Intelligent identification method for tunnel advanced geological disasters based on multi-source information fusion

Abstract

The method comprises the following steps: S1, a combination of advanced geological prediction means comprises multiple prediction methods; a tunnel disaster data set is set, prediction data corresponding to a disaster occurrence position of a target tunnel in each prediction method in the combination of advanced geological prediction means is collected, the collected prediction data is put into the tunnel disaster data set, and a mapping relationship between various geological disasters and construction parameters of the target tunnel is established; S2, a processing system respectively trains a corresponding convolutional neural network model by using each type of prediction data in the tunnel disaster data set, and obtains a convolutional neural network model group; based on multiple single identification results, a multi-expert analysis and decision-making process is simulated, and the purpose of intelligently identifying geological disasters such as karst, fault and gushing water by means of multi-source information fusion is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tunnel geology, in particular to a tunnel advanced geological disaster intelligent identification method based on multi-source information fusion. BACKGROUND

[0002] China is a mountainous country with very complex geological conditions, and mountains and plateaus account for about 60% of the total area of the country. Highways and railways often pass through mountains and hills, and the proportion of tunnels in road engineering is increasing. In the process of tunnel construction, various geological problems induced by excavation are often the main factors that restrict tunnel construction due to their non-selectivity, complexity, particularity and suddenness. How to accurately predict whether there will be water-rich zones, fault fracture zones and karst cave rivers and other adverse geologies in the process of tunnel construction has become a key problem that needs to be solved in the process of tunnel construction, which will play an important guiding role in reducing the blindness in the process of tunnel construction and achieving safe and rapid construction.

[0003] In recent years, artificial intelligence has developed rapidly under the support of deep learning, big data and large-scale parallel computing, especially in image recognition inversion, among which convolutional neural networks have shown high precision modeling capabilities for strong nonlinear mapping. Using convolutional neural networks can effectively reduce the subjectivity of human identification of geological disasters and achieve intelligent identification. At the same time, tunnel advanced geological prediction has developed a variety of geophysical exploration methods such as seismic wave method, direct current method, electromagnetic method and induced polarization method. However, each method is relatively independent and relies too much on professional forecasters, so it is urgent to establish a scientific and intelligent comprehensive analysis system for tunnel advanced geological disaster prediction. SUMMARY

[0004] The present application aims to simulate multi-expert analysis and decision-making process based on multiple single identification results to achieve multi-source information fusion intelligent identification of karst, faults, gushing water and other geological disasters.

[0005] Therefore, a tunnel advanced geological disaster intelligent identification method based on multi-source information fusion is proposed, and the specific technical solutions are as follows:

[0006] The tunnel advanced geological disaster intelligent identification method based on multi-source information fusion is characterized in that:

[0007] S1: The advanced geological prediction means combination includes multiple prediction methods;

[0008] A tunnel disaster data set is provided, which collects prediction data corresponding to the target tunnel disaster site from each prediction method in the advanced geological prediction means combination, and puts the collected prediction data into the tunnel disaster data set to establish a mapping relationship between various geological disasters and various parameters of the target tunnel;

[0009] S2: The processing system respectively trains a corresponding convolutional neural network model for each type of prediction data in the tunnel disaster data set, obtaining a group of convolutional neural network models;

[0010] S3: A set of advanced geological prediction results is set, and the processing system respectively obtains corresponding target section data through each prediction method in the advanced geological prediction means combination, and the processing system inputs the predicted target section data into the corresponding convolutional neural network model, and puts the prediction result into the set of advanced geological prediction results;

[0011] S4: The processing system compares the prediction data in the set of advanced geological prediction results with the field situation:

[0012] If the prediction data in the set of advanced geological prediction results is consistent, go to S5;

[0013] If the prediction data in the set of advanced geological prediction results is inconsistent, go to S6;

[0014] S5: The processing system compares the prediction data with the field situation, if the prediction data in the set of advanced geological prediction results is consistent with the field situation, go to S5-1, if the prediction data in the set of advanced geological prediction results is consistent but inconsistent with the field situation, go to S5-2;

[0015] S5-1: The prediction data is output data;

[0016] S5-2: Return to S2, adjust the convolutional neural network model parameters, and re-optimize the convolutional neural network model;

[0017] S6: If the prediction data in the set of advanced geological prediction results is inconsistent, evaluate all advanced geological predictions by AHP-fuzzy comprehensive evaluation method to obtain optimal prediction data;

[0018] S7: The processing system compares the optimal prediction data with the field actual situation, if the optimal prediction data is consistent with the field actual situation, go to S8;

[0019] Otherwise, the processing system iteratively optimizes the parameters in the AHP-fuzzy comprehensive evaluation method;

[0020] S8: The processing system outputs the optimal prediction data.

[0021] To better realize the present application, further:

[0022] The AHP-fuzzy comprehensive evaluation method in S6 includes the following steps:

[0023] S6-1: Establish a hierarchical structure, and divide the intelligent geological prediction comprehensive evaluation influencing factor system into a target layer, a criterion layer and an index layer;

[0024] The target layer A is a comprehensive evaluation influence factor system of intelligent geological prediction;

[0025] The criterion layer B has five indexes, respectively, the applicability of the method itself B1, the surrounding rock geological condition B2, the prediction original data quality B3, the geophysical implementation unit B4, and the tunnel construction influence B5.

[0026] The index layer C has 25 indexes, respectively:

[0027] The applicability of the method itself corresponds to three secondary indexes of karst sensitivity, gushing water sensitivity, and fault sensitivity.

[0028] The surrounding rock geological condition corresponds to three secondary indexes of lithology, rock layer integrity, and rock strength.

[0029] The prediction original data quality corresponds to 14 secondary indexes of detection frequency, monitoring surface roughness, environmental conductivity, tunnel structure influence, prediction distance, disaster size and shape, same disaster repetition, blast hole layout quality, excitation energy size, mechanical noise interference, loop size and position, measurement voltage, and grounding condition.

[0030] The geophysical implementation unit corresponds to three secondary indexes of geophysical equipment instrument precision, construction personnel technical quality, supervision and attention degree.

[0031] The tunnel construction influence corresponds to two secondary indexes of construction fund and tunnel site influence degree.

[0032] S6-2: Structure judgment matrix, through the way of questionnaire survey, experts judge the importance of each layer index relative to the upper layer index by pairwise comparison, and fill in the index quantified as numerical value through the set scale rule to obtain the judgment matrix of experts on each evaluation index, as shown below:

[0033]

[0034] S6-3: Judgment matrix weight calculation

[0035] The characteristic vector of the expert judgment matrix is calculated respectively, and the weight is calculated, as follows:

[0036] S6-31: Judgment matrix normalization

[0037]

[0038] S6-32: Add the normalized matrix by row

[0039]

[0040] S6-33: Calculate the weight vector

[0041] The normalized vector is the approximate solution of the weight vector.

[0042]

[0043] The vector w = (w1, w2, …, wn) is the solution vector. n

[0044] S6-4: Consistency check and error analysis

[0045] The consistency check is used to determine whether the error of the weight approximate solution is within the allowable range.

[0046] S6-41: Calculate the maximum eigenvalue of the matrix

[0047] Let the maximum eigenvalue of the judgment matrix be λ max , then:

[0048]

[0049] where (Cw) i represents the product of the ith element of the judgment matrix C and its weight, and n represents the order of the matrix.

[0050] S6-42: Consistency index

[0051]

[0052] The average random consistency index is obtained by consulting the standard table.

[0053]

[0054] When CR≤0.1, it is considered that the judgment matrix has satisfactory consistency; when CR>0.1, the judgment matrix must be restructured.

[0055] S6-5: Hierarchical combination weight

[0056] Through the calculation of S6-4, the weight of each index relative to the previous level can be obtained, and through hierarchical combination, the weight of each index relative to the target layer can be obtained.

[0057] Let the first index layer be A, and the next level B have B1, B2, …, B m , and the index weight vector of B is w1 = (b1, b2, …, bn). m The next level of B is C layer, which has C1, C2, …, C n specific indexes, and the weight vector of C is w2 = (c1, c2, …, cn). n The weight of the ith element of C layer relative to the target layer is:

[0058] ​

[0059] wherein a k is the weight of the secondary index B j to which the primary index A k belongs; m is the weight of the tertiary index C i to which the secondary index B m belongs;

[0060] S6-6: Constructing a fuzzy evaluation matrix

[0061] For the five geophysical prospecting methods of ground penetrating radar, induced polarization method, transient electromagnetic method, TSP and HSP, the connection degree of each index is set to five levels, which are: very high, high, general, low and very low;

[0062] Corresponding to 5 scores: 5, 4, 3, 2, 1; 4.5, 3.5, 2.5, 1.5 represent the intermediate state;

[0063] In order to make the evaluation result more objective and accurate, based on the expert evaluation of existing literature, each evaluation index of the five geophysical prospecting methods is objectively scored;

[0064] Secondly, according to the scoring results of each index, a fuzzy evaluation matrix R is constructed;

[0065] S6-7: Calculate the comprehensive evaluation vector

[0066] According to the weight W of each index multiplied by the fuzzy evaluation matrix R, the membership degree vector D of each geophysical prospecting method is obtained, and the optimal geophysical prospecting result is selected as the optimal geophysical prospecting result according to the maximum membership degree principle.

[0067] Further:

[0068] The advanced geological prediction method includes ground penetrating radar method, transient electromagnetic method, induced polarization method, TSP and HSP prediction method.

[0069] Further:

[0070] The target tunnel construction parameters include collecting geophysical prospecting, pre-feasibility geological exploration, auxiliary tunnel, while-drilling parameters and physical parameters.

[0071] The beneficial effects of the present application are: the present application can intelligently identify the geological disaster position, scale and type monitored by various prediction methods such as ground penetrating radar method, transient electromagnetic method, induced polarization method, TSP and HSP, based on multiple single identification results, simulate multi-expert analysis and decision-making process, and achieve the purpose of multi-source information fusion intelligent identification of karst, fault and gushing water and other geological disasters. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 Workflow diagram of the present application;

[0073] Figure 2 HSP geophysical result image label form example diagram;

[0074] Figure 3 Intelligent geological prediction comprehensive evaluation influence factor system diagram. DETAILED DESCRIPTION

[0075] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0076] As Figure 1 shown:

[0077] Tunnel advanced geological disaster intelligent identification method based on multi-source information fusion,

[0078] S1: The advanced geological prediction means combination includes multiple prediction methods.

[0079] A tunnel disaster data set is provided, which collects prediction data corresponding to the disaster occurrence position of the target tunnel for each prediction method in the advanced geological prediction means combination, and puts the collected prediction data into the tunnel disaster data set to establish a mapping relationship between various geological disasters and various parameters of the target tunnel.

[0080] Specifically, the geological prediction picture data obtained by various prediction means is sorted. In this embodiment, HSP is taken as an example. Referring to Figure 2 shown, other geophysical prospecting means are similar. And corresponding labels are marked, mainly including disaster location, scale and type.

[0081] Collect multi-aspect data such as geophysical prospecting, previous geological exploration, auxiliary tunnel, while-drilling parameters, physical property parameters and mapping relationship with various geological disasters, and determine the intelligent geological prediction comprehensive evaluation influence factor system taking the method itself applicability, surrounding rock geological conditions, prediction original data quality, geophysical prospecting implementation unit and tunnel construction influence as main conditions, to provide data support for subsequent AHP-fuzzy comprehensive evaluation method.

[0082] S2: The processing system respectively trains corresponding convolutional neural network models using each type of prediction data in the tunnel disaster data set to obtain a convolutional neural network model group.

[0083] S3: A set of advanced geological prediction results is set, and the processing system obtains corresponding target section data through each prediction method in the advanced geological prediction means combination. The processing system inputs the predicted target section data into the corresponding convolutional neural network model, and places the prediction results into the set of advanced geological prediction results.

[0084] S4: The processing system compares the prediction data in the set of advanced geological prediction results with the field situation.

[0085] If the prediction data in the set of advanced geological prediction results is consistent, S5 is entered.

[0086] If the prediction data in the set of advanced geological prediction results is inconsistent, S6 is entered.

[0087] S5: The processing system compares the prediction data with the field situation. If the prediction data in the set of advanced geological prediction results is consistent with the field situation, S5-1 is entered. If the prediction data in the set of advanced geological prediction results is consistent but inconsistent with the field situation, S5-2 is entered.

[0088] S5-1: The prediction data is output data.

[0089] S5-2: Return to S2, adjust the convolutional neural network model parameters, and re-optimize the convolutional neural network model.

[0090] S6: If the prediction data in the set of advanced geological prediction results is inconsistent, all advanced geological predictions are evaluated by AHP-fuzzy comprehensive evaluation method to obtain optimal prediction data.

[0091] Specifically, five methods of geological radar method, transient electromagnetic method, induced polarization method, TSP, and HSP prediction method are taken as examples.

[0092] The prediction data generated by the above five prediction methods is placed into the set of advanced geological prediction results after being processed by the corresponding convolutional neural network model. In this embodiment, the geological radar method, the induced polarization method, the transient electromagnetic method, the TSP, and the HSP detection result input convolutional neural network recognition result are inconsistent, which are sudden gushing water, sudden gushing water, sudden gushing water, karst, and karst, respectively. Therefore, AHP-fuzzy comprehensive evaluation method is needed to evaluate all advanced geological predictions to obtain optimal prediction data.

[0093] S7: The processing system compares the optimal prediction data with the actual field situation. If the optimal prediction data is consistent with the actual field situation, S8 is entered.

[0094] Otherwise, the processing system iteratively optimizes the parameters in the AHP-fuzzy comprehensive evaluation method.

[0095] S8: The processing system outputs the optimal prediction data.

[0096] The AHP-fuzzy comprehensive evaluation method includes the following steps:

[0097] S6-1: Establish a hierarchy, and divide the intelligent geological prediction comprehensive evaluation influencing factor system into a target layer, a criterion layer, and an index layer;

[0098] The target layer A is the intelligent geological prediction comprehensive evaluation influencing factor system;

[0099] The criterion layer has five indexes, which are method itself applicability B1, surrounding rock geological condition B2, prediction original data quality B3, geophysical implementation unit B4, and tunnel construction influence B5;

[0100] The index layer has 25 indexes, which are:

[0101] The method itself applicability corresponds to karst sensitivity, gushing water sensitivity, and fault sensitivity;

[0102] The surrounding rock geological condition corresponds to lithology, rock layer integrity, and rock strength;

[0103] The prediction original data quality corresponds to detection frequency, monitoring surface roughness, environmental conductivity, tunnel structure influence, prediction distance, disaster scale and shape, same disaster repeated occurrence, blast hole layout quality, excitation energy size, mechanical noise interference, loop size and position, measurement voltage, and grounding condition;

[0104] The geophysical implementation unit corresponds to geophysical equipment instrument precision, construction personnel technical quality, supervision and attention degree;

[0105] The tunnel construction influence corresponds to construction fund and tunnel site influence degree;

[0106] In this embodiment, the intelligent geological prediction comprehensive evaluation influencing factor system is divided into a target layer, a criterion layer, and an index layer.

[0107] The target layer A is the intelligent geological prediction comprehensive evaluation influencing factor system;

[0108] The criterion layer B has five indexes, which are

[0109] B = (B1, B2, B3, B4, B5)

[0110] = (method itself applicability, surrounding rock geological condition, prediction original data quality, geophysical implementation unit, and tunnel construction influence),

[0111] The index layer C has 25 indexes, which are:

[0112] C1 = (C 11 , C12 , C 13 ) = (karst sensitivity, fault sensitivity, gushing water sensitivity)

[0113] C2 = (C 21 , C 22 , C 23 ) = (lithology, rock bed integrity, rock strength)

[0114] C3 = (C 31 , C 32 , C 33 , C 34 , C 35 , C 36 , C 37 , C 38 , C 39 , C 310 , C 311 , C 312 , C 313 , C 314 )

[0115] = (probe frequency, monitoring surface roughness, environmental conductivity, tunnel structure influence,

[0116] forecast distance, disaster size and shape, same disaster repetition, blast hole layout quality excitation energy size, mechanical noise interference, loop size and position, measurement voltage, grounding conditions) C4 = (C 41 , C 42 , C 43 ) = (geophysical equipment instrument precision, construction personnel technical quality, supervision and attention) 51 , C 52 ) = (construction funds, tunnel site impact)

[0117] S6-2: Then, the judgment matrix is constructed. Through questionnaire survey, experts judge the importance of each layer index relative to the upper layer index by pairwise comparison, and quantize it into numerical value by using the scale rule shown in Table 1 to fill in, and get the judgment matrix of each expert on each evaluation index, as shown below:

[0118]

[0119] Table 1 Judgment Matrix C ij Element The specific meaning of

[0120]

[0121]

[0122] The present example is specifically, through the consultation of geological prediction related literature and expert consultation, invite experts in the relevant field to questionnaire form to score the levels of the constructed intelligent geological prediction comprehensive evaluation influence factor system, constitute the judgment matrix A, B1, B2, B3, B4, B5 as shown in the following table.

[0123] Table 2A-B judgment matrix

[0124]

[0125] Table 3B1-C judgment matrix

[0126]

[0127] Table 4B2-C judgment matrix

[0128]

[0129] Table 5B3-C judgment matrix

[0130]

[0131]

[0132] Table 6 B4-C judgment matrix

[0133]

[0134] Table 7 B5-C judgment matrix

[0135]

[0136] S6-3: Judgment matrix weight calculation:

[0137] The sum product method is used to calculate the weight of the analytic hierarchy process, taking the A-B target layer judgment matrix as an example. The subjective weight of the B layer criterion layer is calculated. The A-B judgment matrix C can be obtained from the expert scoring results A-B

[0138]

[0139] S6-31: Normalization of each column of the judgment matrix

[0140]

[0141] Referring to the above formula, each column of the judgment matrix is normalized to obtain

[0142]

[0143] S6-32: Add the normalized matrix by row

[0144]

[0145]

[0146]

[0147]

[0148]

[0149]

[0150] S6-33: Calculate weight vector

[0151] The normalized vector is the approximate solution of the weight vector.

[0152]

[0153]

[0154]

[0155]

[0156]

[0157]

[0158] The vector w = (0.503, 0.068, 0.134, 0.260, 0.035) is the solution vector.

[0159] S6-4: Consistency check and error analysis

[0160] The consistency check is used to determine whether the error of the weight approximate solution is within the allowed range.

[0161] S6-41: Calculate the maximum eigenvalue λ of the judgment matrix max

[0162]

[0163]

[0164] S6-42: Consistency index

[0165]

[0166] The average random consistency index is obtained by referring to Table 8 standard table;

[0167]

[0168] Table 8 Random consistency standard index RI value table

[0169]

[0170] By querying the random consistency index RI value table, RI = 1.12 can be obtained, so the consistency ratio CR can be obtained:

[0171]

[0172] Since the consistency ratio of the A-B hierarchical judgment matrix is 0.054 < 0.1, the judgment matrix has a satisfactory consistency and passes the consistency test.

[0173] According to the above S6-3 and S6-4, the calculation results of other judgment matrices are as follows:

[0174] Table 9 A-B judgment matrix calculation result table

[0175]

[0176] The consistency test of all judgment matrices shows that all matrices pass the consistency test.

[0177] According to the results obtained by the above calculation steps, the hierarchical single ordering of the intelligent geological prediction comprehensive evaluation influence factor system can be carried out, and the subjective weight ordering of each level index is obtained. The subjective weight value and the ordering results are shown in the following table:

[0178] Table 10 Weight value ordering table

[0179]

[0180] S6-5: Hierarchical combination weight

[0181] Through the above calculation, the weight of each index relative to the previous level can be obtained, and through hierarchical combination, the weight of each index relative to the target layer can be obtained;

[0182] Suppose the first level index layer is A, and the next level B has B1, B2, …, B m , the index weight vector of B is w1 = (b1, b2, …, b m ), and the next level of B layer is C layer, which has C1, C2, …, C n specific indexes, and the weight vector of C is w2 = (c1, c2, …, c n ), then the weight of the i-th element of C layer relative to the target layer is:

[0183]

[0184] wherein a k is the weight of the secondary index B j to which the primary index A k belongs, b m is the weight of the tertiary index C i to which the secondary index B m belongs.

[0185] The hierarchical combination weight W is as follows:

[0186] Table 11 Hierarchical combination weight table

[0187]

[0188]

[0189] That is, the hierarchical combination weight W is:

[0190]

[0191] S6-6: Constructing a fuzzy evaluation matrix

[0192] For the five geophysical methods of ground penetrating radar, induced polarization method, transient electromagnetic method, TSP and HSP, the connection degree of each index is set to five levels: very high, high, general, low, and very low, corresponding to five scores: 5, 4, 3, 2, and 1; 4.5, 3.5, 2.5, and 1.5 represent intermediate states. Based on existing literature and expert experience, each evaluation index of the five geophysical methods is objectively scored as shown in Table 12:

[0193] Table 12 Secondary index score table

[0194]

[0195]

[0196] According to the scoring results of each index, a single-factor fuzzy evaluation matrix R is constructed.

[0197]

[0198] S6-7: Calculate the comprehensive evaluation vector

[0199] The fuzzy comprehensive evaluation result vector D is obtained by multiplying the index weight W obtained from the fuzzy evaluation matrix R. According to the maximum membership degree principle, the result measured by the optimal geophysical method is selected as the optimal geophysical result.

[0200] Through AHP-fuzzy comprehensive evaluation, the fuzzy comprehensive evaluation result vector

[0201]

[0202] D = (19.770, 19.840, 19.314, 19.139, 19.139)

[0203] The membership relations of the corresponding determined decision set (geological radar method, induced polarization method, transient electromagnetic method, TSP, HSP) are 19.770, 19.840, 19.314, 19.139 and 19.139 in turn, wherein the induced polarization method occupies the largest proportion, so that the geological prediction disaster comprehensive evaluation result is "sudden gushing water", which is consistent with the actual situation, and meets the characteristics that the induced polarization method is sensitive to water-rich poor geology.

[0204] It is apparent to those skilled in the art that the present application is not limited to the details of the foregoing exemplary embodiments, and that the present application can be implemented in other particular forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered in all respects as illustrative and not restrictive, and the scope of the present application should be defined by the appended claims rather than the above description, and it is intended to include all changes falling within the meaning and range of equivalents of the claims. Any reference signs in the claims should not be construed as limiting the claims to which they belong.

[0205] Furthermore, it should be understood that, although the present specification is described in terms of embodiments, not every embodiment contains only one independent technical solution, and the description of the specification is only for the sake of clarity, and those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be combined appropriately to form other embodiments that those skilled in the art can understand.

Claims

1. A method for intelligent identification of advanced geological hazards in tunnels based on multi-source information fusion, characterized in that: S1: Advanced geological prediction methods include a variety of prediction approaches; A tunnel disaster data set is set up to collect prediction data for the location of the disaster in the target tunnel for each prediction method in the combination of advanced geological prediction methods. The collected prediction data is put into the tunnel disaster data set to establish a mapping relationship between various geological disasters and various parameters of the target tunnel. S2: The processing system uses each type of prediction data in the tunnel disaster dataset to train and generate a corresponding convolutional neural network model, resulting in a group of convolutional neural network models; S3: An advanced geological prediction result set is set up. The processing system obtains the corresponding target section data for each prediction method in the combination of advanced geological prediction methods. The processing system inputs the predicted target section data into the corresponding convolutional neural network model and puts the prediction results into the advanced geological prediction result set. S4: The processing system compares the forecast data in the advanced geological forecast result set with the field conditions: If the prediction data within the advanced geological prediction dataset are consistent, proceed to S5; If the prediction data in the advanced geological prediction dataset are inconsistent, proceed to S6; S5: The processing system compares the predicted data with the field conditions. If the predicted data in the advanced geological prediction dataset are consistent with the field conditions, then proceed to S5-1. If the advanced geological prediction data are consistent but inconsistent with the field conditions, then proceed to S5-2. S5-1: The predicted data is the output data; S5-2: Return to S2, adjust the parameters of the convolutional neural network model, and re-optimize the convolutional neural network model; S6: If the prediction data in the advanced geological prediction dataset are inconsistent, all advanced geological predictions are evaluated using the AHP-fuzzy comprehensive evaluation method to obtain the optimal prediction data; S7: The processing system compares the optimal prediction data with the actual situation on site. If the optimal prediction data is consistent with the actual situation on site, then proceed to S8. Otherwise, the processing system iteratively optimizes the intrinsic parameters of the AHP-fuzzy comprehensive evaluation method; S8: Process the system output of the optimal prediction data.

2. The intelligent identification method for tunnel geological hazards based on multi-source information fusion according to claim 1, characterized in that: The AHP-fuzzy comprehensive evaluation method in S6 includes the following steps: S6-1: Establish a hierarchical structure, dividing the comprehensive evaluation influencing factors system for intelligent geological forecasting into target layer, criterion layer, and indicator layer; Target layer A represents the comprehensive evaluation system of influencing factors for intelligent geological forecasting; The criteria layer B consists of five indicators: the applicability of the method itself (B1), the geological conditions of the surrounding rock (B2), the quality of the original forecast data (B3), the geophysical exploration implementation unit (B4), and the impact of tunnel construction (B5). Indicator layer C contains 25 indicators, namely: The applicability of the method itself corresponds to three secondary indicators: karst sensitivity, sudden water inrush sensitivity, and fault sensitivity. The geological conditions of the surrounding rock correspond to three secondary indicators: lithology, rock strata integrity, and rock strength. The quality of the raw forecast data corresponds to 14 secondary indicators, including detection frequency, surface roughness, environmental conductivity, impact of tunnel structures, forecast distance, disaster size and form, recurrence of the same disaster, borehole layout quality, excitation energy, mechanical noise interference, loop size and location, measurement voltage, and grounding conditions. The geophysical exploration implementation unit corresponds to three secondary indicators: the accuracy of geophysical equipment and instruments, the technical quality of construction personnel, and the degree of supervision and attention. The impact of tunnel construction corresponds to two secondary indicators: construction costs and the degree of impact on the tunnel site. S6-2: Construct a judgment matrix. Through a questionnaire survey, experts compare each indicator pairwise to determine the importance of each level of indicators relative to the previous level. Using predefined scaling rules, the indicators are quantified into numerical values, resulting in the expert judgment matrix for each evaluation indicator, as shown below: S6-3: Calculation of Matrix Weights Calculate the eigenvectors of the expert judgment matrix and their weights, as follows; S6-31: Judgment Matrix Normalization S6-32: Add the normalized matrix row by row. S6-33: Calculate the weight vector Normalizing the vector yields an approximate solution for the weight vector; Vector w = (w1, w2, ..., w n That is, the vector we are looking for; S6-4: Consistency Judgment and Error Analysis Consistency testing is used to determine whether the error of the weighted approximation solution is within the allowable range; S6-41: Calculate the largest eigenvalue of the matrix Let the largest eigenvalue of the judgment matrix be λ. max ,but: Among them (Cw) i This represents the product of the i-th element of the judgment matrix C and its weight, where n refers to the matrix order; S6-42: Consistency Indicators The average random consistency index is obtained by looking up a standard table; When CR≤0.1, the judgment matrix is ​​considered to have satisfactory consistency; when CR>0.1, the judgment matrix must be reconstructed. S6-5: Hierarchical Combination Weights The weight of each indicator relative to the previous level can be obtained through the calculation of S6-4, and the weight of each indicator relative to the target level can be obtained through the combination of levels. Let A be the first-level indicator layer, and let B be the next level layer B, which includes B1, B2, ..., B1. m The index weight vector of B is w1 = (b1, b2, ..., b...). m The layer below layer B is layer C, which has layers C1, C2, ..., C6. n For each specific indicator, the weight vector of C is w2 = (c1, c2, ..., c...). n If the weight of the i-th element in layer C relative to the target layer is: Among them, a k Secondary indicator B j Belongs to Level A Indicator k The weight, b m Level C i Subordinate secondary indicator B m The weights; S6-6: Constructing the fuzzy evaluation matrix For five geophysical exploration methods—ground-penetrating radar, induced polarization method, transient electromagnetic method, TSP, and HSP—the correlation between each indicator and the method is set into five levels: very high, relatively high, average, relatively low, and very low. These correspond to five scores: 5, 4, 3, 2, and 1; 4.5, 3.5, 2.5, and 1.5 represent intermediate scores. To make the evaluation results more objective and accurate, based on expert evaluations of existing literature, objective scores were given to each evaluation index of the five geophysical exploration methods. Secondly, based on the scoring results of each indicator, a fuzzy evaluation matrix R is constructed; S6-7: Calculate the comprehensive evaluation vector The membership vector D of each geophysical method is obtained by multiplying the weights W of each index by the fuzzy evaluation matrix R. The result obtained by the optimal geophysical method is selected as the optimal geophysical result based on the principle of maximizing the membership.

3. The intelligent identification method for tunnel geological hazards based on multi-source information fusion according to claim 1, characterized in that: The advanced geological prediction methods include ground-penetrating radar, transient electromagnetic method, induced polarization method, TSP, and HSP prediction methods.

4. The intelligent identification method for tunnel geological hazards based on multi-source information fusion according to claim 1, characterized in that: The target tunnel construction parameters include geophysical data collection, preliminary geological exploration, auxiliary tunnels, drilling parameters, and physical property parameters.

Images