TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement

Abstract

The invention discloses a TBM surrounding rock classification intelligent identification method based on a penetration index and Bayesian enhancement. The method comprises the steps that 1, surrounding rock classification is conducted; 2, constructing a primary surrounding rock classification model based on a penetration index; 3, constructing a primary surrounding rock classification model sample library; 4, training a primary surrounding rock classification model; 5, testing the primary surrounding rock classification model; step 6, calculating a prior probability; 7, calculating a conditional probability; 8, tunneling is conducted; step 9, calculating a penetration index of the tunneling section; step 10, primary prediction of tunneling section surrounding rock classification; and step 11, Bayesian posteriori estimation is carried out. According to the method, real-time calculation can be carried out according to field machine rock breaking tunneling monitoring data, the field geological exploration and indoor test cost can be effectively reduced, the intelligent prediction method based on Bayesian enhancement can obtain the posterior result of the tunnel face position based on prediction of the excavated section, and then the credibility of the prediction result is improved.

Description

Technical Field

[0001] The invention relates to the application field of TBM construction technology, in particular to a TBM surrounding rock classification intelligent recognition method based on penetration index and Bayesian enhancement. Background Art

[0002] Tunnel boring machines (TBMs) are widely used in the construction of urban underground space projects and tunnel projects in vast deep buried areas in my country. Since the cutter shield of the tunnel boring machine provides a closed construction space for workers to work on site during the excavation process, it has a safe and friendly construction environment. But at the same time, the cutter shield also hinders the geological survey and on-site mapping of the tunnel surrounding rock, especially the tunnel face position, by on-site geological engineers, which leads to serious lags in the assessment of face stability and classification of surrounding rock support. However, tunnel boring machines are very sensitive to changes in the geological environment. Failure to timely and accurately identify the adverse geological conditions and their categories at the face position can easily lead to engineering disasters such as machine jams, while increasing the time for on-site support material preparation and coordination, reducing construction efficiency.

[0003] Traditional geological survey work for TBM excavation faces is mainly based on advance drilling, geological mapping of the intermittent excavation face during machine shutdown, etc. to obtain rock physical mechanics and joint parameters. Related methods need to rely on indoor physical and mechanical strength tests, which have the problems of high test time and economic cost, long result analysis process, and cannot meet the rapid diagnosis and prediction of tunnel construction sites. At the same time, related tests will seriously interfere with the normal excavation construction process of TBM and reduce the excavation speed of the machine.

[0004] Therefore, it is urgent to develop a classification and identification method for the surrounding rock of the tunnel face during the TBM construction process, so as to achieve low-cost, high-quality and rapid identification of the surrounding rock of the tunnel face while reducing the impact on the construction process. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement in view of the deficiencies of the above-mentioned prior art. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement first extracts the penetration index reflecting the difficulty of rock excavation based on the physical rules of the interaction between the cutter head and the rock mass during the TBM construction process. The index can be calculated in real time based on the on-site machine rock breaking and excavation monitoring data, which can effectively reduce the cost of on-site geological surveys and indoor tests. At the same time, the index extraction method of on-site monitoring data can reduce the impact on the construction process. In addition, in order to further eliminate the impact of monitoring data quality, the intelligent prediction method based on Bayesian enhancement can obtain the posterior result of the tunnel face position based on the prediction of the excavated section, thereby improving the credibility of the prediction result. This technical solution can quickly and accurately identify the classification of the tunnel face surrounding rock during TBM construction and excavation, ensure construction safety and guide the selection of on-site support measures.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement includes the following steps.

[0008] Step 1: Classification of surrounding rock: Based on rock strength, degree of integrity, structural surface state, occurrence and groundwater conditions, the surrounding rock is divided into P types; then, a digital classification label C is marked for each type of surrounding rock.

[0009] Step 2: Construct a primary surrounding rock classification model based on penetration index: A random forest algorithm is used to construct a primary surrounding rock classification model based on penetration index for each excavation section during the TBM construction process; the input layer of the primary surrounding rock classification model is the penetration index vector of the excavation section during the TBM construction process. The output layer is the surrounding rock classification label C of the corresponding excavation section; where a is the linear fitting slope of the cutterhead single blade thrust and penetration in the excavation section; b is the initial penetration thrust of the cutterhead roller in the excavation section; It indicates the linear fitting goodness of the thrust and penetration of the cutterhead in the excavation section; TPI is the linear fitting slope of the torque and penetration of the cutterhead in the excavation section; Indicates the goodness of linear fit between single-blade torque and penetration in the tunneling section.

[0010] Step 3: Construct a surrounding rock classification model sample library, which specifically includes the following steps:

[0011] Step 3-1, collect effective excavation section data: collect effective excavation section data of all types of surrounding rock in step 1; wherein, the number of effective excavation sections of each type of surrounding rock is not less than M; M≥100.

[0012] Step 3-2, extracting penetration index of tunneling section: extract the penetration index vector for each tunneling section of each type of surrounding rock, and record the actual surrounding rock classification label C corresponding to each tunneling section, so as to obtain a surrounding rock classification model sample library containing N sample data; each sample data includes

[0013] Step 3-3, sample library classification: divide the sample library into two parts: training set and test set.

[0014] Step 4: training a surrounding rock classification model: using the training set obtained in step 3, the surrounding rock classification model constructed in step 2 is trained to obtain a trained surrounding rock classification model.

[0015] Step 5: Test the surrounding rock classification model: Use the penetration index vector of each sample data in the test set obtained in step 3 And the surrounding rock classification model trained in step 4 is tested to obtain a surrounding rock classification prediction result.

[0016] Step 6, calculate the prior probability: use the sample library of the primary surrounding rock classification model constructed in step 3 to calculate the prior probability of stable surrounding rock and the prior probability of unstable surrounding rock.

[0017] Step 7, calculate conditional probability: compare and count the surrounding rock classification prediction result obtained in step 5 and the actual surrounding rock classification of each sample data in the test set to obtain the conditional probability.

[0018] Step 8, tunnel excavation: The TBM constructs the tunnel to be excavated, and monitors the excavation data of the cutter head single blade thrust and single blade torque for each excavation section; assuming that the current excavation section is t, the excavation data of the t-m+1th excavation section, the t-m+2th excavation section, ..., the tth excavation section are recorded.

[0019] Step 9, calculate the penetration index of the excavation section: according to the excavation data recorded in step 8, calculate the penetration index vectors of the t-m+1th excavation section, the t-m+2th excavation section, ..., the tth excavation section.

[0020] Step 10: Primary prediction of surrounding rock classification of excavation sections: Substitute the penetration index vectors of the t-m+1th excavation section, t-m+2th excavation section, ..., tth excavation section obtained in step 9 into the primary surrounding rock classification model trained in step 4 to obtain the surrounding rock classification sequence [c t-m+1 ,c t-m+2 ,...,c t ].

[0021] Step 11, Bayesian posterior estimation: According to the surrounding rock classification sequence [ct-m+1 ,c t-m+2 ,...,c t ], the prior probability obtained in step 6 and the conditional probability obtained in step 7, so as to obtain the posterior probability of the unstable category of the surrounding rock of the t-th excavation section, and then obtain the surrounding rock classification of the tunnel face.

[0022] In step 1, P=2, which means stable surrounding rock and unstable surrounding rock respectively; according to the national standard "Specification for Geological Investigation of Water Conservancy and Hydropower Engineering GB50487-2008", the surrounding rock is divided into N=5 types, namely Class I, II, III, IV and V surrounding rock; among them, Class I, II and III surrounding rock are stable surrounding rock, and Class IV and V surrounding rock are unstable surrounding rock.

[0023] In step 1, the classification label of stable surrounding rock is marked as 0, and the classification label of unstable surrounding rock is marked as 1.

[0024] In step 8, the calculation formula for m is:

[0025]

[0026] Where L is the length of the cutter shield; d is the maximum travel distance of a single excavation section.

[0027] In step 7, the conditional probability is that the model predicts that the surrounding rock category is c i Under this condition, its true category is c j The probability of j / c i ), then P(c j / c i ) is calculated as:

[0028]

[0029] Where N ji Indicates that the predicted surrounding rock type is c i But the actual surrounding rock type is c j The number of samples at .

[0030] N i Indicates that the predicted surrounding rock type is c i The total number of samples under .

[0031] In step 11, the Bayesian posterior estimation method includes the following steps.

[0032] Step 11-1, calculate the predicted posterior probability P(c i / c j ) t-m+1 , specifically including the following steps.

[0033] Step 11-1A, determine the prior probability of the t-m+1th excavation section: The prior probability of the t-m+1th excavation section includes t -m+1 The prior probability P(c i ) t-m+1 and c t-m+1 Prior probability of the corresponding surrounding rock Among them, P(c i ) t-m+1 Find it from step 6;

[0034] Step 11-1B, select the conditional probability of the t-m+1th excavation section: The prior probability of the t-m+1th excavation section includes P(c j / c i )and in, For c t-m+1 The true category under the opposite type of surrounding rock conditions is c j The probabilities of are obtained from step 7.

[0035] Step 11-1C, calculate P(c i / c j ) t-m+1 :P(c i / c j ) t-m+1 is the same as c in the t-m+1 excavation section t-m+1 The specific calculation formula for the posterior probability of the corresponding surrounding rock is:

[0036]

[0037] Step 11-2, calculate the prior probability of the t-m+2th excavation section: The prior probability of the t-m+2th excavation section includes t -m+2 The prior probability P(c i ) t-m+2 and c t-m+2 Prior probability of opposite type of surrounding rock but:

[0038] P(c i ) t-m+2 =P(c i / c j ) t-m+1

[0039]

[0040] Step 11-3: Repeat steps 11-1B to 11-1C to calculate the t-m+2th excavation section and c t-m+2The posterior probability P(c i / c j ) k-m+2 .

[0041] Step 11-4, repeat steps 11-2 to 11-3, and calculate the posterior probability P(c i / c j ) t-m+3 ,……,P(c i / c j ) t .

[0042] Step 11-5: Secondary determination of surrounding rock classification: Based on the P(c i / c j ) t , the surrounding rock category with the maximum predicted probability is taken as the surrounding rock category of the tunnel face in front of the current t-th excavation section.

[0043] In step 4, the best hyperparameter combination of random forest is selected based on the cross-validation method of grid search.

[0044] In step 4, the cross-validation method selects the best hyperparameter combination of random forests, including the following steps:

[0045] Step 4-1. Design hyperparameter combinations: For random forest, design h hyperparameter combinations.

[0046] Step 4-2: Divide the training subsets: Divide the total training set into n training subsets D by bootstrapping method i .

[0047] Step 4-3, build a random forest: for each training subset D i Independent surrounding rock classification prediction decision trees are constructed, and a total of n surrounding rock classification prediction decision trees are obtained to form a random forest;

[0048] Step 4-4, hyperparameter training: Using the random forest built in step 4-3, each combination of hyperparameters designed in step 4-1 is trained; during the training process, the generation process of each surrounding rock classification prediction decision tree includes the following steps:

[0049] Step 4-4A, calculate the node Gini index: when each node is split, calculate its node Gini index; the calculation formula of the Gini index G(w) of node w is:

[0050]

[0051] In the formula, p cis the probability that the current node w belongs to the surrounding rock category c.

[0052] Step 4-4B, calculate the best split point: For each node, calculate each feature f j All possible split points of f, find a split point that can maximize the information gain; where f j ∈[a,b,TPI], then the calculation formula for the optimal split point s is:

[0053]

[0054] Where D t is the sample set of the current node, and According to f j The left sample and the right sample after the feature split point s segmentation; G(D t ), and Node D t Node D t and nodes The Gini index.

[0055] Step 4-4C, recursive splitting: Repeat steps 4-4A to 4-4B, select the best splitting nodes for the left node and the right node respectively, and recursively split them until the stop growth condition of the surrounding rock classification prediction decision tree is reached.

[0056] Step 4-5, select the optimal hyperparameters: Calculate the average accuracy of each combination of hyperparameters during validation, and select the set of hyperparameters with the highest average accuracy as the optimal hyperparameters for random forest.

[0057] In step 4, when the primary surrounding rock classification model is trained to output the surrounding rock classification label of the excavation section, a prediction model probability threshold is set. The primary surrounding rock classification model compares the probability of each surrounding rock classification with the prediction model probability threshold, and takes the surrounding rock classification greater than the prediction model probability threshold as the output of the primary surrounding rock classification model; wherein the prediction model probability threshold can be adjusted to eliminate the problem of imbalance of samples of different surrounding rock categories.

[0058] In step 4, the method for adjusting the probability threshold of the prediction model includes the following steps:

[0059] Step 4A-1, calculate the maximum ratio of sample categories: calculate the maximum number of samples of different surrounding rock categories in the training set T max and the minimum number of class samples T min The specific calculation formula is:

[0060]

[0061] Step 4A-2, assessment of surrounding rock category sample imbalance: compare the ratio T calculated in step 4A-1 with the set ratio threshold Y. When T≤Y, the surrounding rock category samples are considered balanced and the prediction model probability threshold does not need to be adjusted; otherwise, the surrounding rock category samples are considered unbalanced and proceed to step 4A-3.

[0062] Step 4A-3, determine the initial threshold of the prediction model probability: set the initial threshold of the prediction model probability of stable surrounding rock and unstable surrounding rock to the average value of the two

[0063] Step 4A-4, prediction model probability threshold adjustment: calculate the percentage K of stable surrounding rock and unstable surrounding rock samples in the total samples, and compare K with Compare and according to the comparison results, Make adjustments, specifically:

[0064] A. For The surrounding rock type is increased in sequence according to the initial threshold of the prediction model probability in the order of Δs, 2Δs, 3Δs, ..., RΔs; where:

[0065] B. For The surrounding rock type is reduced in sequence according to the order of Δs, 2Δs, 3Δs, ..., RΔs; where: While keeping all The added value of surrounding rock category is less than The reduction values ​​are the same for the surrounding rock categories.

[0066] Step 4A-5, F1 value evaluation: For each increase or decrease in the prediction model probability threshold, an F1 value evaluation is performed, and a curve of the change of the F1 value with the prediction model probability threshold is constructed.

[0067] Step 4A-6, selecting the optimal prediction model probability threshold: the prediction model probability threshold corresponding to the maximum F1 value in the curve of the change of F1 value with the prediction model probability threshold is used as the optimal prediction model probability threshold of the primary surrounding rock classification model.

[0068] The present invention has the following beneficial effects:

[0069] (1) The present invention uses a Bayesian enhanced ensemble learning method to intelligently identify the surrounding rock classification during TBM construction. The prediction results can evaluate the surrounding rock classification of the tunnel boring machine, especially provide real-time early warning and forecast when encountering extremely unstable sections, which can effectively reduce the risk of engineering instability and reduce accidents and disasters. In addition, the surrounding rock classification evaluation results can provide guidance for the selection of on-site surrounding rock support methods, which is conducive to further reducing support costs and improving construction efficiency.

[0070] (2) The monitoring data of the rock-machine interaction process during TBM construction is used to extract the excavability index parameters as the input parameters of the intelligent prediction model. The measurement of relevant on-site monitoring data can rely on intelligent sensor devices. Compared with indoor tests, advance drilling and other methods, the data acquisition cost is lower and has little impact on the normal construction organization.

[0071] (3) Bayesian updating is used to further perform posterior estimation on the probability sequence prediction results of the machine learning model. The Bayesian enhancement method can effectively reduce the variability of the prediction results caused by the quality of monitoring data, and at the same time consider the influence of the previous monitoring data of the current tunneling section, thereby improving the credibility of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 This is a flow chart of the intelligent identification method for TBM surrounding rock classification based on penetration index and Bayesian enhancement.

[0073] Figure 2 These are the penetration indices extracted from the 57123 position of the Chuojiliao pile in the Yinchuojiliao Tunnel; among them, (a) is the single-blade thrust-penetration index; (b) is the single-blade torque-penetration index.

[0074] Figure 3 Schematic diagram of optimal segmentation features and score selection in the decision tree generation process.

[0075] Figure 4 A comparison chart of the prediction performance indicators using the adjusted prediction model probability threshold.

[0076] Figure 5 Flowchart for posterior estimation of probability results of enhanced ensemble learning based on Bayesian method.

[0077] Figure 6 This is the relationship diagram between the surrounding rock classification probability and pile number in the section from pile number 26+030 to 25+980 of the No. 3 tunnel of Yinchuojiliao.

[0078] Figure 7 This is a comparison chart of the accuracy of the random forest algorithm and the Bayesian enhanced algorithm used in the present invention. DETAILED DESCRIPTION

[0079] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific preferred embodiments.

[0080] The present invention is described in detail with the collected on-site monitoring data of the construction process of the tunnel boring machine of the No. 6 branch tunnel of the No. 2 bidding section of the Inner Mongolia Yinchuojiliao Project.

[0081] like Figure 1 As shown, a TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement includes the following steps.

[0082] Step 1. Classification of surrounding rocks: Based on rock strength, integrity, structural surface state, occurrence and groundwater conditions, the surrounding rocks are preferably divided into P = 2 types of surrounding rocks, namely stable and unstable; then, a digital classification label C is marked for each type of surrounding rock.

[0083] In this embodiment, the main rock breaking parameters of the tunnel boring machine construction process of the Yinchuojiliao 2-6 branch tunnel are collected. The main excavation parameters of the TBM, such as cutterhead thrust, torque, penetration and cutterhead speed parameter values, are collected at a sampling frequency of 1Hz / s. The purpose of the tunnel construction is a cross-regional water transfer tunnel. According to the national standard "Specifications for Geological Survey of Water Conservancy and Hydropower Engineering GB50487-2008", the surrounding rock is divided into N=5 types, namely Class I, II, III, IV and V surrounding rock; among them, Class I, II and III surrounding rock are stable types of surrounding rock, and Class IV and V surrounding rock are unstable types of surrounding rock. Further, the classification label of the stable surrounding rock is marked as 0, and the classification label of the unstable surrounding rock is marked as 1.

[0084] Step 2: Construct a primary surrounding rock classification model based on penetration index: A random forest algorithm is used to construct a primary surrounding rock classification model based on penetration index for each excavation section during the TBM construction process; the input layer of the primary surrounding rock classification model is the penetration index vector of the excavation section during the TBM construction process. The output layer is the surrounding rock classification label C of the corresponding excavation section; where a is the linear fitting slope of the cutterhead single blade thrust and penetration in the excavation section; b is the initial penetration thrust of the cutterhead roller in the excavation section; It indicates the linear fitting goodness of the thrust and penetration of the cutterhead in the excavation section; TPI is the linear fitting slope of the torque and penetration of the cutterhead in the excavation section; Indicates the goodness of linear fit between single-blade torque and penetration in the tunneling section.

[0085] Step 3: Construct a surrounding rock classification model sample library, which specifically includes the following steps:

[0086] Step 3-1, collect effective excavation section data: collect effective excavation section data of all types of surrounding rock in step 1; wherein, the number of effective excavation sections of each type of surrounding rock is not less than M; M≥100.

[0087] In this embodiment, the on-site rock breaking parameters of the 6 tunnels of the Yinchuojiliao No. 2 section include M=5797 complete and effective rock breaking excavation sections. Each set of excavation section data includes a sequence of total cutterhead thrust, cutterhead torque, propulsion speed, and cutterhead rotation speed.

[0088] Step 3-2, extracting penetration index of tunneling section: extract the penetration index vector for each tunneling section of each type of surrounding rock, and record the actual surrounding rock classification label C corresponding to each tunneling section, so as to obtain a surrounding rock classification model sample library containing N sample data; each sample data includes Penetration sequence calculation. Based on the advancement speed v and the cutter head speed n, the penetration p that reflects the depth of the intruded rock mass when the cutter head rotates one circle is calculated. The calculation formula is as follows:

[0089]

[0090] First, the total thrust sequence and torque sequence of the cutterhead are divided by the number of cutters on the TBM cutterhead to obtain the thrust and torque of the single cutter. Then, the thrust and penetration of the single cutter, and the torque and penetration of the single cutter of each excavation section are linearly fitted to obtain the rock penetration index.

[0091] In this embodiment, linear fitting is performed on the single-blade thrust, single-blade torque and penetration of the effective rock-breaking extremes in the excavation section with the pile number 57123 of the No. 6 branch tunnel of the Yinchuojiliao 2 section to obtain the rock penetration index. On this basis, combined with the classification labels of 5797 excavation sections collected in the project, an initial database is constructed, as shown in Table 1 below.

[0092] Table 1 Partial example of the database of the excavation section of the Yinchuojiliao 2-6 branch tunnel

[0093]

[0094] Step 3-3, sample library classification: divide the sample library into two parts: training set and test set.

[0095] In the present invention, a 10-fold algorithm is used to divide the sample library into 10 folds, 9 folds of which are used for model training, and the remaining 1 fold of data is used for verification.

[0096] The sample volume ratio of rock penetration index under stable and unstable working conditions in Tunnel No. 6 of Yinchuojiliao Section 2 is 4949:848. The number of samples under the two working conditions is different. This technical solution uses the method of stratified sampling according to working conditions to divide the data. The specific steps are as follows:

[0097] First, 90% of the 4949 samples in the surrounding rock stability category of working condition 1, namely 4454 samples, are randomly selected as the working condition 1 training set, and the remaining 495 samples are used as the working condition 1 test set.

[0098] Then, 90% of the 848 samples in the stable and indeterminate category of surrounding rock under working condition 2, namely 763 samples, are randomly selected as the training set of working condition 2, and the remaining 85 samples are used as the test set of working condition 2.

[0099] The training sets under the above two working conditions are combined into the total training set, and the test sets under the above two working conditions are combined into the total test set. The final total training set contains 5217 samples, and the test set contains 580 samples.

[0100] Step 4: training a surrounding rock classification model: using the training set obtained in step 3, the surrounding rock classification model constructed in step 2 is trained to obtain a trained surrounding rock classification model.

[0101] In the present invention, a cross-validation method based on grid search is preferably used to select the best hyperparameter combination of random forest.

[0102] In step 4, the cross-validation method selects the best hyperparameter combination of random forests, including the following steps:

[0103] Step 4-1. Design hyperparameter combinations: For random forest, design h hyperparameter combinations.

[0104] In this embodiment, the designed hyperparameter combinations are: number of decision trees [50, 100, 200, 300], minimum number of sample splits [14, 28, 42, 56], maximum number of features [2, 3, 4, 5], node splitting criteria [gini, entropy] 4 hyperparameter combinations. Among them, gini represents the Gini index; entropy represents the cross entropy value.

[0105] Step 4-2: Divide the training subsets: Divide the total training set into n training subsets D by bootstrapping method i .

[0106] Step 4-3, build a random forest: for each training subset D i Independent surrounding rock classification prediction decision trees are constructed, and a total of n surrounding rock classification prediction decision trees are obtained to form a random forest.

[0107] Step 4-4, hyperparameter training: Use the random forest built in step 4-3 to train the hyperparameters of each combination designed in step 4-1.

[0108] like Figure 3 As shown in the figure, during the training process, the generation process of each surrounding rock classification prediction decision tree includes the following steps.

[0109] Step 4-4A, calculate the node Gini index: when each node is split, calculate its node Gini index; the calculation formula of the Gini index G(w) of node w is:

[0110]

[0111] In the formula, p c is the probability that the current node w belongs to the surrounding rock category c.

[0112] Step 4-4B, calculate the best split point: For each node, calculate each feature f j All possible split points of f, find a split point that can maximize the information gain; where f j ∈[a,b,TPI], then the calculation formula for the optimal split point s is:

[0113]

[0114] Where D t is the sample set of the current node, and According to f j The left sample and the right sample after the feature split point s segmentation; G(D t ), and G(D tR ) are nodes D t Node D t and nodes The Gini index.

[0115] In this embodiment, for each node, each feature f is calculated j ∈[a,b,TPI], find a split point that maximizes the information gain. Taking TPI as an example, the initial Gini value minimum split point Gini(TPI)=0.27 is at TPI=3.95, which is smaller than the Gini values ​​of features a and b Gini(a)=0.37, Gini(b)=0.42. Therefore, the optimal split point is at TPI=3.95.

[0116] Step 4-4C, recursive splitting: Repeat steps 4-4A to 4-4B, select the best splitting nodes for the left node and the right node respectively, and recursively split them until the stop growth condition of the surrounding rock classification prediction decision tree is reached.

[0117] Step 4-5, select the optimal hyperparameters: Calculate the average accuracy of each combination of hyperparameters during validation, and select the set of hyperparameters with the highest average accuracy as the optimal hyperparameters for random forest.

[0118] In this embodiment, when the hyperparameter combination is [100, 28, 3, 'gini'], the accuracy is the highest, so the model trained with this set of hyperparameter combination is selected for model testing.

[0119] Furthermore, when the primary surrounding rock classification model is trained to output the surrounding rock classification label of the excavation section, a prediction model probability threshold is set. The primary surrounding rock classification model compares the probability of each surrounding rock classification with the prediction model probability threshold, and takes the surrounding rock classification greater than the prediction model probability threshold as the output of the primary surrounding rock classification model; wherein the prediction model probability threshold can be adjusted to eliminate the problem of imbalance of samples of different surrounding rock categories.

[0120] The method for adjusting the probability threshold of the above prediction model includes the following steps.

[0121] Step 4A-1, calculate the maximum ratio of sample categories: calculate the maximum number of samples of different surrounding rock categories in the training set T max and the minimum number of class samples T min The specific calculation formula is:

[0122]

[0123] In this embodiment, the maximum number of samples T of different surrounding rock categories in the training set is calculated. max =4989 and the minimum number of class samples T min =848, the ratio T is:

[0124]

[0125] Step 4A-2, assessment of surrounding rock category sample imbalance: compare the ratio T calculated in step 4A-1 with the set ratio threshold Y. When T≤Y, the surrounding rock category samples are considered balanced and the prediction model probability threshold does not need to be adjusted; otherwise, the surrounding rock category samples are considered unbalanced and proceed to step 4A-3.

[0126] In this embodiment, the ratio threshold Y is preferably set to 2, 5.88>2, that is, T>Y, and the process goes to step 4A-3.

[0127] Step 4A-3, determine the initial threshold of the prediction model probability: set the initial threshold of the prediction model probability of stable surrounding rock and unstable surrounding rock to the average value of the two

[0128] Step 4A-4, prediction model probability threshold adjustment: calculate the percentage K of stable surrounding rock and unstable surrounding rock samples in the total samples, and compare K with Compare and according to the comparison results, Make adjustments, specifically:

[0129] A. For The surrounding rock type is increased in sequence according to the initial threshold of the prediction model probability in the order of Δs, 2Δs, 3Δs, ..., RΔs; where:

[0130] B. For The surrounding rock type is reduced in sequence according to the order of Δs, 2Δs, 3Δs, ..., RΔs; where: While keeping all The added value of surrounding rock category is less than The reduction values ​​are the same for the surrounding rock categories.

[0131] In this embodiment, the stable category samples account for 0.855 of the total samples, and the unstable category samples account for 0.145 of the total samples.

[0132] Furthermore, preferably, Δs=0.1, then the increased or decreased Δs, 2Δs, 3Δs, ..., RΔs are 0.2, 0.3, 0.4, 0.5, respectively.

[0133] Step 4A-5, F1 value evaluation: For each increase or decrease in the prediction model probability threshold, an F1 value evaluation is performed, and a curve of the F1 value changing with the prediction model probability threshold is constructed; the F1 value is the harmonic mean score of the model precision and recall, which is the prior art.

[0134] Step 4A-6, selecting the optimal prediction model probability threshold: the prediction model probability threshold corresponding to the maximum F1 value in the curve of the change of F1 value with the prediction model probability threshold is used as the optimal prediction model probability threshold of the primary surrounding rock classification model.

[0135] In order to further demonstrate the impact of different prediction probability thresholds on model performance, the curves of model prediction accuracy, precision, recall and F1 value with different probability thresholds under unstable state are obtained as follows: Figure 4 As shown in the figure, the results show that as the judgment threshold increases, the precision of the unstable state gradually increases and the recall rate decreases, which indicates that the increase in the threshold will increase the credibility of the prediction results but also increase the false negative rate. In addition, the accuracy, F1 and weighted F1 values ​​increase first and then decrease with the increase of the threshold, and they show the maximum value when the threshold is 0.4. This shows that under unbalanced samples, the prediction performance can be improved by adjusting the probability threshold. Therefore, adjusting the probability threshold is an effective way to eliminate sample imbalance, and the prediction model is predicted on the test set using the optimal probability threshold.

[0136] Therefore, in this embodiment, the probability threshold of the optimal prediction model of the primary surrounding rock classification model is 0.4.

[0137] Step 5: Test the surrounding rock classification model: Use the penetration index vector of each sample data in the test set obtained in step 3 The surrounding rock classification model trained in step 4 is tested to obtain a surrounding rock classification prediction result.

[0138] Confusion matrix calculation

[0139] The optimal threshold parameters are used to evaluate the performance of the surrounding rock classification prediction model on the test set. The average predicted probability of the sample under different surrounding rock classifications in the decision leaf nodes under each decision tree is used as the predicted probability, and the category corresponding to the maximum probability is used as the prediction result. The confusion matrix describing the accuracy of the test set is obtained based on the predicted category and the true label:

[0140]

[0141] Where N ij It indicates the number of samples in the test set whose true category of surrounding rock is i but is predicted to be category j.

[0142] Taking the binary classification of surrounding rock as an example, unstable rock mass is a positive sample (1), and stable rock mass is a negative sample (0). The confusion matrix is ​​as follows:

[0143]

[0144] Since the random forest method based on ensemble learning only considers the digging index of the current excavation section for classification prediction, the prediction result is affected by the quality of the current excavation section data, so the credibility of the result needs to be improved. Bayesian inference based on more observation data can improve the confidence of the result based on the information of the excavated section. The process is as follows: Figure 5 shown.

[0145] Step 6, calculate the prior probability: use the sample library of the primary surrounding rock classification model constructed in step 3 to calculate the prior probability of stable surrounding rock and the prior probability of unstable surrounding rock.

[0146] In this embodiment, the total number of samples in the database is 5797, of which the number of samples in the stable section is 4949 and the number of samples in the unstable section is 848. Therefore, the prior probabilities under the two working conditions are:

[0147] P(c0)=4949 / 5797=0.85

[0148] P(c1)=848 / 5797=0.15

[0149] Step 7, calculate conditional probability: compare and count the surrounding rock classification prediction result obtained in step 5 and the actual surrounding rock classification of each sample data in the test set to obtain the conditional probability.

[0150] The above conditional probability is that the model predicts that the surrounding rock category is c i Under this condition, its true category is c j The probability of j / c i ), in this embodiment, i = 0 or 1, j = 0 or 1, then P(c j / c i ) is calculated as:

[0151]

[0152] Where N ji Indicates that the predicted surrounding rock type is c i But the actual surrounding rock type is c j The number of samples at time .

[0153] N i Indicates that the predicted surrounding rock type is c i The total number of samples under .

[0154] In this embodiment, the confusion matrix obtained in step 5 is used to calculate:

[0155] P(c1 / c0)=69 / 495=0.16

[0156]

[0157] P(c0 / c1)=20 / 85=0.24

[0158]

[0159] Step 8, tunnel excavation: The TBM constructs the tunnel to be excavated, and monitors the excavation data of the cutter head single blade thrust and single blade torque for each excavation section; assuming that the current excavation section is t, the excavation data of the t-m+1th excavation section, the t-m+2th excavation section, ..., the tth excavation section are recorded.

[0160] The calculation formula of the above m is:

[0161]

[0162] Where L is the length of the cutter shield; d is the maximum travel distance of a single excavation section.

[0163] In this embodiment, the parameter m is set based on the cutter shield length L=6m used in the tunnel excavation project and the maximum travel footage of a single excavation section of the machine d=1.8m and is calculated by the following formula:

[0164]

[0165] Step 9, calculate the penetration index of the excavation section: according to the excavation data recorded in step 8, calculate the penetration index vectors of the t-m+1th excavation section, the t-m+2th excavation section, ..., the tth excavation section.

[0166] In this embodiment, since m=4, the penetration index vectors of the t-3th excavation section, the t-2th excavation section, the t-1th excavation section and the tth excavation section are calculated.

[0167] Step 10: Primary prediction of surrounding rock classification of excavation sections: Substitute the penetration index vectors of the t-m+1th excavation section, t-m+2th excavation section, ..., tth excavation section obtained in step 9 into the primary surrounding rock classification model trained in step 4 to obtain the surrounding rock classification sequence [c t-m+1 ,c t-m+2 ,...,c t ].

[0168] In this embodiment, the surrounding rock classification sequence of the t-3 excavation section, the t-2 excavation section, the t-1 excavation section and the t excavation section

[0169] [c t-3 ,c t-2 ,c t-1 ,c t ]=[1,1,1,1].

[0170] Step 11, Bayesian posterior estimation: According to the surrounding rock classification sequence [c t-m+1 ,c t-m+2 ,...,c t ], the prior probability obtained in step 6 and the conditional probability obtained in step 7, so as to obtain the posterior probability of the unstable category of the surrounding rock of the t-th excavation section, and then obtain the surrounding rock classification of the tunnel face.

[0171] The above-mentioned Bayesian posterior estimation method includes the following steps.

[0172] Step 11-1, calculate the predicted posterior probability P(c i / c j ) t-m+1 , specifically including the following steps:

[0173] Step 11-1A, determine the prior probability of the t-m+1th excavation section: The prior probability of the t-m+1th excavation section includes t -m+1 The prior probability P(c i )t-m+1 and c t-m+1 Prior probability of the corresponding type of surrounding rock Among them, P(c i ) t-m+1 Find it from step 6;

[0174] In this embodiment, due to the t-3 excavation section and c t-3 =1 The prior probability of the opposite type of surrounding rock, that is, stable surrounding rock, is as follows:

[0175] P(c i ) t-3 =P(c0)=4949 / 5797=0.85

[0176] Step 11-1B, select the conditional probability of the t-m+1th excavation section: The prior probability of the t-m+1th excavation section includes P(c j / c i )and in, For c t-m+1 The true category under the opposite type of surrounding rock conditions is c j The probabilities of are obtained from step 7.

[0177] In this embodiment, the prior probability of the t-3th excavation section includes:

[0178] P(c1 / c0)=69 / 495=0.16

[0179]

[0180] Step 11-1C, calculate P(c i / c j ) t-m+1 :P(c i / c j ) t-m+1 is the same as c in the t-m+1 excavation section t-m+1 The specific calculation formula for the posterior probability of the corresponding surrounding rock is:

[0181]

[0182] In this embodiment, the t-3 excavation section and c t-3 =1 The posterior probability of the corresponding surrounding rock is:

[0183]

[0184] Step 11-2, calculate the prior probability of the t-m+2th excavation section: The prior probability of the t-m+2th excavation section includes t -m+2The prior probability P(c i ) t-m+2 and c t-m+2 Prior probability of opposite type of surrounding rock but:

[0185] P(c i ) t-m+2 =P(c i / c j ) t-m+1

[0186]

[0187] In this embodiment, the t-3 excavation section and c t-2 =1The prior probability of the opposite type of surrounding rock, that is, stable surrounding rock, is:

[0188] Penetration index P(c i ) t-2 =P(c0 / c1) t-3 =0.544

[0189]

[0190] Step 11-3: Repeat steps 11-1B to 11-1C to calculate the t-m+2th excavation section and c t-m+2 The posterior probability P(c i / c j ) k-m+2 .

[0191] The posterior probability based on the k-3 prediction result being 1:

[0192]

[0193] P(c0) t-1 =P(c0 / c1) t-2 =0.20

[0194] Step 11-4, repeat steps 11-2 to 11-3, and calculate the posterior probability P(c i / c j ) k-m+3 ,……,P(c i / c j ) t .

[0195] Based on the posterior probability when the k-2th prediction result is 1:

[0196]

[0197] P(c0) t =P(c0 / c1) t-1 =0.05

[0198] Based on the posterior probability when the k-1th prediction result is 1:

[0199]

[0200] Step 11-5: Secondary determination of surrounding rock classification: Based on the P(c i / c j ) t , the surrounding rock category with the maximum predicted probability is taken as the surrounding rock category of the tunnel face in front of the current t-th excavation section.

[0201] Based on the above process, the posterior probability of the stable category based on the four unstable prediction results [1,1,1,1] is 0.01, and the posterior probability of the unstable category based on the four unstable prediction results [1,1,1,1] is 0.99. The probability of the unstable section based on the excavated section is 0.99 using the Bayesian enhancement method.

[0202] The above Bayesian enhanced ensemble learning method was applied to the No. 3 tunnel of the No. 2 section of the Yinchuojiliao project for verification. A length range of about 50m with a pile number range of 26+030 to 25+980 was selected for testing. The rock mass classification in this section was evaluated as Class V, which is unstable rock mass. The prediction probabilities were obtained by using random forest and Bayesian enhanced methods respectively. Figure 6 and Figure 7 As shown, it can be seen that the accuracy of the Bayesian enhancement method is 92%, which is 3.37% higher than the random forest algorithm.

[0203] The present invention can quickly extract the penetration index that reflects the rock-machine interaction process from the rock-breaking parameters monitored during the construction of the tunnel boring machine, realize the intelligent perception of surrounding rock classification, and reduce the time and economic cost of geological exploration. The Bayesian theory is used to enhance the intelligent prediction model, which can consider more observation data and probability results of the excavated section to predict the current excavation section, which can effectively eliminate the impact of the quality of monitoring data on the prediction results and further improve the credibility of the surrounding rock classification results. This method can greatly improve the automation and intelligence of geological condition perception during the mechanized construction of tunnel engineering, and provide guidance for the assessment of surrounding rock stability and the selection of support methods.

[0204] The preferred embodiments of the present invention are described in detail above; however, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all belong to the protection scope of the present invention.

Claims

1. A TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement, characterized by: The steps include: Step 1: Classification of surrounding rocks: Classify surrounding rocks into P types according to rock strength, integrity, structural surface state, occurrence and groundwater conditions; then, mark each type of surrounding rock with a digital classification label C; Step 2: Construct a primary surrounding rock classification model based on penetration index: A random forest algorithm is used to construct a primary surrounding rock classification model based on penetration index for each excavation section during the TBM construction process; the input layer of the primary surrounding rock classification model is the penetration index vector of the excavation section during the TBM construction process. The output layer is the surrounding rock classification label C of the corresponding excavation section; where a is the linear fitting slope of the cutterhead single blade thrust and penetration in the excavation section; b is the initial penetration thrust of the cutterhead roller in the excavation section; It indicates the linear fitting goodness of the thrust and penetration of the cutterhead in the excavation section; TPI is the linear fitting slope of the torque and penetration of the cutterhead in the excavation section; Indicates the linear fit goodness of the single-blade torque and penetration in the excavation section; Step 3: Construct a surrounding rock classification model sample library, which specifically includes the following steps: Step 3-1, collect effective excavation section data: collect effective excavation section data of all types of surrounding rock in step 1; wherein the number of effective excavation sections of each type of surrounding rock is not less than M; M ≥ 100; Step 3-2, extracting penetration index of tunneling section: extract the penetration index vector for each tunneling section of each type of surrounding rock, and record the actual surrounding rock classification label C corresponding to each tunneling section, so as to obtain a surrounding rock classification model sample library containing N sample data; each sample data includes Step 3-3, sample library classification: divide the sample library into two parts: training set and test set; Step 4, training a surrounding rock classification model: using the training set obtained in step 3, training the surrounding rock classification model constructed in step 2 to obtain a trained surrounding rock classification model; Step 5: Test the surrounding rock classification model: Use the penetration index vector of each sample data in the test set obtained in step 3 The primary surrounding rock classification model trained in step 4 is tested to obtain a primary surrounding rock classification prediction result; Step 6, calculating the prior probability: using the sample library of the primary surrounding rock classification model constructed in step 3, calculating the prior probability of stable surrounding rock and the prior probability of unstable surrounding rock; Step 7, calculate conditional probability: compare and count the surrounding rock classification prediction result obtained in step 5 and the actual surrounding rock classification of each sample data in the test set to obtain the conditional probability; Step 8, tunnel excavation: The TBM constructs the tunnel to be excavated, and monitors the excavation data of the cutter head single blade thrust and single blade torque for each excavation section; assuming that the current excavation section is t, the excavation data of the t-m+1th excavation section, the t-m+2th excavation section, ..., the tth excavation section are recorded; Step 9, calculating the penetration index of the excavation section: according to the excavation data recorded in step 8, calculating the penetration index vectors of the t-m+1th excavation section, the t-m+2th excavation section, ..., the tth excavation section; Step 10: Primary prediction of surrounding rock classification of excavation sections: Substitute the penetration index vectors of the t-m+1th excavation section, t-m+2th excavation section, ..., tth excavation section obtained in step 9 into the primary surrounding rock classification model trained in step 4 to obtain the surrounding rock classification sequence [c t-m+1 ,c t-m+2 ,...,c t ]; Step 11, Bayesian posterior estimation: According to the surrounding rock classification sequence [c t-m+1 ,c t-m+2 ,...,c t ], the prior probability obtained in step 6 and the conditional probability obtained in step 7, so as to obtain the posterior probability of the unstable category of the surrounding rock of the t-th excavation section, and then obtain the surrounding rock classification of the tunnel face.

2. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 1 is characterized by: In step 1, the surrounding rock is divided into C = 2 types, namely stable surrounding rock and unstable surrounding rock; according to the national standard "Specification for Geological Survey of Water Conservancy and Hydropower Engineering GB50487-2008", the surrounding rock is divided into N = 5 types, namely Class I, II, III, IV and V surrounding rock; among them, Class I, II and III surrounding rock are stable types of surrounding rock, and Class IV and V surrounding rock are unstable types of surrounding rock.

3. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 2 is characterized by: In step 1, the classification label of stable surrounding rock is marked as 0, and the classification label of unstable surrounding rock is marked as 1.

4. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 1 is characterized by: In step 8, the calculation formula for m is: Where L is the length of the cutter shield; d is the maximum travel distance of a single excavation section.

5. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 1 is characterized by: In step 7, the conditional probability is that the model predicts that the surrounding rock category is c i Under this condition, its true category is c j The probability of j / c i ), then P(c j / c i ) is calculated as: Where N ji Indicates that the predicted surrounding rock type is c i But the actual surrounding rock type is c j The number of samples at time ; N i Indicates that the predicted surrounding rock type is c i The total number of samples under .

6. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 5 is characterized by: In step 11, the Bayesian posterior estimation method includes the following steps: Step 11-1, calculate the predicted posterior probability P(c i / c j ) t-m+1 , specifically including the following steps: Step 11-1A, determine the prior probability of the t-m+1th excavation section: The prior probability of the t-m+1th excavation section includes t-m+1 The prior probability P(c i ) t-m+1 and c t-m+1 Prior probability of the corresponding surrounding rock Among them, P(c i ) t-m+1 Find it from step 6; Step 11-1B, select the conditional probability of the t-m+1th excavation section: The prior probability of the t-m+1th excavation section includes P(c j / c i )and in, For c t-m+1 The true category under the opposite type of surrounding rock conditions is c j The probabilities of are all found from step 7; Step 11-1C, calculate P(c i / c j ) t-m+1 :P(c i / c j ) t-m+1 is the same as c in the t-m+1 excavation section t-m+1 The specific calculation formula for the posterior probability of the corresponding surrounding rock is: Step 11-2, calculate the prior probability of the t-m+2th excavation section: The prior probability of the t-m+2th excavation section includes t-m+2 The prior probability P(c i ) t-m+2 and c t-m+2 Prior probability of opposite type of surrounding rock but: P(c i ) t-m+2 =P(c i / c j ) t-m+1 Step 11-3: Repeat steps 11-1B to 11-1C to calculate the t-m+2th excavation section and c t-m+2 The posterior probability P(c i / c j ) k-m+2 ; Step 11-4, repeat steps 11-2 to 11-3, and calculate the posterior probability P(c i / c j ) k-m+3 ,……,P(c i / c j ) t ; Step 11-5: Secondary determination of surrounding rock classification: Based on the P(c i / c j ) t , the surrounding rock category with the maximum predicted probability is taken as the surrounding rock category of the tunnel face in front of the current t-th excavation section.

7. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 1 is characterized by: In step 4, the best hyperparameter combination of random forest is selected based on the cross-validation method of grid search.

8. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 7 is characterized by: In step 4, the method of selecting the best hyperparameter combination of random forest by cross-validation method includes the following steps: step 4-1, designing hyperparameter combination: for random forest, designing h hyperparameter combinations; Step 4-2: Divide the training subsets: Divide the total training set into n training subsets D by bootstrapping method i ; Step 4-3, build a random forest: for each training subset D i Independent surrounding rock classification prediction decision trees are constructed, and a total of n surrounding rock classification prediction decision trees are obtained to form a random forest; Step 4-4, hyperparameter training: Using the random forest built in step 4-3, each combination of hyperparameters designed in step 4-1 is trained; during the training process, the generation process of each surrounding rock classification prediction decision tree includes the following steps: Step 4-4A, calculate the node Gini index: when each node is split, calculate its node Gini index; the calculation formula of the Gini index G(w) of node w is: In the formula, p c is the probability that the current node w belongs to the surrounding rock category c; Step 4-4B, calculate the best split point: For each node, calculate each feature f j All possible split points of f, find a split point that can maximize the information gain; where f j ∈[a,b,TPI], then the calculation formula for the optimal split point s is: Where D t is the sample set of the current node, and According to f j The left sample and the right sample after the feature split point s segmentation; G(D t ), and Node D t Node D t and nodes The Gini index of Step 4-4C, recursive splitting: repeating steps 4-4A to 4-4B, selecting the best splitting nodes for the left node and the right node respectively for recursive splitting, until the stop growth condition of the surrounding rock classification prediction decision tree is reached; Step 4-5, select the optimal hyperparameters: Calculate the average accuracy of each combination of hyperparameters during validation, and select the set of hyperparameters with the highest average accuracy as the optimal hyperparameters for random forest.

9. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 1 is characterized by: In step 4, when the primary surrounding rock classification model is trained to output the surrounding rock classification label of the excavation section, a prediction model probability threshold is set. The primary surrounding rock classification model compares the probability of each surrounding rock classification with the prediction model probability threshold, and takes the surrounding rock classification greater than the prediction model probability threshold as the output of the primary surrounding rock classification model; wherein the prediction model probability threshold can be adjusted to eliminate the problem of imbalance of samples of different surrounding rock categories.

10. The TBM surrounding rock classification intelligent identification method based on penetration index and Bayesian enhancement according to claim 9 is characterized by: In step 4, the method for adjusting the probability threshold of the prediction model includes the following steps: Step 4A-1, calculate the maximum ratio of sample categories: calculate the maximum number of samples of different surrounding rock categories in the training set T max and the minimum number of class samples T min The specific calculation formula is: Step 4A-2, assessment of surrounding rock category sample imbalance: compare the ratio T calculated in step 4A-1 with the set ratio threshold Y. When T≤Y, the surrounding rock category samples are considered balanced and the prediction model probability threshold does not need to be adjusted; otherwise, the surrounding rock category samples are considered unbalanced and proceed to step 4A-3; Step 4A-3, determine the initial threshold of the prediction model probability: set the initial threshold of the prediction model probability of each surrounding rock to the average value of different categories Step 4A-4, prediction model probability threshold adjustment: calculate the percentage K of stable surrounding rock and unstable surrounding rock samples in the total samples, and compare K with Compare and according to the comparison results, Make adjustments, specifically: A. For The surrounding rock type is increased in sequence according to the initial threshold of the prediction model probability in the order of Δs, 2Δs, 3Δs, ..., RΔs; where: B. For The surrounding rock type is reduced in sequence according to the order of Δs, 2Δs, 3Δs, ..., RΔs; where: While keeping all The added value of surrounding rock category is less than The reduction values ​​for surrounding rock categories are the same; Step 4A-5, F1 value evaluation: perform an F1 value evaluation for each increase or decrease in the prediction model probability threshold, and construct a curve of the change of the F1 value with the prediction model probability threshold; Step 4A-6, selecting the optimal prediction model probability threshold: the prediction model probability threshold corresponding to the maximum F1 value in the curve of the change of F1 value with the prediction model probability threshold is used as the optimal prediction model probability threshold of the primary surrounding rock classification model.

Images