster risk assessment is marked as a risk level jump assessment.

[0083] In step one, the process of determining whether there is a risk level jump includes:

[0084] The proportion of risk level jump assessments in historical disaster risk assessments is used to obtain the risk level jump percentage.

[0085] Compare the percentage of risk level jumps with the preset percentage. If the percentage of risk level jumps is greater than the preset percentage, then the current model has a risk level jump phenomenon when assessing disaster risk.

[0086] It should be noted that the preset percentage is a critical threshold for distinguishing between normal accidental risk jumps and frequent risk jumps caused by current model defects in a specific risk scenario of sandy dolomite tunnels. In essence, it is the upper limit of the acceptable jump case ratio set by the project based on safety tolerance, geological risk level, and industry experience. If the actual risk level jump ratio exceeds this threshold, it means that the model cannot stably capture the risk evolution law and there are frequent jumps due to insufficient nonlinear identification.

[0087] The preset percentage is set by consulting relevant industry specifications, local standards or group standards for sandy dolomite tunnels and extracting the threshold requirements for risk level fluctuations from the reliability indicators of the risk assessment model.

[0088] It is important to emphasize that the purpose of determining whether there is a risk level jump is to screen out abnormal assessments with leaps in risk level from historical assessment data, and to preliminarily determine whether the anomaly is caused by accidental fluctuations in the project or defects in the current model, thus laying the foundation for further in-depth analysis and solving the problems of low efficiency in indiscriminate analysis of historical data and inability to distinguish between accidental and inevitable anomalies.

[0089] Step 2: If it exists, perform nonlinear coupling analysis on the variables of the risk level jump assessment, identify nonlinear variable pairs, and determine whether the risk level jump phenomenon is caused by nonlinear coupling of variables. If so, mark the risk level jump assessment as a nonlinear jump assessment and integrate the nonlinear variable pairs into a nonlinear variable set.

[0090] Please see Figure 2 As shown, in step two, the identification process of the nonlinear variable pair includes:

[0091] The variables used in the risk level jump assessment refer to the variables relied upon when assessing the hazard risk of sandy dolomite tunnels, including but not limited to:

[0092] Geological disaster-causing variables: sandification grade, groundwater pressure, geostress state, groundwater flow velocity, permeability coefficient, etc.

[0093] Construction-related disaster variables include: support lag time, blasting vibration intensity, excavation advance, and grouting quality parameters.

[0094] Load-bearing variables: tunnel cross-sectional dimensions, support structure parameters, lining structure integrity, construction equipment and personnel density, etc.

[0095] The variable pairs for sandy dolomite tunnels include geological-geological variable pairs, geological-construction variable pairs, and construction-stage variable pairs. Nonlinear coupling analysis is performed on any one of these variable pairs, specifically as follows:

[0096] For any pair of variables X and Y, we take the two variables X and Y as independent variables and the risk assessment level Z as the dependent variable, and standardize the independent and dependent variables.

[0097] Based on the standardized independent variables X, Y, and dependent variable Z, linear and nonlinear coupled models are constructed respectively, as follows:

[0098] For any pair of variables:

[0099] The first point to clarify is that the process of constructing a linear model is as follows:

[0100] The linear model is ,in, For variable coefficients, This is the error term;

[0101] The coefficients of the variables are obtained by solving the least squares method. Finally, a linear model of the variable pairs is obtained;

[0102] Secondly, it should be noted that the construction process of the nonlinear coupling model is as follows:

[0103] The nonlinear coupling model is ,in, For variable coefficients, The coupling coefficient is... For variable interaction terms, This is the error term;

[0104] The coefficients of the variables are obtained by solving using the least squares method. and coupling coefficient This ultimately yields a nonlinear coupling model of the variable pairs;

[0105] By comparing the model fitting results, we can determine whether there is nonlinear coupling between variable pairs. Specifically:

[0106] Calculate the goodness of fit and sum of squared residuals for the linear and nonlinear models respectively;

[0107] The difference between the goodness of fit of the nonlinear coupling model and the goodness of fit of the linear model is calculated to obtain the improvement in the goodness of fit of the nonlinear coupling model.

[0108] The deviation between the sum of squared residuals of the linear model and the sum of squared residuals of the nonlinear coupled model is calculated to obtain the reduction in the sum of squared residuals of the nonlinear coupled model.

[0109] Determine the coupling coefficient using a t-test. The significance is as follows:

[0110] Calculate the coupling coefficient of the nonlinear coupling model The t-statistic is used to find the coupling coefficient from the t-distribution table. The p-value is used to determine statistical significance; if p < 0.05, then the statistical significance is considered.

[0111] If the goodness-of-fit improvement of the nonlinear coupling model is greater than 0 and its absolute value is greater than or equal to the threshold, the decrease in the sum of squared residuals is greater than 0 and its absolute value is greater than or equal to the threshold, and the coupling coefficient... If the statistical significance is significant, then there is non-linear coupling between the variable pairs;

[0112] In step two, the process of determining whether the risk level jump phenomenon is caused by nonlinear coupling of variables includes:

[0113] For any risk jump assessment:

[0114] Based on the nonlinear coupling model, the risk contributions of X-only action, Y-only action, and X+Y coupling action are calculated respectively, and the coupling increment is quantified. The specific calculation process of the coupling increment is as follows:

[0115] X function only: Fix Y=0, substitute X=1, and calculate the risk assessment level. ;

[0116] Y only: Fix X=0, substitute Y=1, and calculate the risk assessment level. ;

[0117] Linear superposition expectation: ;

[0118] Nonlinear coupling effect: Substitute X=1, Y=1, calculate ;

[0119] Therefore, the coupling increment is: That is, the coupling coefficient directly corresponds to the coupling increment;

[0120] It is understandable that the physical meaning of coupling increment is: coupling increment refers to the additional risk that nonlinear variable pairs, through synergistic effect, exceed the risk of linear superposition of single variables;

[0121] The contribution ratio of the coupling increment is obtained by proportionally calculating the coupling increment to the nonlinear coupling effect.

[0122] For all nonlinear variable pairs, calculate the change in standardized value between the current assessment and the previous assessment: Change in variable = Current standardized value - Previous standardized value;

[0123] If no single variable changes more than or equal to the preset change, then a sudden change in a single variable that could cause a jump in risk level is excluded.

[0124] For all identified nonlinear variable pairs, calculate the change in coupling increment between two adjacent risk jump assessments. If the change in coupling increment is greater than or equal to the threshold and the contribution ratio of the coupling increment is greater than or equal to the preset ratio, then the risk level jump phenomenon is caused by nonlinear coupling of variables, and the risk level jump assessment is marked as a nonlinear jump assessment.

[0125] Integrate all pairs of nonlinear variables to generate a set of nonlinear variables;

[0126] It is important to emphasize that the role of nonlinear coupling analysis and the construction of nonlinear variable sets is to: clarify the root cause of risk level jumps (whether they are caused by nonlinear coupling of variables) and extract the nonlinear variable pairs that trigger the jumps, providing targets for subsequent model optimization;

[0127] Step 3: By conducting statistical analysis on the nonlinear jump assessment and combining it with the degree of risk level jump in the nonlinear jump assessment, determine whether the current model has a weak ability to identify nonlinear relationships;

[0128] In step three, the process of determining whether the current model has a weak ability to identify nonlinear relationships includes:

[0129] The proportion of nonlinear jump assessments in historical disaster risk assessments is statistically analyzed to obtain the jump ratio.

[0130] Calculate the average risk level jump range of all jump disaster risk assessments, and obtain the average risk level jump rate after normalization.

[0131] The non-linear identification index is obtained by adding the percentage of jumps to the average risk level jump rate.

[0132] The nonlinear identification index is compared with the preset identification index. If the nonlinear identification index is greater than the preset identification index, the current model has a weak ability to identify nonlinear relationships.

[0133] It should be noted that the preset recognition index is the maximum acceptable value of the non-linear recognition index determined for a specific engineering scenario, and is set by reference to industry standards or similar engineering data.

[0134] It is understandable that the physical meaning of the nonlinear identification index is: a comprehensive quantitative index of the number of historical disaster risk assessments and the degree of risk level jumps caused by the coupling of nonlinear variable pairs in the current model. The magnitude of the value directly corresponds to the weakness of the current model's nonlinear identification capability.

[0135] It is important to emphasize that the role of judging the model's nonlinearity recognition capability is to: quantify the degree of deficiency of the current model in nonlinear relationship recognition, give a clear conclusion on whether the model needs to be optimized, and solve the problem of whether the model's defects are serious enough to require adjustment. It is a key decision-making link connecting problem diagnosis and solution.

[0136] Step 4: If so, based on the nonlinear coupling relationship of the nonlinear variable pairs, construct a coupled-superimposed dual-module model with coupling increments as the link;

[0137] In step four, the construction process of the coupled-superimposed dual-module model with coupling increment as the link includes:

[0138] Based on the coupling increments of each nonlinear variable pair, a coupled-superposition dual-module model is constructed with the coupling increments as the link. The model adopts a dual-module serial structure. The first module is the basic risk layer, which calculates the linear superposition risk. The second module is the nonlinear risk layer, which introduces coupling increment correction.

[0139] The first point to clarify is that the construction process of the basic risk layer is as follows:

[0140] The cumulative risk assessment level for each nonlinear variable pair, considering the individual effects of all variables, is calculated and used as the model's base risk output. Specifically:

[0141] The basic risk output of the model is obtained by linearly superimposing the standardized values ​​of each variable with their corresponding linear risk coefficients, as shown in the formula: ,in, For linearly superimposed risks, n is the total number of variables. Let i be the standardized value of the i-th variable. Let be the linear risk coefficient of the i-th variable;

[0142] The linear risk coefficient needs to be obtained through fitting, specifically:

[0143] Multiple linear regression was used to fit the nonlinear jump assessment data, with the historical actual risk assessment level as the dependent variable and the standardized value of the variable as the independent variable. The linear risk coefficient was obtained by standardizing the output regression coefficient.

[0144] Secondly, it should be noted that the construction process of the nonlinear risk layer includes:

[0145] The nonlinear risk layer is the core of the model. By introducing a weighted sum of coupled increments, it corrects the linear superposition risk and obtains the final comprehensive risk index. The core logic is: the higher the coupling strength, i.e., the larger the coupling coefficient, for a pair of nonlinear variables, the greater the weight of its coupling increment in correcting the risk. Specifically:

[0146] For any pair of nonlinear variables

[0147] Calculate the coupling increment of a single pair of nonlinear variables: ,in, Let be the coupling coefficient of the nonlinear variable pair, and S be the coupling state coefficient of the nonlinear variable pair. This represents the theoretical maximum coupling increment.

[0148] Wherein, the coupling state coefficient S is the product of the standardized values ​​of the two nonlinear variables in the current nonlinear variable pair, which describes the actual level of cooperation of the nonlinear variable pair;

[0149] Theoretical maximum coupling increment The coupling increment is the nonlinear variable pair in a high-high cooperative state, i.e., when X=1 and Y=1.

[0150] It is understandable that the physical meaning of the formula for calculating the coupling increment of a single pair of nonlinear variables is: it describes the actual coordination level of the nonlinear variable pair at the current engineering moment. The larger the coupling state coefficient, the higher the coupling effect of the nonlinear variable and the greater the actual contribution of the coupling increment. The smaller the coupling state coefficient, the weaker the coupling effect of the nonlinear variable and the smaller the contribution of the coupling increment.

[0151] It should be noted that the physical meaning of the formula for calculating the coupling increment of a single pair of nonlinear variables is: it is essentially the product of strength × state × limit increment, which accurately quantifies the actual coupling risk contribution under the current engineering conditions. This avoids both overestimating the risk of low-cooperation state with a fixed coupling coefficient and underestimating the risk of high-cooperation state with the limit increment, and ultimately achieves dynamic and accurate quantification of nonlinear coupling effects.

[0152] The coupling weights of each nonlinear variable pair are calculated by combining the coupling strength and the importance of the risk dimension, specifically as follows:

[0153] For any pair of nonlinear variables:

[0154] Coupling strength weight Let be the proportion of the coupling coefficient of a nonlinear variable pair in the sum of the coupling coefficients of all nonlinear variable pairs, where The coupling strength weight for the i-th pair of nonlinear variables;

[0155] Risk dimension weight Here, represents the weights of nonlinear variables with respect to their corresponding disaster types. For the i-th pair of nonlinear variables, the risk dimension weight is denoted as .

[0156] It should be noted that the calculation process for the risk dimension weights is as follows: statistically analyze the historical disaster data of the current sandy dolomite tunnel, calculate the loss ratio of different disaster types, and allocate weights according to the loss ratio;

[0157] The final coupling weights of the nonlinear variable pairs are Where m is the total number of nonlinear variable pairs, Let be the coupling weight of the i-th pair of nonlinear variables;

[0158] Therefore, the total coupling increment correction value output by the nonlinear risk model is Where m is the total number of nonlinear variable pairs, Let be the coupling increment of the i-th pair of nonlinear variables. Let be the coupling weight of the i-th pair of nonlinear variables;

[0159] By superimposing the basic risk layer and the nonlinear risk layer, a coupled-superimposed dual-module model is obtained, which outputs the comprehensive risk index RI, i.e. Where n is the total number of variables. Let i be the standardized value of the i-th variable. Let be the linear risk coefficient of the i-th variable;

[0160] It is important to emphasize that the role of constructing the coupled-superimposed dual-module model is to build a coupled-superimposed dual-module model that can accurately capture the coupling effect to replace the original defective model, which is the solution implementation stage in the process;

[0161] Step 5: For the current disaster risk assessment, compare and analyze the variables with the nonlinear variable set to determine whether there is a risk level jump risk. If so, call the coupled-superimposed dual-module model to assess the current disaster risk.

[0162] In step five, the process of calling the corresponding disaster risk assessment model based on the relevant variables of disaster risk assessment includes:

[0163] List all variable combinations involved in the current disaster risk assessment. Compare with the set of nonlinear variables. If a pair of nonlinear variables exists, calculate the coupling state coefficient of the nonlinear variable pair, which is the product of the standardized values ​​of the two nonlinear variables in the nonlinear variable pair.

[0164] The coupling state coefficient of the nonlinear variable pair is compared with the coordination threshold. If the coupling state coefficient is greater than or equal to the coordination threshold, the current disaster risk assessment has the risk of a risk level jump.

[0165] Call the corresponding coupled-superimposed dual-module model with the current nonlinear variable pair, and output the comprehensive risk index RI;

[0166] Conversely, if there are no nonlinear variable pairs or the coupling state coefficients of nonlinear variable pairs are greater than or equal to the cooperative threshold, then a single linear model is invoked.

[0167] It should be noted that the synergy threshold is the critical value of the coupling state coefficient, used to determine whether the synergy of the nonlinear variable pair has reached the strength that may trigger a jump in risk level. The synergy threshold is determined by finding the boundary interval between the two by statistically analyzing the distribution of the coupling state coefficients of the nonlinear jump assessment and the normal assessment.

[0168] Input the relevant variables of the current disaster risk assessment into the disaster risk assessment model, and output the current disaster risk level;

[0169] It is important to emphasize that the current role of disaster risk assessment and model invocation is to dynamically select a model based on the characteristics of current variables and output an accurate risk level. It is the final application stage of the entire process and addresses the need for real-time risk assessment at engineering sites.

[0170] The technical solution and advantages of this application embodiment are as follows: By performing risk level jump analysis on multiple historical disaster risk assessments of sandy dolomite tunnels, risk level jump assessments are identified, and statistical analysis is performed on these risk level jump assessments to determine whether there is a risk level jump phenomenon when the current model assesses disaster risk. If so, nonlinear coupling analysis is performed on the variables of the risk level jump assessment to identify nonlinear variable pairs and determine whether the risk level jump in the risk jump assessment is caused by nonlinear coupling of variables. If so, the risk jump assessment is marked as a nonlinear jump assessment, and the nonlinear variable pairs are integrated into a nonlinear variable set. By performing statistical analysis on the nonlinear jump assessment and combining it with the degree of risk level jump in the nonlinear jump assessment, it is determined whether the current model has a weak ability to identify nonlinear relationships. If so, based on the nonlinear coupling relationship of the nonlinear variable pairs, a coupled-superimposed dual-module model with coupling increment as the link is constructed. For the current disaster risk assessment, the variables and the nonlinear variable set are compared and analyzed to determine whether there is a risk level jump risk. If so, the coupled-superimposed dual-module model is called to assess the current disaster risk. This application uses historical disaster risk assessment data to perform risk level jump analysis, identifies risk level jump assessments, and statistically determines whether the current model exhibits jump phenomena. If so, it performs nonlinear coupling analysis on the variables to determine whether the risk jump is caused by nonlinear coupling of variables and constructs a set of nonlinear variables. It also uses a nonlinear identification index to determine the current model's ability to identify nonlinear relationships. If the identification ability is weak, it constructs a coupled-superimposed dual-module model with coupling increments as the link. Finally, based on the comparison results between the current variables and the set of nonlinear variables, it dynamically calls the corresponding model to complete real-time risk assessment, solving the problems of existing models ignoring variable nonlinear coupling, frequent risk level jumps, and low assessment accuracy.

[0171] Example 2: Please refer to Figure 3 As shown in the embodiment of the present invention, an intelligent risk assessment system for sand-formed dolomite tunnel disasters includes the following modules:

[0172] Jump identification module: By analyzing the risk level jumps in multiple historical disaster risk assessments of sandy dolomite tunnels, the module identifies risk level jump assessments and performs statistical analysis on the risk level jump assessments to determine whether there are risk level jump phenomena when the current model assesses disaster risks.

[0173] The identification process for the risk level jump assessment includes:

[0174] Obtain historical disaster risk assessment data for the sandy dolomite tunnel multiple times, including:

[0175] Basic disaster information: time of disaster occurrence, tunnel mileage, disaster type, and changes in actual risk level;

[0176] Time series data of variables: monitoring data of disaster-causing variables from 72 hours before the disaster to the time of the disaster;

[0177] Current model assessment data: the fitting results of variable relationships in the current risk assessment model and the risk level assessment results when historical disasters occurred;

[0178] For any given historical disaster risk assessment:

[0179] The previous historical disaster risk assessment, which is adjacent in time and in the same space as the current historical disaster risk assessment, is selected as the comparison benchmark.

[0180] The risk level jump range is obtained by subtracting the risk level of the previous historical disaster risk assessment from the risk level of the previous historical disaster risk assessment. If the risk level jump range is greater than or equal to 2 levels, the historical disaster risk assessment is marked as a risk level jump assessment.

[0181] The process for determining whether there is a risk level jump includes:

[0182] The proportion of risk level jump assessments in historical disaster risk assessments is used to obtain the risk level jump percentage.

[0183] Compare the percentage of risk level jumps with the preset percentage. If the percentage of risk level jumps is greater than the preset percentage, then the current model has a risk level jump phenomenon when assessing disaster risk.

[0184] It should be noted that the preset percentage is a critical threshold for distinguishing between normal accidental risk jumps and frequent risk jumps caused by current model defects in a specific risk scenario of sandy dolomite tunnels. In essence, it is the upper limit of the acceptable jump case ratio set by the project based on safety tolerance, geological risk level, and industry experience. If the actual risk level jump ratio exceeds this threshold, it means that the model cannot stably capture the risk evolution law and there are frequent jumps due to insufficient nonlinear identification.

[0185] The preset percentage is set by consulting relevant industry specifications, local standards or group standards for sandy dolomite tunnels and extracting the threshold requirements for risk level fluctuations from the reliability indicators of the risk assessment model.

[0186] Coupling Analysis Module: If it exists, perform nonlinear coupling analysis on the variables of risk level jump assessment, identify nonlinear variable pairs, and determine whether the risk level jump phenomenon is caused by nonlinear coupling of variables. If so, mark the risk level jump assessment as a nonlinear jump assessment and integrate the nonlinear variable pairs into a nonlinear variable set.

[0187] Please see Figure 2 As shown, the identification process of the nonlinear variable pairs includes:

[0188] The variables used in the risk level jump assessment refer to the variables relied upon when assessing the hazard risk of sandy dolomite tunnels, including but not limited to:

[0189] Geological disaster-causing variables: sandification grade, groundwater pressure, geostress state, groundwater flow velocity, permeability coefficient, etc.

[0190] Construction-related disaster variables include: support lag time, blasting vibration intensity, excavation advance, and grouting quality parameters.

[0191] Load-bearing variables: tunnel cross-sectional dimensions, support structure parameters, lining structure integrity, construction equipment and personnel density, etc.

[0192] The variable pairs for sandy dolomite tunnels include geological-geological variable pairs, geological-construction variable pairs, and construction-stage variable pairs. Nonlinear coupling analysis is performed on any one of these variable pairs, specifically as follows:

[0193] For any pair of variables X and Y, we take the two variables X and Y as independent variables and the risk assessment level Z as the dependent variable, and standardize the independent and dependent variables.

[0194] Based on the standardized independent variables X, Y, and dependent variable Z, linear and nonlinear coupled models are constructed respectively, as follows:

[0195] For any pair of variables:

[0196] The first point to clarify is that the process of constructing a linear model is as follows:

[0197] The linear model is ,in, For variable coefficients, This is the error term;

[0198] The coefficients of the variables are obtained by solving using the least squares method. Finally, a linear model of the variable pairs is obtained;

[0199] Secondly, it should be noted that the construction process of the nonlinear coupling model is as follows:

[0200] The nonlinear coupling model is ,in, For variable coefficients, The coupling coefficient is... For variable interaction terms, This is the error term;

[0201] The coefficients of the variables are obtained by solving using the least squares method. and coupling coefficient This ultimately yields a nonlinear coupling model of the variable pairs;

[0202] By comparing the model fitting results, we can determine whether there is nonlinear coupling between variable pairs. Specifically:

[0203] Calculate the goodness of fit and sum of squared residuals for the linear and nonlinear models respectively;

[0204] The difference between the goodness of fit of the nonlinear coupling model and the goodness of fit of the linear model is calculated to obtain the improvement in the goodness of fit of the nonlinear coupling model.

[0205] The deviation between the sum of squared residuals of the linear model and the sum of squared residuals of the nonlinear coupled model is calculated to obtain the reduction in the sum of squared residuals of the nonlinear coupled model.

[0206] Determine the coupling coefficient using a t-test. The significance is as follows:

[0207] Calculate the coupling coefficient of the nonlinear coupling model The t-statistic is used to find the coupling coefficient from the t-distribution table. The p-value is used to determine statistical significance; if p < 0.05, then the statistical significance is considered.

[0208] If the goodness-of-fit improvement of the nonlinear coupling model is greater than 0 and its absolute value is greater than or equal to the threshold, the decrease in the sum of squared residuals is greater than 0 and its absolute value is greater than or equal to the threshold, and the coupling coefficient... If the statistical significance is significant, then there is non-linear coupling between the variable pairs;

[0209] The process for determining whether the risk level jump phenomenon is caused by nonlinear coupling of variables includes:

[0210] For any risk jump assessment:

[0211] Based on the nonlinear coupling model, the risk contributions of X-only action, Y-only action, and X+Y coupling action are calculated respectively, and the coupling increment is quantified. The specific calculation process of the coupling increment is as follows:

[0212] X function only: Fix Y=0, substitute X=1, and calculate the risk assessment level. ;

[0213] Y only: Fix X=0, substitute Y=1, and calculate the risk assessment level. ;

[0214] Linear superposition expectation: ;

[0215] Nonlinear coupling effect: Substitute X=1, Y=1, calculate ;

[0216] Therefore, the coupling increment is: That is, the coupling coefficient directly corresponds to the coupling increment;

[0217] It is understandable that the physical meaning of coupling increment is: coupling increment refers to the additional risk that nonlinear variable pairs, through synergistic effect, exceed the risk of linear superposition of single variables;

[0218] The contribution ratio of the coupling increment is obtained by proportionally calculating the coupling increment to the nonlinear coupling effect.

[0219] For all nonlinear variable pairs, calculate the change in standardized value between the current assessment and the previous assessment: Change in variable = Current standardized value - Previous standardized value;

[0220] If no single variable changes more than or equal to the preset change, then a sudden change in a single variable that could cause a jump in risk level is excluded.

[0221] For all identified nonlinear variable pairs, calculate the change in coupling increment between two adjacent risk jump assessments. If the change in coupling increment is greater than or equal to the threshold and the contribution ratio of the coupling increment is greater than or equal to the preset ratio, then the risk level jump phenomenon is caused by nonlinear coupling of variables, and the risk jump assessment is marked as a nonlinear jump assessment.

[0222] Integrate all pairs of nonlinear variables to generate a set of nonlinear variables;

[0223] Capability assessment module: By performing statistical analysis on the nonlinear jump assessment and combining it with the degree of risk level jump in the nonlinear jump assessment, it determines whether the current model has a weak ability to identify nonlinear relationships;

[0224] The process of determining whether the current model has a weak ability to identify nonlinear relationships includes:

[0225] The proportion of nonlinear jump assessments in historical disaster risk assessments is statistically analyzed to obtain the jump ratio.

[0226] Calculate the average risk level jump range of all jump disaster risk assessments, and obtain the average risk level jump rate after normalization.

[0227] The non-linear identification index is obtained by adding the percentage of jumps to the average risk level jump rate.

[0228] The nonlinear identification index is compared with the preset identification index. If the nonlinear identification index is greater than the preset identification index, the current model has a weak ability to identify nonlinear relationships.

[0229] It should be noted that the preset recognition index is the maximum acceptable value of the non-linear recognition index determined for a specific engineering scenario, and is set by reference to industry standards or similar engineering data.

[0230] It is understandable that the physical meaning of the nonlinear identification index is: a comprehensive quantitative index of the number of historical disaster risk assessments and the degree of risk level jumps caused by the coupling of nonlinear variable pairs in the current model. The magnitude of the value directly corresponds to the weakness of the current model's nonlinear identification capability.

[0231] Model building module: If so, construct a coupled-superimposed dual-module model with coupling increment as the link, based on the nonlinear coupling relationship of the nonlinear variable pairs;

[0232] The construction process of the coupled-superimposed dual-module model with coupling increment as the link includes:

[0233] Based on the coupling increments of each nonlinear variable pair, a coupled-superposition dual-module model is constructed with the coupling increments as the link. The model adopts a dual-module serial structure. The first module is the basic risk layer, which calculates the linear superposition risk. The second module is the nonlinear risk layer, which introduces coupling increment correction.

[0234] The first point to clarify is that the construction process of the basic risk layer is as follows:

[0235] The cumulative risk assessment level for each nonlinear variable pair, considering the individual effects of all variables, is calculated and used as the model's base risk output. Specifically:

[0236] The basic risk output of the model is obtained by linearly superimposing the standardized values ​​of each variable with their corresponding linear risk coefficients, as shown in the formula: ,in, For linearly superimposed risks, n is the total number of variables. Let i be the standardized value of the i-th variable. Let be the linear risk coefficient of the i-th variable;

[0237] The linear risk coefficient needs to be obtained through fitting, specifically:

[0238] Multiple linear regression was used to fit the nonlinear jump assessment data, with the historical actual risk assessment level as the dependent variable and the standardized value of the variable as the independent variable. The linear risk coefficient was obtained by standardizing the output regression coefficient.

[0239] Secondly, it should be noted that the construction process of the nonlinear risk layer includes:

[0240] The nonlinear risk layer is the core of the model. By introducing a weighted sum of coupled increments, it corrects the linear superposition risk and obtains the final comprehensive risk index. The core logic is: the higher the coupling strength, i.e., the larger the coupling coefficient, for a pair of nonlinear variables, the greater the weight of its coupling increment in correcting the risk. Specifically:

[0241] For any pair of nonlinear variables

[0242] Calculate the coupling increment of a single pair of nonlinear variables: ,in, Let be the coupling coefficient of the nonlinear variable pair, and S be the coupling state coefficient of the nonlinear variable pair. This represents the theoretical maximum coupling increment.

[0243] Wherein, the coupling state coefficient S is the product of the standardized values ​​of the two nonlinear variables in the current nonlinear variable pair, which describes the actual level of cooperation of the nonlinear variable pair;

[0244] Theoretical maximum coupling increment The coupling increment is the nonlinear variable pair in a high-high cooperative state, i.e., when X=1 and Y=1.

[0245] It is understandable that the physical meaning of the formula for calculating the coupling increment of a single pair of nonlinear variables is: it describes the actual coordination level of the nonlinear variable pair at the current engineering moment. The larger the coupling state coefficient, the higher the coupling effect of the nonlinear variable and the greater the actual contribution of the coupling increment. The smaller the coupling state coefficient, the weaker the coupling effect of the nonlinear variable and the smaller the contribution of the coupling increment.

[0246] It should be noted that the physical meaning of the formula for calculating the coupling increment of a single pair of nonlinear variables is: it is essentially the product of strength × state × limit increment, which accurately quantifies the actual coupling risk contribution under the current engineering conditions. This avoids both overestimating the risk of low-cooperation state with a fixed coupling coefficient and underestimating the risk of high-cooperation state with the limit increment, and ultimately achieves dynamic and accurate quantification of nonlinear coupling effects.

[0247] The coupling weights of each nonlinear variable pair are calculated by combining the coupling strength and the importance of the risk dimension, specifically as follows:

[0248] For any pair of nonlinear variables:

[0249] Coupling strength weight Let be the proportion of the coupling coefficient of a nonlinear variable pair in the sum of the coupling coefficients of all nonlinear variable pairs, where The coupling strength weight for the i-th pair of nonlinear variables;

[0250] Risk dimension weight Here, represents the weights of nonlinear variables with respect to their corresponding disaster types. For the i-th pair of nonlinear variables, the risk dimension weight is denoted as .

[0251] It should be noted that the calculation process for the risk dimension weights is as follows: statistically analyze the historical disaster data of the current sandy dolomite tunnel, calculate the loss ratio of different disaster types, and allocate weights according to the loss ratio;

[0252] The final coupling weights of the nonlinear variable pairs are Where m is the total number of nonlinear variable pairs, Let be the coupling weight of the i-th pair of nonlinear variables;

[0253] Therefore, the total coupling increment correction value output by the nonlinear risk model is Where m is the total number of nonlinear variable pairs, Let be the coupling increment of the i-th pair of nonlinear variables. Let be the coupling weight of the i-th pair of nonlinear variables;

[0254] By superimposing the basic risk layer and the nonlinear risk layer, a coupled-superimposed dual-module model is obtained, which outputs the comprehensive risk index RI, i.e. Where n is the total number of variables. Let i be the standardized value of the i-th variable. The linear risk coefficient of the i-th variable

[0255] Assessment module: For the current disaster risk assessment, the variables are compared and analyzed with the nonlinear variable set to determine whether there is a risk level jump risk. If so, the coupled-superimposed dual-module model is invoked to assess the current disaster risk.

[0256] The process of calling the corresponding disaster risk assessment model based on disaster risk assessment-related variables includes:

[0257] List all variable combinations involved in the current disaster risk assessment. Compare with the set of nonlinear variables. If a pair of nonlinear variables exists, calculate the coupling state coefficient of the nonlinear variable pair, which is the product of the standardized values ​​of the two nonlinear variables in the nonlinear variable pair.

[0258] The coupling state coefficient of the nonlinear variable pair is compared with the coordination threshold. If the coupling state coefficient is greater than or equal to the coordination threshold, the current disaster risk assessment has the risk of a risk level jump.

[0259] Call the corresponding coupled-superimposed dual-module model with the current nonlinear variable pair, and output the comprehensive risk index RI;

[0260] Conversely, if there are no nonlinear variable pairs or the coupling state coefficients of nonlinear variable pairs are greater than or equal to the cooperative threshold, then a single linear model is invoked.

[0261] It should be noted that the synergy threshold is the critical value of the coupling state coefficient, used to determine whether the synergy of the nonlinear variable pair has reached the strength that may trigger a jump in risk level. The synergy threshold is determined by finding the boundary interval between the two by statistically analyzing the distribution of the coupling state coefficients of the nonlinear jump assessment and the normal assessment.

[0262] Input the relevant variables of the current disaster risk assessment into the disaster risk assessment model, and output the current disaster risk level.

[0263] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent assessment of hazard risk in sandy dolomite tunnels, characterized in that: Comprise: Step one: through the risk grade jump analysis of the risk assessment of the sanding dolomite tunnel history multiple disasters, identify the risk grade jump assessment, and statistically analyze the risk grade jump assessment, judge whether there is a risk grade jump phenomenon when the current model assesses the disaster risk; Step two: if so, the nonlinear coupling analysis of the variable of the risk grade jump assessment is carried out, the nonlinear variable pair is identified, and it is judged whether the risk grade jump phenomenon is caused by the nonlinear coupling of the variable, if so, the risk grade jump assessment is marked as nonlinear jump assessment, and the nonlinear variable pair is integrated into a nonlinear variable set; Step three: through statistical analysis of the nonlinear jump assessment, and combining the risk grade jump degree of the nonlinear jump assessment, judge whether the current model is weak in identifying nonlinear relationship; Step four: if so, according to the nonlinear coupling relationship of the nonlinear variable pair, a coupling-addition double module model is constructed with coupling increment as the link; The calculation method of the coupling increment is: Based on the nonlinear coupling model, the risk assessment level when one variable is 0 and the other variable is 1 is calculated respectively, and the linear addition expectation is obtained by adding them up; Let both variables be 1, substitute into the nonlinear coupling model to get the risk assessment level under the nonlinear coupling effect; The risk assessment level under the nonlinear coupling effect is subtracted from the linear addition expectation to get the coupling increment; The construction method of the coupling-addition double module model is: The standardized value of each variable is linearly added to the corresponding linear risk coefficient to obtain the basic risk output of the model; Wherein, the linear risk coefficient is the regression coefficient standardized by fitting the historical actual risk assessment level as the dependent variable and the standardized value of the variable as the independent variable; For any one nonlinear variable pair; The coupling increment of a single nonlinear variable pair is the product of the coupling coefficient, the coupling state coefficient and the theoretical maximum coupling increment of the nonlinear variable pair; Wherein, the coupling state coefficient S is the product of the standardized values of the two nonlinear variables in the current nonlinear variable pair, and the theoretical maximum coupling increment is the coupling increment output by the nonlinear coupling model when the nonlinear variable pair is 1; Step five: for the current disaster risk assessment, compare and analyze the variables and the nonlinear variable set to determine whether there is a risk grade jump risk, if so, call the coupling-addition double module model to assess the current disaster risk.

2. The sanding dolomite tunnel disaster risk intelligent assessment method according to claim 1, characterized in that: The identification method of the risk grade jump assessment is: For any one historical disaster risk assessment: The risk grade jump amplitude is obtained by subtracting the risk grade of the previous historical disaster risk assessment from the risk grade of the historical disaster risk assessment, if the risk grade jump amplitude is greater than or equal to 2 levels, the historical disaster risk assessment is marked as risk grade jump assessment; The proportion of risk grade jump in historical disaster risk assessment is obtained by statistical analysis of the risk grade jump assessment; Compare the risk grade jump proportion with the preset proportion, if the risk grade jump proportion is greater than the preset proportion, there is a risk grade jump phenomenon when the current model assesses the disaster risk. 3.The method according to claim 1, wherein: the identification method of the nonlinear variable pair is: for any variable pair, taking the two variables as independent variables and the risk assessment level as dependent variable, and standardizing the independent variables and the dependent variable; constructing a linear model and a nonlinear coupling model according to the standardized independent variables and dependent variables; calculating the goodness of fit and the residual sum of squares of the linear model and the nonlinear model; calculating the difference between the goodness of fit of the nonlinear coupling model and the goodness of fit of the linear model to obtain the goodness of fit improvement of the nonlinear coupling model; calculating the deviation between the residual sum of squares of the linear model and the residual sum of squares of the nonlinear coupling model to obtain the residual sum of squares reduction of the nonlinear coupling model; and judging the statistical significance of the coupling coefficient by t-test, wherein the coupling coefficient is the coefficient of the interaction term of the two variables in the nonlinear coupling model; and if the goodness of fit improvement of the nonlinear coupling model and the residual sum of squares reduction both meet the requirements and the coupling coefficient is statistically significant, the variable pair has nonlinear coupling. 4.The method according to claim 3, wherein: the judgment method of whether the risk level jump phenomenon is caused by variable nonlinear coupling is: calculating the coupling increment of each nonlinear variable pair based on the nonlinear coupling model; calculating the contribution ratio of the coupling increment by proportioning the coupling increment and the nonlinear coupling effect; calculating the change amount of the standardized value of the variable between the current assessment and the previous assessment: variable change amount = current variable standardized value - previous variable standardized value; if no single variable change amount is greater than or equal to the preset change amount, the risk level jump caused by the dramatic change of the single variable is excluded; calculating the coupling increment change amount of the adjacent two risk jump assessments for all identified nonlinear variable pairs; if the coupling increment change amount is greater than or equal to the threshold value and the contribution ratio of the coupling increment is greater than or equal to the preset ratio, the risk level jump phenomenon is caused by variable nonlinear coupling, and the risk level jump assessment is marked as nonlinear jump assessment. 5.The method according to claim 1, wherein: the judgment method of whether the current model has weak nonlinear relationship identification ability is: calculating the proportion of nonlinear jump assessment in the historical disaster risk assessment to obtain the jump proportion; calculating the average risk level jump amplitude of all jump disaster risk assessments and obtaining the average risk level jump rate after normalization; adding the jump proportion and the average risk level jump rate to obtain the nonlinear identification index; and if the nonlinear identification index is greater than the preset identification index, the current model has weak nonlinear relationship identification ability. 6.The method according to claim 1, wherein: the calculation method of the coupling weight of the nonlinear variable pair is: the coupling strength weight is the proportion of the coupling coefficient of the nonlinear variable pair in the total coupling coefficient of all nonlinear variable pairs. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ The weight of the risk dimension weight nonlinear variable pair corresponding to the disaster type is calculated in the following manner: historical disaster data of the current dolomitic sandstone tunnel is counted, the loss proportion of different disaster types is calculated, and the weight is allocated according to the loss proportion; The coupling weight of the final nonlinear variable pair is the product of the coupling strength weight and the risk dimension weight, and the proportion of the product in the total sum of the coupling strength weight and the risk dimension weight of all nonlinear variable pairs.

7. The dolomitic sandstone tunnel disaster risk intelligent assessment method according to claim 1, characterized in that: The judgment method of whether the current disaster risk assessment exists risk grade jump risk is: List all variable combinations involved in the current disaster risk assessment, and compare with the nonlinear variable set. If there is a nonlinear variable pair, calculate the coupling state coefficient of the nonlinear variable pair, that is, the product of the standardized values of the two nonlinear variables in the nonlinear variable pair; Compare the coupling state coefficient of the nonlinear variable pair with the synergy threshold. If the coupling state coefficient is greater than or equal to the synergy threshold, the current disaster risk assessment exists risk grade jump risk; Call the coupling-superposition double module model to output the comprehensive risk index.

8. A dolomitic sandstone tunnel disaster risk intelligent assessment system, comprising the following modules: Jump identification module: analyze the risk grade jump of the historical multiple disaster risk assessments of the dolomitic sandstone tunnel, identify the risk grade jump assessment, and statistically analyze the risk grade jump assessment to determine whether the current model exists risk grade jump phenomenon when assessing the disaster risk; Coupling analysis module: if so, perform nonlinear coupling analysis on the variables of the risk grade jump assessment, identify the nonlinear variable pair, and determine whether the risk grade jump phenomenon is caused by the nonlinear coupling of the variables. If so, mark the risk grade jump assessment as nonlinear jump assessment, and integrate the nonlinear variable pair into the nonlinear variable set; Capability judgment module: statistically analyze the nonlinear jump assessment, and determine whether the current model has weak nonlinear relationship identification capability by combining the risk grade jump degree of the nonlinear jump assessment; Model construction module: if so, construct a coupling-superposition double module model based on the nonlinear coupling relationship of the nonlinear variable pair, with coupling increment as the link; The calculation method of the coupling increment is: Based on the nonlinear coupling model, calculate the risk assessment level when one variable is 0 and the other variable is 1 respectively, and add them to get the linear superposition expectation; Let both variables be 1, and substitute them into the nonlinear coupling model to get the risk assessment level under the nonlinear coupling effect; The coupling increment is obtained by subtracting the risk assessment level under the nonlinear coupling effect from the linear superposition expectation; The construction method of the coupling-superposition double module model is: Linearly superimpose the standardized values of each variable and the corresponding linear risk coefficient to obtain the basic risk output of the model; Wherein, the linear risk coefficient is the regression coefficient standardized by fitting the historical actual risk assessment level as the dependent variable and the standardized value of the variable as the independent variable; For any nonlinear variable pair; The coupling increment of the single set of nonlinear variable pairs is the product of the coupling coefficient, the coupling state coefficient and the theoretical maximum coupling increment of the nonlinear variable pairs; Wherein, the coupling state coefficient S is the product of the normalized values of the two nonlinear variables in the current nonlinear variable pair, and the theoretical maximum coupling increment is the coupling increment of the nonlinear coupling model output when the nonlinear variable pairs are all 1; The evaluation calling module: for the current disaster risk assessment, the variable and the nonlinear variable set are compared and analyzed to determine whether there is a risk level jump risk, and if there is, the coupling-superposition double module model is called to assess the current disaster risk.

Images