Full-face tunneling machine tunneling roadway surrounding rock strength prediction model training method and device

Abstract

The application provides a full-face tunneling machine tunneling roadway surrounding rock strength prediction model training method and related equipment, which can train a model capable of effectively predicting the point load strength of a coal mine roadway TBM tunneling roadway surrounding rock. The method comprises the following steps: obtaining first original data corresponding to a target tunneling machine; removing data in which the target tunneling machine is in a shutdown section and an empty pushing section from the first original data to obtain second original data; processing abnormal values in the second original data to obtain target operation data; extracting feature parameters from the target operation data; normalizing the feature parameters to obtain a model training data set; constructing a Stacking algorithm model according to a base learner and a meta learner, wherein the base learner comprises a plurality of regression trees, and the output result of the base learner is used as the input of the meta learner; and training the Stacking algorithm model based on the model training data set to obtain a full-face tunneling machine tunneling roadway surrounding rock strength prediction model.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of tunneling, in particular to a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model training method and equipment. BACKGROUND

[0002] Tunnel Boring Machine (TBM) tunneling is an efficient, safe and environmentally friendly roadway construction method. TBM construction is sensitive to geological conditions, and the point load strength of surrounding rock, as a key indicator for measuring the drivability of surrounding rock and determining roadway support parameters, has important practical significance for optimizing tunneling parameters and guiding roadway support design. However, with the increasing complexity of underground strata environment, current surrounding rock strength information is mainly obtained through static rock mechanics experiments and geological exploration, which has obvious hysteresis and cannot be obtained in real time. Therefore, there is an urgent need for a TBM tunneling roadway rock mass point load strength prediction technology. This technology will be able to quickly and real-time identify rock mass strength and provide accurate rock mass strength information for underground tunneling operations, effectively dealing with the working environment of TBM in complex strata conditions.

[0003] In recent years, with the explosive growth of data and the development of artificial intelligence, various machine learning methods have been gradually introduced into the field of rock mass perception. Mainly from two directions of image and tunneling parameter analysis to establish a surrounding rock perception model. In terms of images, features of rock mass or hyperspectral images are extracted and segmented through deep learning to realize the identification of rock mass state. Although the image method has achieved good results in rock mass identification, due to the complex working conditions of coal mine roadway TBM tunneling, it is difficult to collect clear rock mass pictures, so it is not suitable for rock mass state perception in coal mine roadway. SUMMARY

[0004] The embodiment of the present application provides a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model training method and equipment, which can train a model that can effectively predict the point load strength of the surrounding rock of the coal mine roadway TBM tunneling roadway.

[0005] The first aspect of the present application provides a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model training method, comprising:

[0006] Obtaining first original data corresponding to the target tunneling machine, the first original data including tunneling machine running state data, cutterhead vibration acceleration data and surrounding rock point load strength data;

[0007] Eliminating the data of the target tunneling machine in the stop section and the empty pushing section in the first original data to obtain second original data;

[0008] Processing the abnormal values in the second original data to obtain target running data;

[0009] extracting feature parameters in the target operation data, the feature parameters including cutter thrust, cutter torque, penetration, cutter rotating speed, cutter vibration acceleration and rock mass geomechanics property parameters;

[0010] normalizing the feature parameters to obtain a model training data set;

[0011] constructing a Stacking algorithm model according to a base learner and a meta learner, the base learner including multiple regression trees, and an output result of the base learner being used as an input of the meta learner;

[0012] training the Stacking algorithm model based on the model training data set to obtain a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model.

[0013] The second aspect of the present application provides a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model training device, comprising:

[0014] an acquisition module configured to acquire first original data corresponding to a target tunnel boring machine, the first original data including tunnel boring machine operation state data, cutter vibration acceleration data and surrounding rock point load strength data;

[0015] a rejection module configured to reject data in which the target tunnel boring machine is in a shutdown section and an empty pushing section in the first original data to obtain second original data;

[0016] an outlier processing module configured to process outliers in the second original data to obtain target operation data;

[0017] a feature extraction module configured to extract feature parameters in the target operation data, the feature parameters including cutter thrust, cutter torque, penetration, cutter rotating speed, cutter vibration acceleration and rock mass geomechanics property parameters;

[0018] a normalization module configured to normalize the feature parameters to obtain a model training data set;

[0019] a construction module configured to construct a Stacking algorithm model according to a base learner and a meta learner, the base learner including multiple regression trees, and an output result of the base learner being used as an input of the meta learner;

[0020] a training module configured to train the Stacking algorithm model based on the model training data set to obtain a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model.

[0021] The third aspect of the present application provides an electronic device comprising a memory and a processor, wherein the processor is configured to execute the steps of the full-face tunnel boring machine tunneling surrounding rock strength prediction model training method according to the first aspect.

[0022] The fourth aspect of the present application provides a computer readable storage medium, which stores a computer management program, wherein the computer management program is configured to execute the steps of the full-face tunnel boring machine tunneling surrounding rock strength prediction model training method according to the first aspect when executed by a processor.

[0023] In summary, it can be seen that, in the embodiments of the present application, the related data in the tunneling process is obtained by the field collection method, the invalid data and abnormal data in the original data are removed, the input features of the model are determined based on the rock-machine interaction analysis, the gradient boosting decision tree (GBDT), random forest (RF) and support vector regression (SVR) three machine learning algorithms are used as the base learner, and the linear regression (LR) algorithm is used as the meta-learner, so as to train the prediction model based on the Stacking integrated algorithm. Compared with the prediction model of a single machine learning algorithm, the prediction model has the highest goodness of fit and the smallest error index value, has higher prediction accuracy, and can train an effective model for predicting the point load strength of the surrounding rock of the coal mine tunnel TBM. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A flowchart of the full-face tunnel boring machine tunneling surrounding rock strength prediction model training method provided by the embodiments of the present application is shown in the figure.

[0025] Figure 2 A schematic diagram of the original data provided by the embodiments of the present application is shown in the figure.

[0026] Figure 3 A schematic diagram of the complete field tunneling data provided by the embodiments of the present application is shown in the figure.

[0027] Figure 4 A schematic diagram of a group of tunneling cycle data provided by the embodiments of the present application is shown in the figure.

[0028] Figure 5 A schematic diagram of the box plot composition structure provided by the embodiments of the present application is shown in the figure.

[0029] Figure 6 A schematic diagram of the rock-machine interaction relationship provided by the embodiments of the present application is shown in the figure.

[0030] Figure 7 A schematic diagram of the input feature sampling interval provided by the embodiments of the present application is shown in the figure.

[0031] Figure 8A Stacking integrated algorithm principle schematic diagram provided for an embodiment of the present application is shown in FIG. 1.

[0032] Figure 9a A GBDT model prediction result schematic diagram provided for an embodiment of the present application is shown in FIG. 2.

[0033] Figure 9b An RF model prediction result schematic diagram provided for an embodiment of the present application is shown in FIG. 3.

[0034] Figure 9c An SVR model prediction result schematic diagram provided for an embodiment of the present application is shown in FIG. 4.

[0035] Figure 9d A Stacking model prediction result schematic diagram provided for an embodiment of the present application is shown in FIG. 5.

[0036] Figure 10 A virtual structure schematic diagram of a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model training device provided for an embodiment of the present application is shown in FIG. 6.

[0037] Figure 11 A hardware structure schematic diagram of a full-face tunnel boring machine tunneling roadway surrounding rock strength prediction model training device provided for an embodiment of the present application is shown in FIG. 7.

[0038] Figure 12 An embodiment schematic diagram of an electronic device provided for an embodiment of the present application is shown in FIG. 8.

[0039] Figure 13 An embodiment schematic diagram of a computer readable storage medium provided for an embodiment of the present application is shown in FIG. 9. DETAILED DESCRIPTION

[0040] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0041] The tunneling parameter identification method has lower requirements for the site environment and higher practicability. For example, if the TBM site tunneling data is collected, a rock mass excavability grade perception model based on random forest (RF) can be established to accurately predict the relevant characteristics of the surrounding rock. The quantum genetic algorithm (QGA) and the radial basis function neural network (RBF) are combined to realize the advanced identification of the surrounding rock category. A surrounding rock state perception model is established based on the degree neural network (DNN) with 10 tunneling parameters related to the surrounding rock conditions as input features. A long short-term memory (LSTM) model is constructed using 5 selected feature input parameters to identify the broken state of the surrounding rock. The back propagation (BP) neural network is used to establish the nonlinear relationship between the rock mass state and the input parameters. The various parameters in the TBM tunneling process are used as input features to establish a prediction model for the surrounding rock state, which has high accuracy and practicability. However, a single machine learning algorithm often cannot fully utilize the information of the data, and the prediction effect may vary greatly for different data distributions and feature combinations.

[0042] Therefore, the full-face tunneling machine tunneling surrounding rock strength prediction model training method provided by the present application fuses multiple models, combines different machine learning algorithms, and optimizes the multiple algorithms to improve the prediction accuracy of the model.

[0043] The full-face tunneling machine tunneling surrounding rock strength prediction model training method provided by the present application will be described from the perspective of a full-face tunneling machine tunneling surrounding rock strength prediction model training device. The full-face tunneling machine tunneling surrounding rock strength prediction model training device can be a server or a service unit in a server, and the specific implementation is not limited.

[0044] Please refer to Figure 1 , Figure 1 The full-face tunneling machine tunneling surrounding rock strength prediction model training method provided by the present application will be described from the perspective of a full-face tunneling machine tunneling surrounding rock strength prediction model training device. The full-face tunneling machine tunneling surrounding rock strength prediction model training device can be a server or a service unit in a server, and the specific implementation is not limited.

[0045] 101, Obtain the first original data corresponding to the target tunneling machine.

[0046] In this embodiment, the full-face tunneling machine tunneling surrounding rock strength prediction model training device can obtain the first original data corresponding to the target tunneling machine, which includes tunneling machine running state data, cutterhead vibration acceleration data, and surrounding rock point load strength data. Among them, the full-face tunneling machine tunneling surrounding rock strength prediction model training device exports data through the on-board system to obtain 164 tunneling machine running mechanical, electrical and hydraulic parameter data in real time, including tunneling machine cutterhead cutting state, tunneling machine walking state, anchor rod drilling machine drilling state and belt machine running state, and the sampling frequency is 1Hz, please refer toFigure 2 , Figure 2 The schematic diagram of the original data provided by the embodiment of the present application. Since the strength change of the surrounding rock is only closely related to the cutting state of the cutter head, the redundant data is eliminated, and the cutting state parameter data is retained. The vibration of the cutter head when breaking rock can also reflect the change of the rock mass in front to a certain extent, so the cutter head vibration acceleration is also used as the main parameter for collection, and the sampling frequency is 100 Hz. The point load test is used to collect the point load strength data of the rock mass, and Table 1 is the collection result of the point load data:

[0047] Table 1

[0048] Rock sample class Point load strength IS (MPa) Mean value (MPa) Sandy mudstone 2.17、2.05、2.32、2.45、2.34、1.93 2.21 Fine sandstone 3.22、3.65、3.78、3.41、3.88、3.04 3.50 Mudstone 0.88、0.94、0.54、1.05、0.64、0.73 0.79

[0049] 102、Eliminate the data of the target tunneling machine in the stop section and the empty pushing section in the first original data to obtain the second original data.

[0050] In this embodiment, generally speaking, the operation process of the TBM tunneling cycle can be regarded as a "tunneling-stop-tunneling" stage, and the cycle is repeated to advance forward. Therefore, based on the original data collected on site, the data of the tunneling and stopping stages are also contained (as shown in Figure 3 ). Among them, the data of the tunneling section can be divided into the empty pushing section, the rising section and the steady state section (as shown in Figure 4 ). In the empty pushing stage, the cutter head speed increases and approaches the rock mass of the working face; in the rising stage, the cutter head contacts the working face, the disc cutter gradually penetrates into the rock mass, and the torque, the pushing force and other values gradually increase to the set rock breaking parameter requirements; in the steady state stage, the cutter head rotates to break rock, and the values tend to be stable, and remain in a small range of fluctuation around a value. The key to establishing the surrounding rock strength prediction model lies in analyzing the relationship between the rock-machine interaction data generated during the contact process of the cutter head of the TBM and the coal rock mass, therefore, it is necessary to eliminate the stop section and the empty pushing section data in the original data, and retain the tunneling data of the rising stage and the steady state stage. Specifically, the first sub-data of the target tunneling machine in the stop section and the second sub-data of the target tunneling machine in the tunneling stage in the first original data can be determined by the following formula:

[0051] ;

[0052] Among them, is the cutter head thrust, is the cutter head torque, is the cutter head speed, is the discriminant function, , ;

[0053] According to the second sub-data, third sub-data in which the target boring machine is in the empty pushing stage is determined, that is, the first boring data in the second sub-data is determined, and a target time is determined based on a time of the first boring data and a preset time length. Finally, data between the time of the first boring data and the target time is determined as the third sub-data. In order to delete the empty pushing stage data in the boring stage, by comparing a large amount of collected original data, it is found that the data in this stage are all within 40s before the boring stage data (here, the preset time length is taken as 40s for example, and of course it can also be adjusted according to actual conditions, and the specific is not limited), as shown in Figure 4 Therefore, the data points within 40s before the boring stage are deleted by using the manual processing method. The first sub-data and the third sub-data are removed from the first original data, and the second original data is obtained.

[0054] 103. Process the abnormal values in the second original data to obtain target running data.

[0055] In the embodiment, in the TBM boring process, due to the influence of the poor working environment and equipment failure, there are some abnormal data in the original data collected on site. These abnormal data not only interfere with subsequent model training, leading to a decrease in model accuracy, but also increase the amount of calculation and increase the complexity of model learning. Therefore, the abnormal data needs to be removed, and specifically:

[0056] Based on the box plot, the upper and lower whiskers corresponding to the second original data are determined by the following formula:

[0057] ;

[0058] ;

[0059] wherein, is the upper whisker, is the lower whisker, is the upper quartile corresponding to the box plot, is the lower quartile corresponding to the box plot, is the lower quartile difference corresponding to the box plot, ;

[0060] The data values in the second original data that are not between the upper and lower whiskers are determined as abnormal values;

[0061] The abnormal values in the second original data are removed to obtain third running data;

[0062] The missing values in the third running data are supplemented based on the linear interpolation method to obtain target running data.

[0063] Referring to Figure 5 , Figure 5The box plot composition structure schematic diagram provided by the embodiment of the application can determine the upper whisker and the lower whisker, and determine the data value not between the upper whisker and the lower whisker in the second original data as an abnormal value. For the abnormal value in the original data, the traditional abnormal value elimination method 3σ criterion is also called the Relyda criterion, and this method needs to assume that the original data conforms to the normal distribution in advance, but the actual data often does not conform to this rule, and this method takes the mean value and the standard deviation as the abnormal value judgment standard, and is easy to be disturbed by the abnormal value. Compared with the 3σ criterion, the box plot abnormal value elimination method takes the quartile and the quartile distance as the basis for abnormal value judgment, and these statistics have a certain resistance to abnormal values, and this method does not need to assume that the data conforms to the normal distribution. Therefore, the box plot method is used to eliminate the abnormal values in the original data. The box plot is composed of three main parts: the box, the whisker and the abnormal value. The box is composed of the upper quartile, the median and the lower quartile, and the whisker is divided into the upper whisker and the lower whisker. The values not between the upper whisker and the lower whisker in the box plot are determined as abnormal values and are eliminated, and finally the missing values are supplemented by linear interpolation.

[0064] 104. Extracting the feature parameters in the target operation data.

[0065] In the embodiment, in the process of TBM tunneling, the cutter head directly contacts the rock mass of the tunneling face, and the cutting part parameters are directly affected by the state of the rock mass in front. With the progress of the cutting process, the working face forms a circular broken section under the continuous action of the rock breaking of the rolling cutter. When the state of the cutting surrounding rock changes, the existence of the broken section of the surrounding rock not only affects the cutting torque T and the rotating speed RPM of the working face, but also causes the dynamic change of the direction of the reaction force of the working face on the rolling cutter. Therefore, the cutting thrust F of the cutter head should be included in the category of key parameters of main inputs. With the continuous rotation of the cutter head, the rolling cutter periodically breaks the rock. In a rock breaking period, the maximum impact state occurs when the rolling cutter contacts the rock mass, at which time the rock mass is compressed, causing the load of the cutting part to rise rapidly. When the stress reaches its own material limit, the rock mass breaks, and the cutting part is unloaded immediately. The continuous step-by-step breaking process causes the acceleration to change rapidly, and the vibration signal A and the penetration depth signal REV generated thereby should also be included in the category of key parameters reflecting the state of the rock mass. Therefore, the cutter thrust F, the cutter torque T, the penetration depth REV, the cutter rotating speed RPM and the cutter vibration acceleration A are determined as the main inputs of the model. The rock-machine interaction relationship is as shown in Figure 6

[0066] For the steady state section of the tunneling cycle, the reflection of the TBM equipment state parameters on the geological conditions of the working face mainly embodies the mean value and the fluctuation amplitude of the tunneling parameters in the interval. From Figure 7 ​It can be seen that in the steady state stage, the tunneling parameters fluctuate up and down around a certain value, and the data in the first 50s can basically reflect the law and characteristics of the subsequent changes. Considering that in the actual excavation process of the tunnel, it is more beneficial to identify the strength value of the surrounding rock mass in front as soon as possible for the tunneling process. Therefore, in the present application, the mean value (F m , T m , Rev m , RPM m , A m ) and the standard deviation (F n , T n , Rev n , RPM n , A n ) of the data of the first 50s in the steady state stage of the TBM tunneling are selected as the input parameters of the model, and finally the input parameters include 10 characteristic values.

[0067] 105. Normalize the characteristic parameters to obtain a model training data set.

[0068] In this embodiment, the tunneling rock strength prediction model training device of the full-face tunneling machine can normalize the characteristic parameters (the purpose of the normalization is to eliminate the dimensional and scale differences of the data and improve the stability and accuracy of the model) to obtain a model training data set. Specifically, the characteristic parameters can be normalized to obtain a model training data set by the following formula:

[0069]

[0070] wherein, is the normalized characteristic parameter, is the characteristic parameter, and are the maximum and minimum values of the characteristic parameters, respectively.

[0071] It should be noted that in this embodiment, the tunneling cycle corresponding to the target operation data has 524 cycles, and the data preprocessing method is used to process the five tunneling state parameters of the tunneling machine (cutter thrust F, cutter torque T, penetration REV, cutter speed RPM, and cutter vibration acceleration A) and one rock mass geomechanical property parameter (point load strength value IS) to form a rock-machine interaction database containing mean value and standard deviation, a total of 524 data amounts, providing a data basis for subsequent algorithm model learning and verification.

[0072] 106. Construct a Stacking algorithm model according to the base learner and the meta-learner.

[0073] In this embodiment, a Stacking ensemble algorithm is used to construct the model. The Stacking ensemble algorithm is a hierarchical algorithm that takes the outputs of multiple different base learners in the first layer as new input features, and then uses a meta-learner in the second layer to learn and integrate, thereby optimizing and integrating the base learners and improving the prediction accuracy of the prediction model. The principle of the Stacking ensemble algorithm is as shown in Figure 8

[0074] The base learners and meta-learners are important components of the Stacking ensemble algorithm, and they directly affect the performance and effect of the ensemble model. Different types of base learners should be selected for the first layer of the Stacking algorithm model to increase the diversity of the model. A simple model should be selected for the second layer meta-learner to prevent overfitting. Since the coal rock mass geomechanical parameters are continuous variables, the above mutual feedback relationship is a regression problem between multiple independent variables and a dependent variable. Therefore, in this application, a variety of machine learning algorithms that can be used for regression analysis are used as base learners in the first layer of the Stacking model, including GBDT (Gradient Boosting Decision Tree), RF (Random Forest), and SVR (Support Vector Regression). GBDT (Gradient Boosting Decision Tree) is an ensemble learning algorithm based on boosting algorithm, which uses multiple regression trees as base learners and uses a series of optimization techniques to speed up the training of the model and improve the accuracy of the model. RF (Random Forest) algorithm is an ensemble learning method composed of multiple decision trees, which votes or averages the prediction results of all decision trees to improve the accuracy of the prediction model. SVR (Support Vector Regression) is a supervised learning method for regression problems, which can avoid overfitting and better handle nonlinear regression problems. These base learners have different learning abilities and characteristics, and can learn and analyze data from different angles and levels, thereby improving the diversity and complementarity of the ensemble model. In the second layer, LR (Linear Regression) is selected as the meta-learner. Linear regression model is a simple and effective model used to fit the linear relationship between features and labels in the training set, avoiding overfitting.

[0075] 107. Train the Stacking algorithm model based on the model training data set to obtain the full-face tunneling machine tunneling roadway surrounding rock strength prediction model.

[0076] In this embodiment, after the Stacking algorithm model is constructed, the full-face tunneling machine tunneling roadway surrounding rock strength prediction model training device can train the Stacking algorithm model based on the model training data set to obtain the full-face tunneling machine tunneling roadway surrounding rock strength prediction model.

[0077] ​In one embodiment, the Stacking algorithm model is trained based on the model training data set to obtain the full-face tunneling machine tunneling roadway surrounding rock strength prediction model, which comprises the following steps:

[0078] The model training data is divided into a training set and a test set;

[0079] The training set is divided into multiple training subsets;

[0080] The first training subset is determined as a validation set, and the other training subsets in the multiple training subsets except the first training subset are used as a training set. Each regression tree in the multiple regression trees is trained by K-fold cross-validation to obtain multiple first prediction results corresponding to each regression tree.

[0081] Each regression tree is tested based on the corresponding test data in the test set to obtain a second prediction result corresponding to each regression tree;

[0082] The multiple first prediction results are input into a meta-learner for training, and the trained meta-learner is tested based on the second prediction results to obtain the full-face tunneling machine tunneling roadway surrounding rock strength prediction model.

[0083] In this embodiment, the model training data set (524 groups of data) is divided into a training set and a test set according to a ratio of 5:1, obtaining 437 groups of training data and 87 groups of test data. The training data is used to train the Stacking integrated model, and the test data is used to verify the prediction effect of the model. In order to verify that the Stacking integrated model has good prediction effect on point load strength of different lithology, the data amount of sandstone, fine sandstone and mudstone is ensured to be consistent, i.e., 29 groups. In this way, the prediction deviation of the model caused by uneven distribution of lithology can be avoided.

[0084] The full-face tunneling machine tunneling roadway surrounding rock strength prediction model training device can divide the training set into multiple training subsets (for example, 4 or 5, which is not limited, and 4 is taken as an example for description), train1, train2, train3, and train4. The base learner is selected, and three regression trees of GBDT (gradient boosting decision tree), RF (random forest) and SVR (support vector regression algorithm) are selected as the base learner.

[0085] For example, GBDT: train1, train2, train3 and train4 are used as the validation set in turn, and the remaining 3 are used as the training set, and k-fold cross-validation is performed for model training; then prediction is performed on the test set. In this way, 4 prediction results trained by the GBDT algorithm on the training set and 1 prediction value B1 on the test set are obtained, and the 4 prediction results are vertically overlapped and combined to obtain A1. The above is described by taking GBDT as an example, and RF algorithm and SVR algorithm are similar to GBDT, and details are not repeated here.

[0086] After the three regression trees in the base learner are trained, the prediction values of the three regression trees on the training set are used as 3 features A1, A2 and A3 respectively, and the meta-learner (here, the LR model is used as the meta-learner) is used for training to establish the LR model. Using the trained LR model, the prediction is performed on the values of the three features (B1, B2, B3) constructed by the prediction values of the three base models on the test set, and the final prediction category or probability is obtained.

[0087] It should be noted that in machine learning research, the performance of the model is often affected by the hyperparameters, so in the present application, the enumeration method is used to test and optimize the hyperparameters of the Stacking algorithm model multiple times, and finally the hyperparameters of the Stacking algorithm model are determined. At the same time, in order to evaluate the performance and effect of the prediction model, four evaluation indexes, mean square error (MSE), mean absolute error (MAE), mean absolute percentage error (MAPE), and goodness of fit (R 2 ) are used as the model performance evaluation standard. The calculation formula of these evaluation indexes is as follows:

[0088] ;

[0089] ;

[0090] ;

[0091] .

[0092] It should be noted that in order to evaluate the prediction effect of the whole-face tunneling machine tunneling surrounding rock strength prediction model, three control models are established by using single GBDT, RF and SVR respectively, and the training process of the model and the data used are completely the same, wherein the hyperparameters of the model are set by using the enumeration method to take multiple values, and the optimal value is selected. The prediction accuracy of different models is discussed by using the evaluation standard, and the prediction effect of the model is evaluated.

[0093] The comparison between the predicted results of the surrounding rock point load strength of the stacking integrated model (i.e., the full-face tunnel boring machine tunneling surrounding rock strength prediction model) and the other three control models on 87 test samples and the actual values is shown in FIG. 8. Figure 9a-9d As can be seen from the figure, the different models have good fitting degree and explanation degree for the prediction of the surrounding rock point load strength, among which the prediction effect of the stacking model is the best, the error between the prediction results of the model and the true values is small on most of the test samples, and only a few individual samples have errors with the true values. The GBDT model and the RF model are the second, and the prediction results of the two models are also close to the true values, but there is a large deviation on some samples. The effect of the SVR model is the worst, and the difference between the prediction results of the model and the true values is large, indicating that the prediction ability of the model for the surrounding rock point load strength is weak.

[0094] The prediction results of the models are calculated by using the evaluation indexes, and the calculation results are shown in Table 2. As can be seen from the data in the table, the fitting degree R2 value of the stacking integrated model on the test set is the largest, which is 0.939, which is improved by 0.03, 0.092 and 0.148 compared with 0.909, 0.847 and 0.791 of the GBDT, RF and SVR three single models respectively, and the values of the stacking integrated model in the remaining MSE, MAE and MAPE indexes are the smallest. For a model, the closer the R2 value is to 1, the smaller the MSE, MAE and MAPE are, indicating that the prediction ability of the model is better. Therefore, it is indicated that the model has higher prediction accuracy, and the stacking integrated model is used to predict the point load strength of the surrounding rock in the TBM tunneling process of the coal mine roadway, which can better establish the mutual feedback relationship model representing the rock-machine interaction in the tunneling process, and also provides a basis for the tunneling control and the adjustment of the roadway support parameters.

[0095] Table 2

[0096] Model name MSE MAE R2 MAPE GBTD 0.105 0.252 0.909 0.167 RF 0.177 0.304 0.847 0.204 SVR 0.241 0.383 0.791 0.282 Stacking 0.071 0.186 0.939 0.105

[0097] In summary, it can be seen that in the embodiments provided by the present application, the related data in the tunneling process is obtained by the field acquisition method, and the invalid data and abnormal data in the original data are removed; based on the rock-machine interaction analysis, the input features of the model are determined, three machine learning algorithms of gradient boosting decision tree (GBDT), random forest (RF) and support vector regression (SVR) are used as base learners, and linear regression (LR) algorithm is used as meta-learner, and a prediction model based on the stacking integrated algorithm is trained. Compared with the prediction model of single machine learning algorithm, the fitting degree is the highest, the error index value is the smallest, and the prediction accuracy is higher, which can effectively predict the point load strength of the surrounding rock of the TBM tunneling roadway of the coal mine roadway.

[0098] The full-face tunneling machine tunneling roadway surrounding rock strength prediction model training method is described above, and the full-face tunneling machine tunneling roadway surrounding rock strength prediction model training device is described below.

[0099] Please refer to Figure 10 The virtual structure schematic diagram of the full-face tunneling machine tunneling roadway surrounding rock strength prediction model training device in the embodiment of the application is shown in the figure. The full-face tunneling machine tunneling roadway surrounding rock strength prediction model training device 1000 comprises:

[0100] The acquisition module 1001 is configured to acquire first original data corresponding to a target tunneling machine. The first original data comprises tunneling machine running state data, cutterhead vibration acceleration data, and surrounding rock point load strength data.

[0101] The elimination module 1002 is configured to eliminate data of a shutdown section and an empty pushing section of the target tunneling machine in the first original data to obtain second original data.

[0102] The outlier processing module 1003 is configured to process outliers in the second original data to obtain target running data.

[0103] The feature extraction module 1004 is configured to extract feature parameters in the target running data. The feature parameters comprise cutterhead thrust, cutterhead torque, penetration, cutterhead rotation speed, cutterhead vibration acceleration, and rock mass geomechanics property parameters.

[0104] The normalization module 1005 is configured to perform normalization processing on the feature parameters to obtain a model training data set.

[0105] The construction module 1006 is configured to construct a Stacking algorithm model according to a base learner and a meta learner. The base learner comprises a plurality of regression trees, and an output result of the base learner is used as an input of the meta learner.

[0106] The training module 1007 is configured to train the Stacking algorithm model based on the model training data set to obtain a full-face tunneling machine tunneling roadway surrounding rock strength prediction model.

[0107] In one possible design, the elimination module 1002 is specifically configured to:

[0108] The first sub-data of the target tunneling machine in the shutdown section and the second sub-data of the target tunneling machine in the tunneling section in the first original data are determined by the following formula:

[0109] ;

[0110] wherein, is a thrust of the cutter head, is a torque of the cutter head, is a rotation speed of the cutter head, is a discriminant function, , ;

[0111] determining third sub-data according to the second sub-data, the third sub-data indicating that the target tunneling machine is in an empty thrust stage;

[0112] eliminating the first sub-data and the third sub-data from the first original data to obtain second original data.

[0113] In a possible design, the elimination module 1002 is further specifically configured to:

[0114] determining first tunneling data in the second sub-data;

[0115] determining a target time based on a time of the first tunneling data and a preset time length;

[0116] determining data between the time of the first tunneling data and the target time as the third sub-data.

[0117] In a possible design, the outlier processing module 1003 is specifically configured to:

[0118] determining upper and lower whiskers corresponding to the second original data based on a box plot and by using the following formula:

[0119] ;

[0120] ;

[0121] wherein, is the upper whisker, is the lower whisker, is an upper quartile corresponding to the box plot, is a lower quartile corresponding to the box plot, is a lower quartile deviation corresponding to the box plot, ;

[0122] determining, as the outliers, data values in the second original data that are not between the upper whisker and the lower whisker;

[0123] eliminating the outliers in the second original data to obtain third running data;

[0124] complementing missing values in the third running data based on a linear interpolation manner to obtain the target running data.

[0125] In a possible design, the model training module 1007 is specifically configured to:

[0126] divide the model training data into a training set and a test set;

[0127] divide the training set into multiple training subsets;

[0128] determine a first training subset as a validation set, and determine other training subsets except the first training subset in the multiple training subsets as training sets, train each regression tree in the multiple regression trees in a manner of K-fold cross validation to obtain multiple first prediction results corresponding to each regression tree;

[0129] test each regression tree corresponding to each test data in the test set based on the training of the regression tree, to obtain a second prediction result corresponding to each regression tree;

[0130] input the multiple first prediction results into the meta-learner to perform training, and test the meta-learner after training based on the second prediction result, to obtain the full-face tunnel boring machine tunneling surrounding rock strength prediction model.

[0131] In a possible design, the normalization module 1005 is specifically configured to:

[0132] normalize the feature parameter by using the following formula:

[0133]

[0134] wherein, is the feature parameter after normalization, is the feature parameter, and are the maximum value and the minimum value in the feature parameter, respectively.

[0135] The above Figure 10 The full-face tunnel boring machine tunneling surrounding rock strength prediction model training apparatus in the embodiment of the present application is described from the perspective of a modular functional entity, and the full-face tunnel boring machine tunneling surrounding rock strength prediction model training apparatus in the embodiment of the present application is described in detail from the perspective of hardware processing. Please refer to Figure 10 FIG. 1, which is an embodiment schematic diagram of the full-face tunnel boring machine tunneling surrounding rock strength prediction model training apparatus 1000 in the embodiment of the present application. The full-face tunnel boring machine tunneling surrounding rock strength prediction model training apparatus 1000 includes:

[0136] The input device 1001, the output device 1002, the processor 1003 and the memory 1004 (wherein the number of the processor 1003 can be one or more, Figure 10 In some embodiments of the present application, the input device 1001, the output device 1002, the processor 1003 and the memory 1004 can be connected through a communication bus or other means, wherein, Figure 10 In some embodiments of the present application, the input device 1001, the output device 1002, the processor 1003 and the memory 1004 can be connected through a communication bus or other means, wherein,

[0137] The processor 1003 is configured to execute the following steps by invoking the operation instructions stored in the memory 1004:

[0138] Obtaining first original data corresponding to the target boring machine, wherein the first original data comprises boring machine running state data, cutterhead vibration acceleration data and surrounding rock point load strength data;

[0139] Removing the data of the target boring machine in the shutdown section and the empty pushing section in the first original data to obtain second original data;

[0140] Processing the abnormal values in the second original data to obtain target running data;

[0141] Extracting feature parameters in the target running data, wherein the feature parameters comprise cutterhead thrust, cutterhead torque, penetration, cutterhead rotation speed, cutterhead vibration acceleration and rock mass geomechanics property parameters;

[0142] Normalizing the feature parameters to obtain a model training data set;

[0143] Constructing a Stacking algorithm model according to a base learner and a meta-learner, wherein the base learner comprises a plurality of regression trees, and the output result of the base learner is used as the input of the meta-learner;

[0144] Training the Stacking algorithm model based on the model training data set to obtain a full-face boring machine tunneling surrounding rock strength prediction model.

[0145] The processor 1003 is further configured to execute the following steps by invoking the operation instructions stored in the memory 1004: Figure 1 Any of the manners in the corresponding embodiments.

[0146] Please refer to Figure 11 , Figure 11 The embodiment of the electronic device provided by the present application is shown in the embodiment of the electronic device provided by the present application.

[0147] As Figure 11As shown, the embodiment of the present application provides an electronic device, which comprises a memory 1110, a processor 1120, and a computer program 1111 stored in the memory 1110 and capable of running on the processor 1120, and the processor 1120 implements the following steps when executing the computer program 1111:

[0148] Obtaining first original data corresponding to the target boring machine, wherein the first original data comprises boring machine running state data, cutterhead vibration acceleration data, and surrounding rock point load strength data;

[0149] Removing data of the target boring machine in the shutdown section and the empty pushing section in the first original data to obtain second original data;

[0150] Processing abnormal values in the second original data to obtain target running data;

[0151] Extracting feature parameters in the target running data, wherein the feature parameters comprise cutterhead thrust, cutterhead torque, penetration, cutterhead rotation speed, cutterhead vibration acceleration, and rock mass geomechanics property parameters;

[0152] Normalizing the feature parameters to obtain a model training data set;

[0153] Constructing a Stacking algorithm model according to a base learner and a meta learner, wherein the base learner comprises a plurality of regression trees, and an output result of the base learner is used as an input of the meta learner;

[0154] Training the Stacking algorithm model based on the model training data set to obtain a full-face boring machine tunneling surrounding rock strength prediction model.

[0155] In the specific implementation process, when the processor 1120 executes the computer program 1111, the following steps can be implemented Figure 1 Any of the embodiments of the corresponding embodiments.

[0156] Since the electronic device introduced in the embodiment is the device used to implement the full-face boring machine tunneling surrounding rock strength prediction model training device in the embodiment of the present application, based on the method introduced in the embodiment of the present application, those skilled in the art can understand the specific implementation of the electronic device of the embodiment and its various forms, so the electronic device how to implement the method in the embodiment of the present application is not introduced in detail, as long as the device used by those skilled in the art to implement the method in the embodiment of the present application belongs to the scope of the present application.

[0157] Please refer to Figure 12 , Figure 12An embodiment of a computer readable storage medium provided by the embodiment of the present application is shown in the figure.

[0158] As shown in Figure 12 The embodiment of the present application also provides a computer readable storage medium 1200, and the computer readable storage medium 1200 stores a computer program 1211, and the computer program 1211 is executed by a processor to realize the following steps:

[0159] Obtaining first original data corresponding to a target boring machine, the first original data including boring machine running state data, cutterhead vibration acceleration data and surrounding rock point load strength data;

[0160] Removing data of a shutdown section and an empty pushing section of the target boring machine in the first original data to obtain second original data;

[0161] Processing an abnormal value in the second original data to obtain target running data;

[0162] Extracting a feature parameter in the target running data, the feature parameter including cutterhead thrust, cutterhead torque, penetration, cutterhead rotation speed, cutterhead vibration acceleration and rock mass geomechanics property parameter;

[0163] Normalizing the feature parameter to obtain a model training data set;

[0164] Constructing a Stacking algorithm model according to a base learner and a meta learner, the base learner including a plurality of regression trees, and an output result of the base learner being used as an input of the meta learner;

[0165] Training the Stacking algorithm model based on the model training data set to obtain a full-face boring machine tunneling surrounding rock strength prediction model.

[0166] In the specific implementation process, the computer program 1211 is executed by the processor to realize Figure 1 Any embodiment in the corresponding embodiment.

[0167] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0168] The embodiment of the present application also provides a computer program product, and the computer program product includes computer software instructions, when the computer software instructions run on a processing device, the processing device executes the flow in the Figure 1 Corresponding embodiment.

[0169] The above-described and above-embodied examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the present application; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions recorded in the foregoing examples can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A full-face tunneling machine tunneling roadway surrounding rock strength prediction model training method, characterized in that, The method comprises the following steps: obtaining first original data corresponding to a target boring machine, wherein the first original data comprises boring machine running state data, cutterhead vibration acceleration data, and surrounding rock point load strength data; removing data of the target boring machine in a shutdown section and an empty pushing section from the first original data to obtain second original data; processing abnormal values in the second original data to obtain target running data; extracting feature parameters in the target running data, wherein the feature parameters comprise cutterhead thrust, cutterhead torque, penetration, cutterhead rotation speed, cutterhead vibration acceleration, and rock mass geomechanics property parameters; performing normalization processing on the feature parameters to obtain a model training data set; constructing a Stacking algorithm model according to a base learner and a meta learner, wherein the base learner comprises a plurality of regression trees, and an output result of the base learner is used as an input of the meta learner; training the Stacking algorithm model based on the model training data set to obtain a full-face boring machine tunneling surrounding rock strength prediction model; the removing of the data of the target boring machine in the shutdown section and the empty pushing section from the first original data to obtain the second original data comprises: determining first sub-data of the target boring machine in the shutdown section and second sub-data of the target boring machine in the tunneling section in the first original data according to the following formula: ； wherein, is the tool thrust force, is the tool torque, is the tool rotational speed, is the discriminant function, , ; determining third sub-data of the target boring machine in the empty pushing section according to the second sub-data, comprising: determining a first tunneling data in the second sub-data; determining a target time based on a time of the first tunneling data and a preset time length; determining data between the time of the first tunneling data and the target time as the third sub-data; removing the first sub-data and the third sub-data from the first original data to obtain the second original data.

2. The method of claim 1, wherein, the processing of the abnormal values in the second original data to obtain the target running data comprises: determining upper and lower whiskers corresponding to the second original data based on a box plot according to the following formula: ； ； wherein, is the upper whisker, is the lower whisker, is the upper quartile corresponding to the box plot, is the lower quartile corresponding to the box plot, is the lower quartile deviation corresponding to the box plot, ; determining data values in the second original data that are not between the upper and lower whiskers as the abnormal values; removing the abnormal values in the second original data to obtain third running data; complementing missing values in the third running data based on a linear interpolation method to obtain the target running data.

3. The method of any one of claims 1-2, wherein, the training of the Stacking algorithm model based on the model training data set to obtain the full-face boring machine tunneling surrounding rock strength prediction model comprises: dividing the model training data into a training set and a test set; dividing the training set into a plurality of training subsets; determining a first training subset as a validation set, and using other training subsets in the plurality of training subsets except the first training subset as training sets, training each regression tree in the plurality of regression trees through K-fold cross-validation to obtain a plurality of first prediction results corresponding to each regression tree; Test each regression tree corresponding to each test data in the test set respectively to obtain a second prediction result corresponding to each regression tree; Input the multiple first prediction results into the meta-learner for training, and test the trained meta-learner based on the second prediction result to obtain the full-face tunnel boring machine tunneling surrounding rock strength prediction model.

4. The method of any one of claims 1 to 2, characterized in that, The normalization processing on the feature parameters comprises: The normalization processing on the feature parameters comprises: wherein, is the normalized feature parameter, is the feature parameter, and are the maximum and minimum values of the feature parameter, respectively.

5. A full-face tunneling machine tunneling roadway surrounding rock strength prediction model training device, characterized in that, Comprise: The acquisition module is configured to acquire first original data corresponding to a target tunneling machine, wherein the first original data comprises tunneling machine running state data, cutterhead vibration acceleration data, and surrounding rock point load strength data. The elimination module is configured to eliminate data of a shutdown section and an empty pushing section of the target tunneling machine in the first original data to obtain second original data. The outlier processing module is configured to process outliers in the second original data to obtain target running data. The feature extraction module is configured to extract feature parameters in the target running data, wherein the feature parameters comprise cutterhead thrust, cutterhead torque, penetration, cutterhead rotation speed, cutterhead vibration acceleration, and rock mass geomechanics property parameters. The normalization module is configured to normalize the feature parameters to obtain a model training data set. The construction module is configured to construct a Stacking algorithm model according to a base learner and a meta-learner, wherein the base learner comprises multiple regression trees, and an output result of the base learner is used as an input of the meta-learner. The training module is configured to train the Stacking algorithm model based on the model training data set to obtain a full-face tunnel boring machine tunneling surrounding rock strength prediction model. The elimination module is specifically configured to: The first sub-data of the target tunneling machine in the shutdown section and the second sub-data of the target tunneling machine in the tunneling section in the first original data are determined by the following formula: ； wherein, is the tool thrust force, is the tool torque, is the tool rotational speed, is the discriminant function, , ; The third sub-data of the target tunneling machine in the empty pushing section is determined according to the second sub-data. The first sub-data and the third sub-data are eliminated from the first original data to obtain the second original data. The elimination module is further specifically configured to: determine a first tunneling data in the second sub-data; determine a target time based on a time of the first tunneling data and a preset time length; and determine data between the time of the first tunneling data and the target time as the third sub-data.

6. An electronic device, comprising: Comprise: The memory and the processor are configured to implement the full-face tunnel boring machine tunneling surrounding rock strength prediction model training method in any one of claims 1 to 4 when the computer management program is executed by the processor.

7. A computer readable storage medium having stored thereon a computer management class program, the program comprising: The computer management program is configured to implement the full-face tunnel boring machine tunneling surrounding rock strength prediction model training method in any one of claims 1 to 4 when executed by the processor.

Images