Tunnel advanced geological prediction method based on analytic hierarchy process

Abstract

The present application relates to the field of geological prediction technology, and discloses a method for advanced geological prediction of tunnels based on the analytic hierarchy process, which aims to solve the problems of low accuracy and efficiency of existing advanced geological prediction schemes for tunnels. The scheme mainly includes: selecting a preferred geophysical method from a variety of geophysical methods based on the analytic hierarchy process; performing geological prediction based on a first advanced geological prediction model constructed by the preferred geophysical method, and if there are unfavorable geological conditions, selecting a secondary geophysical method from each geophysical method based on the analytic hierarchy process according to the type of the unfavorable geological conditions, and performing geological prediction based on a second advanced geological prediction model constructed by the secondary geophysical method; judging whether there are two geological prediction results that both have unfavorable geological conditions and are of the same type, and if so, taking them as the final geological prediction results. The present application improves the accuracy and efficiency of advanced geological prediction, and is suitable for tunnel engineering.

Description

Technical Field

[0001] The present application relates to the technical field of geological prediction, and in particular to a method for advanced tunnel geological prediction based on the analytic hierarchy process. Background Art

[0002] Tunnel engineering is an important part of water conservancy and hydropower engineering, such as water diversion tunnels and underground powerhouses. During the construction process, the poor geological conditions ahead seriously affect the construction progress and safety. Therefore, tunnel advanced geological prediction is introduced into tunnel engineering as an important part. The main advanced geological prediction methods include geological analysis and geophysical exploration. Geological analysis includes geological survey, drilling, etc. Geophysical exploration includes seismic reflection wave method, land sonar method, TRT (True Reflection Tomography, true seismic reflection imaging method), geological radar method, transient electromagnetic method, etc.

[0003] At present, the advanced geological prediction technology of tunnels has made great progress in both theoretical research and engineering practice, and has covered most of the adverse geological conditions. In order to improve the accuracy of advanced geological prediction in the prior art, the solution usually adopted is to improve the accuracy of the prediction model itself or the obtained on-site construction data. For example, application publication number CN113779690A discloses a method and system for judging the accuracy of the advanced geological prediction system, by reviewing and evaluating the geological objects of the engineering excavation section, obtaining the real geological conditions of the entire section; comparing the actual geological conditions of the entire section with the prediction conclusions, and obtaining the basic prediction comprehensive score and prediction accuracy of each object under the prediction system. For another example, application publication number CN113189672A discloses a method for advanced geological prediction of tunnels based on multi-attribute inversion, which obtains the reflection wave data of the prediction area in front of the tunnel face through geophysical methods, and obtains various physical parameters of the surrounding rock, so as to solve the problem of inaccurate geological interpretation in the geophysical method prediction in the prior art.

[0004] Since geophysical data often have multiple solutions in the interpretation process, when encountering adverse geological conditions, a single prediction method may differ greatly from engineering data or other special circumstances. Therefore, different types of geophysical methods may need to be used to verify each other to improve the accuracy of advanced geological prediction. And according to current engineering practice applications, the establishment of a comprehensive interpretation system in which various geophysical methods verify each other can feedback as much geological information as possible and is relatively accurate. However, the advanced geological prediction technology combining multiple geophysical methods usually adopts the solution of pre-constructing prediction models corresponding to multiple geophysical methods, and using each prediction model for geological prediction. This solution is heavy in workload, and will increase the workload when the geological conditions are good, wasting manpower and material resources. In addition, the geophysical interpretation process is complicated, time-consuming and labor-intensive, and easily affected by subjective factors. Summary of the invention

[0005] The present application aims to solve the problems of low accuracy and efficiency of existing tunnel advance geological prediction schemes, and proposes a tunnel advance geological prediction method based on hierarchical analysis method.

[0006] The technical solution adopted by this application to solve the above technical problems is:

[0007] A tunnel advanced geological prediction method based on the hierarchical analysis method, the method comprising:

[0008] S1. Determine multiple geophysical exploration methods and select the preferred geophysical exploration method from each geophysical exploration method based on the hierarchical analysis method;

[0009] S2, constructing a first advanced geological prediction model based on a deep learning network according to the preferred geophysical prospecting method, and performing geological prediction according to the first advanced geological prediction model to obtain a first geological prediction result. If there are unfavorable geological conditions in the first geological prediction result, proceeding to S3;

[0010] S3. According to the type of adverse geological conditions in the first geological prediction result, a secondary geophysical method is selected from various geophysical methods based on the hierarchical analysis method, a second advanced geological prediction model based on a deep learning network is constructed according to the secondary geophysical method, and geological prediction is performed according to the second advanced geological prediction model to obtain a second geological prediction result;

[0011] S4. Determine whether there are two geological prediction results with the same type and both of which have adverse geological conditions. If so, take the two identical geological prediction results as the final geological prediction results. Otherwise, re-enter S3.

[0012] Furthermore, the preferred geophysical exploration method is selected from various geophysical exploration methods based on the hierarchical analysis method, including:

[0013] Determine the factors that affect the selection of each geophysical exploration method, construct a target hierarchy model based on each factor, and set importance values, construct a judgment matrix based on the target hierarchy model and importance values, and calculate the characteristic weight vector corresponding to each judgment matrix, and determine the preferred geophysical exploration method based on the characteristic weight vector.

[0014] Furthermore, in the target hierarchy model, the first layer is the preferred geophysical method, the second layer is accuracy, reliability and economy, the third layer is factors corresponding to accuracy, reliability and economy respectively, and the fourth layer is multiple geophysical methods; the factors corresponding to accuracy include faults, rock cracks, groundwater, caves and weak rock masses, the factors corresponding to reliability include the maturity of the technology, the degree of resistance to environmental interference and the accuracy of the results, and the factors corresponding to economy include price, required construction period and required personnel.

[0015] Furthermore, the judgment matrix includes: a first judgment matrix of importance values ​​of each geophysical prospecting method relative to each factor, a second judgment matrix of importance values ​​of each factor corresponding to accuracy relative to accuracy, importance values ​​of each factor corresponding to reliability relative to reliability, and importance values ​​of each factor corresponding to economy relative to economy, and a third judgment matrix of importance values ​​of accuracy, reliability and economy relative to the preferred geophysical prospecting method;

[0016] The feature weight vectors include a first feature weight vector corresponding to the first judgment matrix in one-to-one correspondence, a second feature weight vector corresponding to the second judgment matrix in one-to-one correspondence, and a third feature weight vector corresponding to the third judgment matrix in one-to-one correspondence.

[0017] Further, determining the preferred geophysical exploration method according to the characteristic weight vector specifically includes:

[0018] The first feature weight vector, the second feature weight vector and the third feature weight vector are correspondingly multiplied to obtain the first weight corresponding to each geophysical exploration method, and the geophysical exploration method corresponding to the maximum value of the first weight is used as the preferred geophysical exploration method.

[0019] Furthermore, based on the analytic hierarchy process, the secondary geophysical exploration methods are selected from various geophysical exploration methods, including:

[0020] According to the type of adverse geological conditions in the first geological prediction result, the factors corresponding to accuracy are determined, and the first judgment matrix and the second judgment matrix are screened according to the determined factors corresponding to accuracy. The first characteristic weight vector corresponding to the screened first judgment matrix, the second characteristic weight vector corresponding to the screened second judgment matrix, and the third characteristic weight vector corresponding to the third judgment matrix are multiplied accordingly to obtain the second weight corresponding to each geophysical exploration method, and the geophysical exploration method corresponding to the maximum value of the second weight is used as the secondary geophysical exploration method.

[0021] Furthermore, a first advanced geological prediction model based on a deep learning network is constructed according to the preferred geophysical prospecting method, specifically comprising:

[0022] Collecting a first image training set labeled with types of adverse geological conditions, the first image training set is labeled using a preferred geophysical prospecting method, training a deep learning network based on the first image training set, and obtaining a first advanced geological prediction model;

[0023] According to the secondary geophysical prospecting method, a second advanced geological prediction model based on a deep learning network is constructed, specifically including:

[0024] A second image training set marked with types of adverse geological conditions is collected, the second image training set is marked using a secondary geophysical prospecting method, and a deep learning network is trained based on the second image training set to obtain a second advanced geological prediction model.

[0025] Furthermore, when training the deep learning network, the learning rate selection strategy is set according to the training cycle. The selection strategy is as follows:

[0026] ;

[0027] in, represents the selected learning rate, represents the maximum learning rate for training, represents the corrected set learning rate, Indicates the current training cycle, Indicates the preset training cycle. represents the total training cycle, Represents pi.

[0028] Furthermore, the deep learning network includes a backbone extraction network, an RPN network and a ROI Align network combining ResNet50 and an improved FPN network, and the improved FPN network adds a 3×3 convolution kernel with a dilation rate of 3, a 3×3 convolution kernel with a dilation rate of 5 and a 1×1 convolution kernel.

[0029] Furthermore, the multiple geophysical exploration methods include at least seismic reflection wave method, geological radar method and transient electromagnetic method.

[0030] The beneficial effects of the present application are as follows: the tunnel advanced geological prediction method based on the analytic hierarchy process provided by the present application selects the preferred geophysical method and the secondary geophysical method from a variety of geophysical methods based on the analytic hierarchy process, and then constructs the corresponding advanced geological prediction model after the screening, thereby realizing the mutual verification of different types of geophysical methods and improving the accuracy of the tunnel advanced geological prediction. The present application selects geophysical methods based on the analytic hierarchy process, and only needs to construct the advanced geological prediction model corresponding to the geophysical methods obtained by the screening, and only needs to use a small number of targeted advanced geological prediction models for advanced geological prediction, thereby reducing the workload of comprehensive geological prediction, simplifying the geophysical interpretation process, and improving the efficiency of advanced geological prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic diagram of a process flow of a tunnel advanced geological prediction method based on the analytic hierarchy process provided in an embodiment of the present application;

[0032] Figure 2 A schematic diagram of a target hierarchy model provided in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of the structure of a deep learning network provided in an embodiment of the present application;

[0034] Figure 4 A schematic diagram of the structure of an FPN network provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0036] In some of the processes described in the specification of the present application and the above-mentioned figures, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel, and the sequence numbers of the operations are only used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel.

[0037] The technical solution of the embodiment of the present application is applicable to the application scenarios of tunnel advanced geological prediction, such as advanced geological prediction of tunnel projects such as water diversion tunnels and underground power plants.

[0038] At present, the advanced geological prediction technology combining multiple geophysical exploration usually adopts the scheme of pre-building prediction models corresponding to multiple geophysical exploration methods, and using each prediction model for geological prediction. The inventor found through research that the model training process and geophysical exploration interpretation process of this scheme have a large workload and low efficiency, especially when the geological conditions are good, which will increase the workload and waste manpower and material resources.

[0039] Based on this, the technical solution of the present application is proposed. In an embodiment of the present application, a plurality of geophysical exploration methods are determined, and a preferred geophysical exploration method is screened out from each geophysical exploration method based on the hierarchical analysis method; a first advanced geological prediction model based on a deep learning network is constructed according to the preferred geophysical exploration method, and geological prediction is performed according to the first advanced geological prediction model to obtain a first geological prediction result; if there are adverse geological conditions in the first geological prediction result, a secondary geophysical exploration method is screened out from each geophysical exploration method based on the hierarchical analysis method according to the type of the adverse geological conditions in the first geological prediction result, and a second advanced geological prediction model based on a deep learning network is constructed according to the secondary geophysical exploration method, and geological prediction is performed according to the second advanced geological prediction model to obtain a second geological prediction result; it is determined whether there are two geological prediction results in which both have adverse geological conditions and are of the same type. If so, the two identical geological prediction results are used as the final geological prediction results; otherwise, the above steps are repeated.

[0040] Specifically, the embodiment of the present application first selects the preferred geophysical method from a plurality of geophysical methods based on the hierarchical analysis method, and constructs a first advanced geological prediction model corresponding to the preferred geophysical method, and then uses the first advanced geological prediction model to obtain the first geological prediction result. If there are bad geological conditions in the first geological prediction result, it is necessary to use different geophysical methods for verification. Specifically, according to the type of bad geological conditions in the first geological prediction result and based on the hierarchical analysis method, a secondary geophysical method is selected from a plurality of geophysical methods, and a second advanced geological prediction model corresponding to the secondary geophysical method is constructed, and then the second geological prediction result is obtained using the second advanced geological prediction model. If the second geological prediction result and the first geological prediction result both have bad geological conditions, and the types of the bad geological conditions are the same, then it is used as the final geological prediction result. If they are different, the secondary geophysical methods are repeatedly selected until there are two geological prediction results with bad geological conditions and the same type, and they are used as the final geological prediction result. The embodiment of the present application realizes mutual verification using different types of geophysical methods, thereby improving the accuracy of tunnel advanced geological prediction. Moreover, based on the hierarchical analysis method to screen geophysical methods, it is only necessary to construct an advanced geological prediction model corresponding to the screened geophysical methods, and only a small number of targeted advanced geological prediction models need to be used for advanced geological prediction, thereby reducing the workload of comprehensive geological prediction, simplifying the geophysical interpretation process, and improving the efficiency of advanced geological prediction.

[0041] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with 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 of the embodiments.

[0042] Figure 1 A schematic diagram of a tunnel geological prediction method based on the hierarchical analysis method is shown in Figure 1 , the method comprises the following steps:

[0043] S1. Determine multiple geophysical exploration methods and select the preferred geophysical exploration method from each geophysical exploration method based on the hierarchical analysis method.

[0044] In the embodiment of the present application, the geophysical exploration method includes seismic reflection wave method, geological radar method and transient electromagnetic method.

[0045] In practical applications, based on the previous exploration data and geological catalog data, combined with the accuracy, reliability and economy of the geophysical exploration method as the criteria, the preferred geophysical exploration method is selected from various geophysical exploration methods based on the hierarchical analysis method. Specifically including:

[0046] Determine the factors that affect the selection of each geophysical exploration method, construct a target hierarchy model based on each factor, construct a judgment matrix based on the target hierarchy model and importance values, calculate the characteristic weight vector corresponding to each judgment matrix, and determine the preferred geophysical exploration method based on the characteristic weight vector.

[0047] See also Figure 2 In the embodiment of the present application, the first layer of the target hierarchy model is the preferred geophysical exploration method, the second layer is accuracy, reliability and economy, the third layer includes factors corresponding to accuracy, reliability and economy, and the fourth layer is multiple geophysical exploration methods; the factors corresponding to the accuracy include faults, rock cracks, groundwater, caves and weak rock masses, the factors corresponding to the reliability include the maturity of the technology, the degree of resistance to environmental interference and the accuracy of the results, and the factors corresponding to the economy include price, required construction period and required personnel.

[0048] Among them, the judgment matrix includes: a first judgment matrix of the importance values ​​of each geophysical exploration method relative to each factor, a second judgment matrix of the importance values ​​of each factor corresponding to accuracy relative to accuracy, the importance values ​​of each factor corresponding to reliability relative to reliability, and the importance values ​​of each factor corresponding to economy relative to economy, and a third judgment matrix of the importance values ​​of accuracy, reliability and economy relative to the preferred geophysical exploration method.

[0049] The feature weight vectors include a first feature weight vector corresponding to the first judgment matrix in one-to-one correspondence, a second feature weight vector corresponding to the second judgment matrix in one-to-one correspondence, and a third feature weight vector corresponding to the third judgment matrix in one-to-one correspondence.

[0050] In practical applications, we first need to set the relative importance values ​​of the two indicators based on geological survey data, please refer to Table 1.

[0051] Table 1 Importance value table

[0052]

[0053] Then, a plurality of first judgment matrices of the importance values ​​of each geophysical exploration method (fourth layer) relative to each factor (third layer) are constructed, and the first characteristic weight vector corresponding to each first judgment matrix is ​​determined.

[0054] Among them, the feature weight in the feature weight vector can be discriminated and solved by inputting the judgment matrix into the MATLAB software. Specifically, in the MATLAB software, the first line of code is the judgment matrix input, the second line of code is the feature weight vector and eigenvalue of the judgment matrix, the third line of code is the first column of the judgment matrix, and the last line of code is the normalization operation. The MATLAB software can specifically use the arithmetic mean method, the geometric mean method or the eigenvalue method. The method of solving the feature weight vector by the judgment matrix belongs to the prior art, and the embodiment of the present application will not be repeated here.

[0055] For example, based on the previous data, the importance values ​​of various geophysical exploration methods relative to faults are shown in Table 2.

[0056] Table 2 The importance of each geophysical method relative to the fault

[0057]

[0058] The first judgment matrix constructed according to the above importance value table is as follows:

[0059] ;

[0060] The first judgment matrix The corresponding first feature weight vector is [0.57, 0.33, 0.1].

[0061] By analogy, the following first judgment matrix of each geophysical exploration method relative to rock mass fractures is constructed: , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical prospecting method relative to groundwater , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical exploration method relative to the cave , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical exploration method relative to weak rock mass , and determine the corresponding first feature weight vector ; The first judgment matrix of the maturity of each geophysical exploration method relative to the technology , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical exploration method relative to the degree of resistance to environmental interference , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical method relative to the accuracy of the result , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical exploration method relative to price , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical exploration method relative to the required construction period , and determine the corresponding first feature weight vector ; The first judgment matrix of each geophysical exploration method relative to the required personnel , and determine the corresponding first feature weight vector The first feature weight vector obtained is shown in Table 3.

[0062] Table 3 Correspondence between the first judgment matrix and the first feature weight vector

[0063]

[0064] Then, a plurality of second judgment matrices of the importance values ​​of the factors corresponding to each accuracy (third layer) relative to accuracy, reliability and economy (second layer) are constructed, and the second characteristic weight vector corresponding to each second judgment matrix is ​​determined.

[0065] For example, based on the previous data, the importance values ​​of the factors corresponding to each accuracy relative to accuracy are shown in Table 4.

[0066] Table 4 The importance of each accuracy factor relative to accuracy

[0067]

[0068] The second judgment matrix constructed according to the above importance value table is as follows:

[0069] ;

[0070] The second judgment matrix The corresponding second feature weight vector is [0.42, 0.27, 0.19, 0.09, 0.03].

[0071] By analogy, the following second judgment matrix is ​​constructed: , and determine its corresponding second feature weight vector ; The second judgment matrix of each factor corresponding to the economy relative to the economy , and determine its corresponding second feature weight vector The obtained second feature weight vector is shown in Table 5.

[0072] Table 5 Correspondence between the second judgment matrix and the second feature weight vector

[0073]

[0074] Then, a plurality of third judgment matrices of the importance values ​​of accuracy, reliability and economy (second layer) relative to the preferred geophysical exploration method (first layer) are constructed, and the third characteristic weight vector corresponding to each third judgment matrix is ​​determined.

[0075] For example, based on the preliminary data, the values ​​of the importance of accuracy, reliability and economy relative to the preferred geophysical exploration method are shown in Table 6.

[0076] Table 6 Numerical table of the importance of accuracy, reliability and economy relative to the preferred geophysical method

[0077]

[0078] The third judgment matrix constructed according to the above importance value table is as follows:

[0079] ;

[0080] The third judgment matrix The corresponding third feature weight vector is [0.49, 0.34, 0.17].

[0081] Finally, the first characteristic weight vector, the second characteristic weight vector and the third characteristic weight vector are correspondingly multiplied to obtain the first weight corresponding to each geophysical exploration method, and the geophysical exploration method corresponding to the maximum value of the first weight is used as the preferred geophysical exploration method. For example, the first weight corresponding to the seismic reflection wave method is 0.66×0.68×0.34=0.153, and the first weight is the maximum value of the first weight, so the seismic reflection wave method is used as the preferred geophysical exploration method.

[0082] S2. Construct a first advanced geological prediction model based on a deep learning network according to the preferred geophysical prospecting method, and perform geological prediction according to the first advanced geological prediction model to obtain a first geological prediction result. If there are adverse geological conditions in the first geological prediction result, enter S3.

[0083] In an embodiment of the present application, a method for constructing a first advanced geological prediction model includes:

[0084] A first image training set labeled with types of adverse geological conditions is collected, the first image training set is labeled using a preferred geophysical exploration method, a deep learning network is trained based on the first image training set, and a first advanced geological prediction model is obtained.

[0085] For example, if the preferred geophysical exploration method is seismic reflection wave method, the results of seismic reflection wave method are collected to make the first image training set, the unfavorable geological types in the image are marked, and then it is input into the deep learning network for training to obtain the first advanced geological prediction model. Among them, the deep learning network includes the backbone extraction network, RPN network and ROI Align network combined with ResNet50 and FPN network, please refer to Figure 3 ,The training steps include inputting the labeled image into the backbone extraction network, performing the feature extraction function, then inputting the extracted backbone feature map into the RPN network, generating the interested pre-fetch, and finally performing ROIAlign operation, correcting the feature map, and then mapping the corrected feature map into a fixed vector, and finally outputting the recognition result.

[0086] In the embodiment of the present application, when training a deep learning network, a learning rate selection strategy is set according to the training cycle, and the selection strategy is as follows:

[0087] ;

[0088] in, represents the selected learning rate, represents the maximum learning rate for training, represents the corrected set learning rate, Indicates the current training cycle, Indicates the preset training cycle. represents the total training cycle, Represents pi.

[0089] In the early training, indivual Internal The learning rate is , in the later stage of training, Corrected setting of learning rate to reduce late oscillations.

[0090] See also Figure 4 In the embodiment of the present application, the FPN network is improved to enhance the feature recognition effect of adverse geological phenomena in the forecast results. The main method is to add a dilated convolution to calculate the feature map. The FPN network includes a 3×3 convolution kernel with a dilation rate of 3, a 3×3 convolution kernel with a dilation rate of 5, and a 1×1 convolution kernel, so as to obtain feature maps of different receptive fields, and finally use feature fusion to output the image. Figure 4 In , C2, C3, C4, and C5 represent the generated feature matrices, and P2, P3, P4, and P5 represent the generated prediction feature layers of different scales.

[0091] After the first advanced geological prediction model is obtained through training, the images to be predicted collected on-site are input into the first advanced geological prediction model to obtain the corresponding first geological prediction results. The first geological prediction results include whether there are adverse geological conditions and the types of adverse geological conditions when they exist. If there are no adverse geological conditions in the first geological prediction results, the process ends and construction can continue. If there are adverse geological conditions in the first geological prediction results, the subsequent steps are entered, and different geophysical methods are used for cross-validation to improve the accuracy of the advanced geological prediction.

[0092] S3. According to the type of adverse geological conditions in the first geological prediction result, a secondary geophysical method is screened out from various geophysical methods based on the hierarchical analysis method, and a second advanced geological prediction model based on a deep learning network is constructed according to the secondary geophysical method. Geological prediction is performed according to the second advanced geological prediction model to obtain a second geological prediction result.

[0093] In the embodiment of the present application, the secondary geophysical exploration method is screened out from various geophysical exploration methods based on the hierarchical analysis method, specifically including:

[0094] According to the type of adverse geological conditions in the first geological prediction result, the factors corresponding to accuracy are determined, and the first judgment matrix and the second judgment matrix are screened according to the determined factors corresponding to accuracy. The first characteristic weight vector corresponding to the screened first judgment matrix, the second characteristic weight vector corresponding to the screened second judgment matrix, and the third characteristic weight vector corresponding to the third judgment matrix are multiplied accordingly to obtain the second weight corresponding to each geophysical exploration method, and the geophysical exploration method corresponding to the maximum value of the second weight is used as the secondary geophysical exploration method.

[0095] For example, the preferred geophysical method is the seismic reflection wave method, and its corresponding first geological prediction result is groundwater enrichment. In this case, only groundwater is retained in the factors corresponding to accuracy, and the first judgment matrix and the second judgment matrix are screened according to the retained groundwater, that is, only the first judgment matrix and the second judgment matrix related to groundwater are retained. Then, the first characteristic weight vector corresponding to the screened first judgment matrix, the second characteristic weight vector corresponding to the screened second judgment matrix, and the third characteristic weight vector corresponding to the third judgment matrix are multiplied accordingly to obtain the second weight corresponding to each geophysical method, and the geophysical method corresponding to the maximum value of the second weight is used as the secondary geophysical method. For example, the second weight corresponding to the transient electromagnetic method is 0.68×0.19×0.49=0.06, and the second weight is the maximum value of the second weight. In this case, the transient electromagnetic method is used as the secondary geophysical method.

[0096] After determining the secondary geophysical exploration method, a second advanced geological prediction model corresponding to the secondary geophysical exploration method can be constructed. The construction method of the second advanced geological prediction model is the same as the construction method of the first advanced geological prediction model, and will not be repeated in the embodiments of this application.

[0097] After constructing the second advanced geological prediction model, the images to be predicted collected on-site are input into the second advanced geological prediction model to obtain the corresponding second geological prediction results, which include whether there are adverse geological conditions and the types of adverse geological conditions when they exist.

[0098] S4. Determine whether there are two geological prediction results with the same type and both of which have adverse geological conditions. If so, take the two identical geological prediction results as the final geological prediction results. Otherwise, re-enter S3.

[0099] In practical applications, if the second geological prediction result and the first geological prediction result both contain unfavorable geological conditions and the type of unfavorable geological conditions is the same, then it will be used as the final geological prediction result. If the second geological prediction result is different from the first geological prediction result, the secondary geophysical methods will be repeatedly screened until there are two geological prediction results that both contain unfavorable geological conditions and the same type, and then they will be used as the final geological prediction result. In practical applications, the unfavorable geological conditions in the final geological prediction result must be dealt with through corresponding engineering measures before construction can continue. Cross-validation is performed through a variety of different geophysical methods to improve the accuracy of tunnel advance geological prediction.

[0100] In summary, the tunnel advanced geological prediction method based on the analytic hierarchy process provided in the embodiment of the present application selects the preferred geophysical method and the secondary geophysical method from a variety of geophysical methods based on the analytic hierarchy process, and then constructs the corresponding advanced geological prediction model after screening, thereby realizing the mutual verification of different types of geophysical methods and improving the accuracy of the tunnel advanced geological prediction. The present application selects geophysical methods based on the analytic hierarchy process, and only needs to construct the advanced geological prediction model corresponding to the geophysical methods obtained by screening, and only needs to use a small number of targeted advanced geological prediction models for advanced geological prediction, thereby reducing the workload of comprehensive geological prediction, simplifying the geophysical interpretation process, and improving the efficiency of advanced geological prediction.

Claims

1. A tunnel advanced geological prediction method based on the analytic hierarchy process, characterized in that: The method comprises: S1. Determine multiple geophysical exploration methods and select the preferred geophysical exploration method from each geophysical exploration method based on the hierarchical analysis method; S2, constructing a first advanced geological prediction model based on a deep learning network according to the preferred geophysical prospecting method, and performing geological prediction according to the first advanced geological prediction model to obtain a first geological prediction result. If there are unfavorable geological conditions in the first geological prediction result, proceeding to S3; S3. According to the type of adverse geological conditions in the first geological prediction result, a secondary geophysical method is selected from various geophysical methods based on the hierarchical analysis method, a second advanced geological prediction model based on a deep learning network is constructed according to the secondary geophysical method, and geological prediction is performed according to the second advanced geological prediction model to obtain a second geological prediction result; S4, judging whether there are two geological prediction results with bad geological conditions and the same type, if so, taking the two same geological prediction results as the final geological prediction results, otherwise, re-entering S3; According to the preferred geophysical prospecting method, a first advanced geological prediction model based on a deep learning network is constructed, specifically including: Collecting a first image training set labeled with types of adverse geological conditions, the first image training set is labeled using a preferred geophysical prospecting method, training a deep learning network based on the first image training set, and obtaining a first advanced geological prediction model; According to the secondary geophysical prospecting method, a second advanced geological prediction model based on a deep learning network is constructed, specifically including: Collecting a second image training set marked with types of adverse geological conditions, the second image training set is marked using a secondary geophysical prospecting method, and training a deep learning network based on the second image training set to obtain a second advanced geological prediction model; When training a deep learning network, the learning rate selection strategy is set according to the training cycle. The selection strategy is as follows: Among them, Lr represents the selected learning rate, Lr max Indicates the maximum learning rate of training, Lr setting represents the corrected set learning rate, epoch represents the current training cycle, t represents the preset training cycle, T represents the total training cycle, and Π represents pi; The deep learning network includes a backbone extraction network, an RPN network and a ROI Align network that are a combination of ResNet50 and an improved FPN network. The improved FPN network adds a 3×3 convolution kernel with a dilation rate of 3, a 3×3 convolution kernel with a dilation rate of 5 and a 1×1 convolution kernel.

2. The tunnel advanced geological prediction method based on the analytic hierarchy process according to claim 1 is characterized in that: Based on the analytic hierarchy process, the preferred geophysical exploration method is selected from various geophysical exploration methods, including: Determine the factors that affect the selection of each geophysical exploration method, construct a target hierarchy model based on each factor, and set importance values, construct a judgment matrix based on the target hierarchy model and importance values, and calculate the characteristic weight vector corresponding to each judgment matrix, and determine the preferred geophysical exploration method based on the characteristic weight vector.

3. The tunnel advanced geological prediction method based on the analytic hierarchy process according to claim 2 is characterized in that: In the target hierarchy model, the first layer is the preferred geophysical method, the second layer is accuracy, reliability and economy, the third layer is factors corresponding to accuracy, reliability and economy respectively, and the fourth layer is multiple geophysical methods; the factors corresponding to accuracy include faults, rock fissures, groundwater, caves and weak rock masses, the factors corresponding to reliability include the maturity of the technology, the degree of resistance to environmental interference and the accuracy of the results, and the factors corresponding to economy include price, required construction period and required personnel.

4. The tunnel advanced geological prediction method based on the analytic hierarchy process according to claim 3 is characterized in that: The judgment matrix includes: a first judgment matrix of the importance values ​​of each geophysical prospecting method relative to each factor, a second judgment matrix of the importance values ​​of each factor corresponding to accuracy relative to accuracy, the importance values ​​of each factor corresponding to reliability relative to reliability, and the importance values ​​of each factor corresponding to economy relative to economy, and a third judgment matrix of the importance values ​​of accuracy, reliability and economy relative to the preferred geophysical prospecting method; The feature weight vectors include a first feature weight vector corresponding to the first judgment matrix in one-to-one correspondence, a second feature weight vector corresponding to the second judgment matrix in one-to-one correspondence, and a third feature weight vector corresponding to the third judgment matrix in one-to-one correspondence.

5. The tunnel advanced geological prediction method based on the analytic hierarchy process according to claim 4 is characterized in that: Determining the preferred geophysical exploration method according to the characteristic weight vector specifically includes: The first feature weight vector, the second feature weight vector and the third feature weight vector are correspondingly multiplied to obtain the first weight corresponding to each geophysical exploration method, and the geophysical exploration method corresponding to the maximum value of the first weight is used as the preferred geophysical exploration method.

6. The tunnel advanced geological prediction method based on the analytic hierarchy process according to claim 4 is characterized in that: Based on the analytic hierarchy process, the secondary geophysical exploration methods are selected from various geophysical exploration methods, including: According to the type of adverse geological conditions in the first geological prediction result, the factors corresponding to accuracy are determined, and the first judgment matrix and the second judgment matrix are screened according to the determined factors corresponding to accuracy. The first characteristic weight vector corresponding to the screened first judgment matrix, the second characteristic weight vector corresponding to the screened second judgment matrix, and the third characteristic weight vector corresponding to the third judgment matrix are multiplied accordingly to obtain the second weight corresponding to each geophysical exploration method, and the geophysical exploration method corresponding to the maximum value of the second weight is used as the secondary geophysical exploration method.

7. The tunnel advanced geological prediction method based on the analytic hierarchy process according to claim 1 is characterized in that: The multiple geophysical exploration methods include at least seismic reflection wave method, geological radar method and transient electromagnetic method.

Images