Method and device for constructing rock mass condition prediction model, equipment and storage medium

Abstract

The application provides a rock mass condition prediction model construction method and device, equipment and a storage medium. The method comprises the following steps: obtaining original tunneling data of a tunnel boring machine and a rock mass state grade corresponding to the original tunneling data as original data; performing feature screening and feature construction according to the original data to form a model training data set; performing model training by using multiple machine learning model training algorithms according to the model training data set; in the model training process, a northern goshawk optimization algorithm improved by a vertical and horizontal cross strategy is used to optimize hyperparameters in the model training process; and the machine learning model with the best model performance is used as the rock mass condition prediction model according to the model performance of the multiple machine learning models obtained by training.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of TBM intelligent sensing construction, and particularly relates to a rock mass condition prediction model construction method and device, equipment and a storage medium. BACKGROUND

[0002] Compared with the drill-and-blast method construction, the tunnel boring machine (TBM) construction method has the characteristics of high efficiency, high efficiency, environmental protection, and small influence on the surrounding environment. However, when the TBM is excavating in a stratum with complex and variable structure, it is extremely easy to cause accidents such as machine jamming or equipment damage due to large deformation of the rock mass. Therefore, it is of great significance to master and understand the tunnel rock mass condition in real time for construction safety, strengthening of support, and reduction of construction risks.

[0003] Most of the existing rock mass condition information comes from pre-construction drilling exploration, advanced drilling during construction, TRT or TSP geophysical prospecting methods, which require special detection equipment and professional personnel. Moreover, most of the methods cannot allow the TBM to normally excavate forward during the detection process, which will affect the progress of the project to some extent. Therefore, it is urgent to provide a more convenient rock mass condition prediction method. SUMMARY

[0004] The application provides a rock mass condition prediction model construction method and device, equipment and a storage medium, which can form a rock mass condition prediction model with high prediction accuracy.

[0005] In one aspect, a rock mass condition prediction model construction method is provided, comprising:

[0006] Obtaining original excavation data of a tunnel boring machine and a rock mass state grade corresponding to the original excavation data as original data;

[0007] According to the original data, feature screening and feature construction are performed to form a model training data set;

[0008] According to the training set in the model training data set, a plurality of machine learning model training algorithms are used for model training. In the model training process, a Northern Goshawk optimization algorithm improved by a vertical and horizontal cross strategy is used to optimize the hyperparameters in the model training process.

[0009] According to the model performance of the plurality of machine learning models obtained by training, the machine learning model with the best model performance is used as the rock mass condition prediction model.

[0010] Optionally, according to the original data, feature screening and feature construction are performed to form a model training data set, comprising:

[0011] Isolated forest algorithm is adopted to remove outliers in the original data, and the null values in the original data are filled to form a first tunneling parameter data set;

[0012] The first tunneling parameter data set is subjected to feature screening, the variance of a single feature and the correlation between features are calculated, features with a variance below a variance threshold or a correlation above a correlation threshold are removed, and a second tunneling parameter data set is obtained;

[0013] According to the data in the first tunneling parameter data set, feature construction is performed to obtain a tunneling performance feature data set, which includes a thrust penetration index, a torque penetration index and a specific penetration degree;

[0014] The second tunneling parameter data set and the tunneling performance feature data set are merged to obtain a model training data set.

[0015] Optionally, according to the data in the first tunneling parameter data set, feature construction is performed to obtain a tunneling performance feature data set, which includes a thrust penetration index, a torque penetration index and a specific penetration degree, including:

[0016] According to the formula feature construction is performed to obtain a field thrust penetration degree index , wherein represents thrust, the number of represents the number of cutters, and represents penetration depth;

[0017] According to the formula feature construction is performed to obtain a field torque penetration degree index , wherein represents cutter torque, and represents cutter diameter;

[0018] According to the formula feature construction is performed to obtain a specific penetration degree .

[0019] Optionally, in the model training process, the Northern Goshawk optimization algorithm improved by the vertical and horizontal cross strategy is adopted to optimize the hyperparameters in the model training process, including:

[0020] Step 1, initialize the Northern Goshawk population , , wherein represents the population matrix of the Northern Goshawk, represents the position of the th Northern Goshawk, and represents the th Northern Goshawk. The position of each dimension of the Northern Goshawk represents the value of a hyperparameter.

[0021] Step 2: Determine the value of a certain model performance index of the machine learning model corresponding to the Northern Goshawk based on the location of the Northern Goshawk, and use it as the fitness value; the model performance index includes any one of the following: the weighted average of accuracy and recall, accuracy, and the area under the receiver operating feature curve.

[0022] Step 3, use the formula , , To identify and attack prey; Indicates the first After the first Northern Goshawk identified and attacked its prey, the second... Wei's new position; Indicates the first Fitness value of a Northern Goshawk; Indicates the first The location of the prey of the northern eagle; Indicates the first The first of the prey of the northern goshawk The position of the dimension; Indicates the first The fitness value of the prey of the Northern Goshawk; Indicates the first Target location update during prey identification and attack operations by a Northern Goshawk; Indicates the first step in prey identification and attack operations. The fitness value of the target location of a Northern Goshawk is updated. for Random numbers within; It is a random number, either 1 or 2;

[0023] Step 4, use the formula , , They conducted chase and escape maneuvers; among them, Indicates the first After the Northern Goshawk completed its pursuit and escape maneuvers, the... Wei's new position; Indicates the first The target location is updated when a Northern Goshawk performs a chase and escape maneuver. Indicates the first The fitness value of the target's updated position when a Northern Goshawk performs a chase and escape maneuver; Indicates the current iteration number; Indicates the maximum number of iterations;

[0024] Step 5, performing a horizontal crossover operation on the positions of the northern hawk in step 2 and the positions of the northern hawk obtained in step 4;

[0025] Step 6, performing a vertical crossover operation on the positions of the northern hawk in step 2;

[0026] Step 7, comparing the fitness values of the northern hawk before and after the vertical crossover in step 6, and retaining the northern hawk with the higher fitness value;

[0027] Step 8, repeating steps 2 to 7 until the maximum number of iterations is reached , outputting the coordinates of the northern hawk with the highest fitness value as the hyperparameter combination of the machine learning model.

[0028] Optionally, the horizontal crossover operation on the positions of the northern hawk in step 2 and the positions of the northern hawk obtained in step 4 includes:

[0029] According to the formula , the first dimension of the position of the northern hawk in step 2 and the first dimension of the position of the northern hawk obtained in step 4 are horizontally crossed; wherein, represents the first dimensional position of the th northern hawk in step 2; represents the first dimensional position of the th northern hawk obtained in step 4; represents the first dimension of the offspring individual obtained by performing a horizontal crossover on the th northern hawk in step 2; represents the first dimension of the offspring individual obtained by performing a horizontal crossover on the th northern hawk obtained in step 4; and is a random number between

[0030] Optionally, the vertical crossover operation on the positions of the northern hawk in step 2 includes:

[0031] According to the formula , the first dimension and the second dimension of the northern hawk in step 2 are vertically crossed; wherein, is the first dimension of the northern hawk in step 2 and the second The first The first

[0032] Optionally, according to the model performance of the plurality of machine learning models obtained through training, the machine learning model with the best model performance is taken as the rock mass condition prediction model, including:

[0033] The weighted average of the accuracy and the recall rate of the machine learning model obtained through training, the accuracy, and the area under the curve of the receiver operating characteristic curve are accumulated to obtain the model performance index accumulation of the machine learning model.

[0034] The machine learning model with the maximum model performance index accumulation is taken as the rock mass condition prediction model.

[0035] In another aspect, a rock mass condition prediction model construction device is provided, including:

[0036] The acquisition module is configured to acquire original tunneling data of a TBM and a rock mass state grade corresponding to the original tunneling data as original data.

[0037] The data processing module is configured to perform feature screening and feature construction according to the original data to form a model training data set.

[0038] The model training module is configured to perform model training on the training set in the model training data set by using a plurality of machine learning model training algorithms, to optimize hyperparameters in the model training process by using a Northern Goshawk optimization algorithm improved through a longitudinal and transverse cross strategy, and to take the machine learning model with the best model performance as the rock mass condition prediction model according to the model performance of the plurality of machine learning models obtained through training.

[0039] In another aspect, an electronic device is provided, including the rock mass condition prediction model construction device as described above.

[0040] In another aspect, a computer readable storage medium is provided, and the computer readable storage medium stores at least one program code, and the program code is executed by a processor to implement the rock mass condition prediction model construction method as described in any one of the above.

[0041] The technical scheme provided by the embodiments of the present disclosure has the following beneficial effects:

[0042] In the embodiments of the present disclosure, a method for constructing a rock mass condition prediction model is provided. The method comprises the following steps: obtaining original tunneling data of a tunnel boring machine and a rock mass state grade corresponding to the original tunneling data as original data, performing feature screening and feature construction on the original data to form a model training data set, and the data in the model training data set is more conducive to the model performance of a machine learning model obtained by training than the original data. In the model training process, the hyperparameters in the model training process are optimized by using the Northern Goshawk optimization algorithm improved by a vertical-horizontal cross strategy, so as to ensure that the combination of the hyperparameters used in the machine learning model training process is optimal, thereby further improving the model performance of the machine learning model. Finally, the model performance index of the machine learning model is evaluated, the model with the optimal performance is taken as the rock mass condition prediction model, and the prediction accuracy of the model is ensured. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0044] Figure 1 A method flowchart of a method for constructing a rock mass condition prediction model provided by the embodiments of the present disclosure is provided.

[0045] Figure 2 A TBM tunneling operation parameter complete cycle diagram provided by the embodiments of the present disclosure is provided.

[0046] Figure 3 A flowchart of an original data data processing process provided by the embodiments of the present disclosure is provided.

[0047] Figure 4 A flowchart of model training provided by the embodiments of the present disclosure is provided.

[0048] Figure 5 A structural block diagram of a rock mass condition prediction model construction device provided by the embodiments of the present disclosure is provided.

[0049] Figure 6 A structural block diagram of an electronic device provided by the embodiments of the present disclosure is provided.

[0050] The reference signs are as follows:

[0051] 21: acquisition module; 22: data processing module; 23: model construction module;

[0052] 31: processor; 32: memory. DETAILED DESCRIPTION

[0053] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below in conjunction with the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0054] In essence, the tunneling operation parameters generated by the TBM in real time are the result of the interaction between the TBM and the rock mass, so the tunneling operation parameters of the TBM can represent the current tunneling state of the TBM in real time, and reflect the geological conditions of the surrounding rock. Therefore, by further mining the tunneling operation parameters of the TBM, a machine learning model based on the operation parameters is established, which can predict the output rock mass conditions in real time, and is beneficial to the real-time decision-making of the TBM operator and the construction personnel, the selection of appropriate support and control measures, the reduction of the increase of construction risk, and the important significance to the safety construction of the TBM tunnel project.

[0055] Figure 1 A method flowchart of a method for constructing a rock mass condition prediction model is provided for the embodiments of the present disclosure. Referring to Figure 1 , the method comprises:

[0056] S11, obtaining original tunneling data of a TBM and a rock mass state grade corresponding to the original tunneling data as original data.

[0057] The original tunneling state data of the tunnel boring machine (TBM) includes main tunneling operation parameters (cutterhead rotation speed, penetration, thrust, torque, etc.), mechanical motor current and output power parameters, mechanical pressure and pressure parameters, mechanical oil level and temperature, etc. The original tunneling data output by different types of TBM will be different.

[0058] In the embodiments of the present disclosure, as for the rock mass state grade, if the TBM excavates a highway tunnel, the rock mass state grade is divided into six levels, I, II, III, IV, V and VI, according to the hardness and completeness of the rock mass, and reference is made to the “Highway Tunnel Design Specification” (JTG 3370.1-2018); if the TBM excavates a water conveyance tunnel, the rock mass state grade is divided into five categories, I, II, III, IV and V, according to the rock strength, integrity, structure surface state, underground water and structure surface occurrence, and reference is made to the “Water Conservancy and Hydropower Engineering Geological Exploration Specification” (GB 50487-2008).

[0059] S12, performing feature screening and feature construction according to the original data to form a model training data set.

[0060] In an example, step S12 comprises:

[0061] Step 1, adopt the isolated forest algorithm to remove the outliers in the original data, and fill in the null values in the original data to form a first tunneling parameter data set.

[0062] In the embodiments of the present disclosure, the method steps of the isolated forest are as follows:

[0063] Step 1, select a plurality of groups of small-volume samples.

[0064] Step 2, construct an isolated tree for each group of samples, randomly select a feature at each node, and randomly select a split value between the maximum and minimum values of the feature, and divide the samples into left and right sub-trees based on the split value, and the end condition is that the tree reaches a limited height or the number of node samples reaches a certain number;

[0065] Step 3, repeat steps 1 and 2 to construct a certain number of isolated trees to form an isolated forest;

[0066] Step 4, calculate the average path length required for the sample to be isolated from the root node of the tree to the node where the sample is isolated in the isolated forest, and the sample with a shorter average path length is identified as an abnormal sample, and the sample is removed. In addition, for the case where there are null values in the recorded feature, the average value of the adjacent time feature variable is selected to fill in the null value, and finally a first tunneling parameter data set is formed.

[0067] Step 2, perform feature screening on the first tunneling parameter data set, calculate the variance of a single feature and the correlation between features, remove features with a variance below a variance threshold or a correlation above a correlation threshold, and obtain a second tunneling parameter data set.

[0068] In an example, step 2 comprises:

[0069] Step 1, according to the formula the variance of a single feature is calculated, wherein represents the number of data participating in the calculation of a certain feature, represents a certain data of a certain feature, represents the mean value of the data participating in the calculation of a certain feature.

[0070] Step 2, according to the Pearson correlation coefficient calculation formula the correlation between features is calculated. Wherein, and respectively represent the data variables (i.e. a certain data in the data participating in the calculation of the feature) of the two features, and represent the average value of the data variable (i.e. the average value of the data participating in the calculation of the feature).

[0071] Thirdly, remove the low-variance features (variance is lower than the variance threshold) and the high-correlation features (correlation is higher than the correlation threshold).

[0072] For example, the variance threshold can be 1.5, and the correlation threshold can be 0.8.

[0073] Of course, the above thresholds are only an example provided by the embodiments of the present disclosure, and the variance threshold and the correlation threshold can be selected optionally, and the present disclosure does not limit this.

[0074] Step 3: According to the data in the first tunneling parameter data set, feature construction is performed to obtain a tunneling performance feature data set, wherein the tunneling performance feature data set includes a thrust penetration index, a torque penetration index and a specific penetration.

[0075] In an example, step 3 includes:

[0076] Firstly, according to the formula feature construction is performed to obtain a field thrust penetration index , wherein represents the thrust, and represents the number of cutters, and represents the penetration.

[0077] Secondly, according to the formula feature construction is performed to obtain a field torque penetration index , wherein represents the cutterhead torque, and represents the cutterhead diameter.

[0078] Thirdly, according to the formula feature construction is performed to obtain a specific penetration .

[0079] In the embodiments of the present disclosure, the feature construction is performed by the above formula, and the , and are calculated according to the original data, which can describe the tunneling performance features of the tunnel boring machine, improve the richness of the training data, and be beneficial to the prediction accuracy of the trained model.

[0080] Step 4: Merge the second tunneling parameter data set and the tunneling performance feature data set to obtain a model training data set.

[0081] In the embodiments of the present disclosure, by performing the outlier removal, the null value filling, the feature screening and the feature construction on the original data, on the one hand, the integrity of the original data can be ensured, and on the other hand, the training data can all use the better data, which is beneficial to the prediction accuracy of the trained model.

[0082] S13, training the model according to the model training data set by using multiple machine learning model training algorithms; in the model training process, the hyperparameters in the model training process are optimized by using the northern goshawk optimization algorithm improved by the vertical and horizontal cross strategy.

[0083] In an example, step S13 comprises:

[0084] Step 1, initializing the northern goshawk population , , the indicates the population matrix of the northern goshawk, indicates the position of the th northern goshawk, and the indicates the position of the th northern goshawk in the th dimension. The position of each dimension of the northern goshawk indicates the value of a hyperparameter.

[0085] Step 2, determining the value of a certain model performance indicator of the machine learning model corresponding to the northern goshawk according to the position of the northern goshawk as the fitness value, and the model performance indicator includes any one of the weighted average (F1) of accuracy and recall rate, accuracy (Acurracy), and the area under the curve (AUC) of the receiver operating characteristic curve.

[0086] Step 3, performing prey recognition and attack operation by using the formula , , indicates the new position of the th northern goshawk in the th dimension after performing prey recognition and attack operation; indicates the fitness value of the th northern goshawk; indicates the position of the prey of the th northern goshawk; indicates the position of the prey of the th northern goshawk in the th dimension; indicates the fitness value of the prey of the th northern goshawk; indicates the target update position of the th northern goshawk when performing prey recognition and attack operation; indicates the fitness value of the target update position of the th northern goshawk when performing prey recognition and attack operation; is a random number in ; and is a random number 1 or 2. ​

[0087] Step 4, use the formula , , They conducted chase and escape maneuvers; among them, Indicates the first After the Northern Goshawk completed its chase and escape maneuvers, the... Wei's new position; Indicates the first The target location is updated when a Northern Goshawk performs a chase and escape maneuver. Indicates the first The fitness value of the target's updated position when a Northern Goshawk performs a chase and escape maneuver; Indicates the current iteration number; This indicates the maximum number of iterations.

[0088] Step 5: Perform a horizontal cross operation between the position of the Northern Goshawk obtained in Step 2 and the position of the Northern Goshawk obtained in Step 4.

[0089] In one example, step 5 includes:

[0090] According to the formula , Parental individuals and the offspring individuals formed in step 4 The Dimensions intersect horizontally; among them, and Representing the parent individuals and offspring individuals The dimension, and Representing the parent individuals Offspring individuals The first generation of offspring obtained by lateral crossover dimension, and for A random number between [a certain number of points].

[0091] In this embodiment of the disclosure, the above formula can be used to achieve horizontal cross-operation between parent and offspring individuals, further enhancing the richness of the Northern Goshawk population.

[0092] Step 6: Perform a vertical cross operation on the position of the Northern Eagle in Step 2.

[0093] In one example, step 6 includes:

[0094] According to the formula Parental individuals The peacekeeping the first dimension and the second dimension are crossed longitudinally, the parent individual the first dimension and the second dimension of the offspring obtained after the first dimension and the second dimension are crossed longitudinally. dimension.

[0095] In the embodiments of the present disclosure, the parent individual of the northern goshawk is subjected to longitudinal crossing operation through the above formula, and the richness of the population is improved.

[0096] Step 7, comparing the fitness values of the northern goshawk before and after the longitudinal crossing in step 6, and retaining the northern goshawk with the higher fitness value.

[0097] Step 8, repeating steps 2 to 7 until the maximum number of iterations is met, and outputting the coordinates of the northern goshawk with the highest fitness value as the hyperparameter combination of the machine learning model.

[0098] In the embodiments of the present disclosure, by using the longitudinal and transverse crossing strategy to improve the northern goshawk optimization algorithm, the ability of the northern goshawk algorithm to jump out of the local optimal value can be enhanced, and the model performance of the trained machine learning model can be improved.

[0099] In step S13, first, the data in the model training data set is divided into a training set (80%) and a test set (20%), and then in a ten-fold cross-validation manner, machine learning algorithms such as Light Gradient Boosting Machine (LightGBM) algorithm, Extreme Gradient Boosting (XGBoost) algorithm, Random Forest (RF) algorithm, Support Vector Machine (SVM) algorithm, and Extra-Trees (ET) are used for model training. In the training process, the hyperparameters are optimized, for example, the F1 is used as the optimization target to optimize the hyperparameters, and the hyperparameter combination is obtained for model training.

[0100] S14, according to the model performance of the plurality of machine learning models trained, the machine learning model with the best model performance is used as the rock mass condition prediction model.

[0101] In an example, step S14 includes:

[0102] In the first step, the model performance indicators of the trained machine learning model are accumulated to obtain the accumulated sum of the model performance indicators of the machine learning model, including the weighted average (F1) of the accuracy and recall rate, the accuracy (Acurracy), and the area under the curve (AUC) of the receiver operating characteristic curve.

[0103] In the second step, the machine learning model with the maximum accumulated sum of the model performance indicators is taken as the rock mass condition prediction model.

[0104] In the embodiments of the present disclosure, the model performance indicators are used to evaluate the model performance of the trained machine learning model, and the model performance of the finally trained machine learning models is evaluated by determining the sizes of F1, Acurracy, and AUC in the model performance indicators and determining the sum of F1, Acurracy, and AUC, and the one with the optimal performance can be used for rock mass condition prediction.

[0105] In the embodiments of the present disclosure, a method for constructing a rock mass condition prediction model is provided, in which the original tunneling data of a tunnel boring machine and the rock mass state grade corresponding to the original tunneling data are obtained as original data, the original data are subjected to feature screening and feature construction to form a model training data set, and the data in the model training data set are more conducive to the model performance of the trained machine learning model than the original data. In the model training process, the hyperparameters in the model training process are optimized by using the Northern Goshawk optimization algorithm improved by the vertical-horizontal cross strategy, so that the combination of the hyperparameters used in the machine learning model training process is optimal, thereby further improving the model performance of the machine learning model. Finally, the model with the optimal performance is taken as the rock mass condition prediction model by evaluating the model performance indicators of the machine learning model, so as to ensure the prediction accuracy of the model.

[0106] Figure 2 A TBM tunneling operation parameter complete cycle diagram is provided for the embodiments of the present disclosure.

[0107] Figure 3 A flowchart of the data processing process of the original data is provided for the embodiments of the present disclosure. Referring to Figure 3 , Figure 3 The processing process of the original data is shown, and the data serial number represents the processing step.

[0108] Figure 4 A flowchart of the model training is provided for the embodiments of the present disclosure. Referring to Figure 4 , Figure 4 The process of the model training is shown in

[0109] Figure 5 A structural block diagram of a construction device of a rock mass condition prediction model is provided for the embodiments of the present disclosure. Referring toFigure 5 ,include:

[0110] The acquisition module 21 is used to acquire the original tunneling data of the TBM and the rock mass condition level corresponding to the original tunneling data, as the original data.

[0111] The data processing module 22 is used to perform feature filtering and feature construction based on the raw data to form a model training dataset.

[0112] The model training module 23 is used to train the model using multiple machine learning model training algorithms based on the model training dataset. During the model training process, the Northern Eagle optimization algorithm, which is improved through a cross-cutting strategy, is used to optimize the hyperparameters during the model training process. Based on the model performance of the multiple machine learning models obtained from the training, the machine learning model with the best performance is selected as the rock mass condition prediction model.

[0113] Optionally, the data processing module 22 is used for:

[0114] The Isolation Forest algorithm is used to remove outliers from the original data and fill in the null values ​​in the original data to form the first tunneling parameter dataset;

[0115] Feature filtering is performed on the first tunneling parameter dataset. The variance of a single feature and the correlation between features are calculated. Features with variances below the variance threshold or correlations above the correlation threshold are removed to obtain the second tunneling parameter dataset.

[0116] Based on the data in the first tunneling parameter dataset, feature construction is performed to obtain a tunneling performance feature dataset, which includes thrust penetration index, torque penetration index and specific penetration.

[0117] The second tunneling parameter dataset and the tunneling performance feature dataset are merged to obtain the model training dataset.

[0118] Optionally, the data processing module 22 is used for:

[0119] According to the formula Feature construction was performed to obtain the field thrust penetration index. The Indicates thrust, the Indicates the number of hobs, the Indicates penetration degree;

[0120] According to the formula Feature construction was performed to obtain the field torque penetration index. The Indicates the torque of the cutter head, the Indicates the diameter of the cutter head;

[0121] According to the formula Feature construction is performed to obtain the penetration ratio .

[0122] Optionally, the model training module 23 is configured to:

[0123] Step 1, initialize the northern goshawk population , , the represents the population matrix of the northern goshawk, represents the position of the th northern goshawk, and the represents the position of the th northern goshawk in the th dimension, and the position of each dimension of the northern goshawk represents the value of a hyperparameter;

[0124] Step 2, determining the value of a certain model performance indicator of the machine learning model corresponding to the northern goshawk according to the position of the northern goshawk as the fitness value; the model performance indicator includes any one of the weighted average of the accuracy and the recall rate, the accuracy, and the area under the curve of the receiver operating characteristic curve;

[0125] Step 3, performing prey recognition and attack operation by using the formula , , represents the new position of the th northern goshawk in the th dimension after the prey recognition and attack operation; represents the fitness value of the th northern goshawk; represents the position of the prey of the th northern goshawk; represents the position of the prey of the th northern goshawk in the th dimension; represents the fitness value of the prey of the th northern goshawk; represents the target update position of the th northern goshawk during the prey recognition and attack operation; represents the fitness value of the target update position of the th northern goshawk during the prey recognition and attack operation; is a random number within ; is a random number 1 or 2;

[0126] Step 4, performing prey recognition and attack operation by using the formula , , ​They conducted chase and escape maneuvers; among them, Indicates the first After the Northern Goshawk completed its pursuit and escape maneuvers, the... Wei's new position; Indicates the first The target location is updated when a Northern Goshawk performs a chase and escape maneuver. Indicates the first The fitness value of the target's updated position when a Northern Goshawk performs a chase and escape maneuver; Indicates the current iteration number; Indicates the maximum number of iterations;

[0127] Step 5: Perform a horizontal cross operation between the position of the Northern Goshawk obtained in Step 2 and the position of the Northern Goshawk obtained in Step 4.

[0128] Step 6: Perform a vertical cross-operation on the position of the Northern Goshawk in Step 2;

[0129] Step 7: Compare the fitness values ​​of the Northern Goshawks before and after the longitudinal crossing in Step 6, and retain the Northern Goshawks with higher fitness values.

[0130] Step 8: Repeat steps 2 through 7 until the maximum number of iterations is met. At that time, the coordinates of the Northern Goshawk with the highest fitness value are output as the hyperparameter combination of the machine learning model.

[0131] Optionally, the model training module 23 is used for:

[0132] According to the formula , The location of the Northern Goshawk in step 2 The location of the Northern Goshawk obtained in step 4 The Dimensions intersect horizontally; among them, This indicates the first step in step 2. The first Northern Eagle Dimensional position; This indicates the result obtained in step 4. The first Northern Eagle Dimensional position; This indicates the first step in step 2. The offspring of the Northern Goshawk obtained by lateral crossbreeding dimension; This indicates the result obtained in step 4. The offspring of the Northern Goshawk obtained by lateral crossbreeding dimension; and for A random number between [a certain number of points].

[0133] Optionally, the model training module 23 is configured to:

[0134] According to the formula The first dimension and the first dimension of the Northern Goshawk in step 2 are longitudinally crossed; wherein, The first dimension of the offspring individual obtained after the first dimension and the first dimension of the Northern Goshawk in step 2 are longitudinally crossed.

[0135] Optionally, the model training module 23 is configured to:

[0136] The weighted average of the accuracy and recall rate, the accuracy, and the area under the curve of the receiver operating characteristic curve of the trained machine learning model are accumulated to obtain a model performance index accumulation sum of the machine learning model.

[0137] The machine learning model with the largest model performance index accumulation sum is taken as the rock mass condition prediction model.

[0138] Figure 6 A structural block diagram of an electronic device provided by the embodiments of the present disclosure is provided. Referring to Figure 6 , the electronic device can include Figure 5 the rock mass condition prediction model construction apparatus described above. Generally, the electronic device includes a processor 31 and a memory 32.

[0139] ​​The processor 31 can include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 31 can be implemented in at least one of a hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), a PLA (Programmable Logic Array). The processor 31 can also include a main processor and a coprocessor, the main processor being a processor for processing data in a wake-up state, also referred to as a CPU (Central Processing Unit), and the coprocessor being a low-power processor for processing data in a standby state. The memory 32 can include one or more computer-readable storage media, which can be non-transitory. The memory 32 can also include a high-speed random access memory, and a non-volatile memory such as one or more disk storage devices, flash memory devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 32 is used to store at least one instruction for being executed by the processor 31 to implement the method for constructing a rock mass condition prediction model performed by an electronic device provided by the method embodiments in the present application.

[0140] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; 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 method of constructing a rock mass condition prediction model, characterized by, The method comprises the following steps: Obtain original tunneling data of a tunnel boring machine and rock mass state grades corresponding to the original tunneling data as original data; Perform feature screening and feature construction according to the original data to form a model training data set; According to the model training data set, a plurality of machine learning model training algorithms are used for model training; in the model training process, the Northern Goshawk optimization algorithm improved by the longitudinal and transverse cross strategy is used to optimize the hyperparameters in the model training process; According to the model performance of the plurality of machine learning models obtained by training, the machine learning model with the best model performance is used as the rock mass condition prediction model; In the model training process, the Northern Goshawk optimization algorithm improved by the longitudinal and transverse cross strategy is used to optimize the hyperparameters in the model training process, comprising: Step 1, initializing a northern goshawk population , , the represents a population matrix of northern goshawks, represents the position of the th northern goshawk, the represents the position of the th northern goshawk in the th dimension, and the position of each dimension of the northern goshawk represents the value of a hyperparameter; Step 2, determine the value of a certain model performance index of the machine learning model corresponding to the Northern Goshawk according to the position of the Northern Goshawk, as the fitness value; the model performance index includes any one of the weighted average of accuracy and recall rate, accuracy, and the area under the receiver operating characteristic curve; Step 3, using the formula , , to perform prey recognition and attack operation; represents the new position of the th northern goshawk after performing prey recognition and attack operation; represents the fitness value of the th northern goshawk; represents the position of the prey of the th northern goshawk; represents the position of the prey of the th northern goshawk in the th dimension; represents the fitness value of the prey of the th northern goshawk; represents the target update position of the th northern goshawk when performing prey recognition and attack operation; represents the fitness value of the target update position of the th northern goshawk when performing prey recognition and attack operation; is a random number within ; and is a random number 1 or 2.​ Step 4, the pursuit and escape operation is performed using the formula , , , wherein represents the new position of the th northern hawk in the th dimension after the pursuit and escape operation; represents the target updated position of the th northern hawk when performing the pursuit and escape operation; represents the fitness value of the target updated position of the th northern hawk when performing the pursuit and escape operation; represents the current iteration number; represents the maximum iteration number; Step 5, transversely cross the positions of the Northern Goshawk in step 2 and step 4; Step 6, longitudinally cross the position of the Northern Goshawk in step 2; Step 7, compare the fitness values of the Northern Goshawk before and after the longitudinal cross in step 6, and retain the Northern Goshawk with the higher fitness value; Step 8, re-perform steps 2 to 7 until the maximum number of iterations is met When the fitness value is the highest, the coordinates of the northern goshawk are output as the hyperparameter combination of the machine learning model. The transverse cross operation of the positions of the Northern Goshawk in step 2 and the positions of the Northern Goshawk obtained in step 4 comprises: According to the formula , the position of the northern goshawk in step 2 and the position of the northern goshawk obtained in step 4 are crossbred in the first dimension; wherein, represents the first dimensional position of the n th northern goshawk in step 2; represents the first dimensional position of the n th northern goshawk obtained in step 4; represents the first dimensional position of the n th northern goshawk in step 2 after crossbreeding; represents the first dimensional position of the n th northern goshawk obtained in step 4 after crossbreeding; and is a random number between ; The longitudinal cross operation of the position of the Northern Goshawk in step 2 comprises: According to the formula The northern goshawk in step 2 The first Dimension and the first Dimension of the offspring individual after longitudinal crossing of the northern goshawk in step 2 The first Dimension and the first Dimension of the offspring individual after longitudinal crossing of the northern goshawk in step 2 Dimension of the offspring individual after longitudinal crossing of the northern goshawk in step 2 Dimension of the offspring individual after longitudinal crossing of the northern goshawk in step 2 2. The method of constructing a rock mass condition prediction model according to claim 1, characterized in that, According to the original data, perform feature screening and feature construction to form a model training data set, comprising: An isolated forest algorithm is used to remove outliers in the original data, and missing values in the original data are filled to form a first tunneling parameter data set; Perform feature screening on the first tunneling parameter data set, calculate the variance of a single feature and the correlation between features, remove features with a variance below a variance threshold or a correlation above a correlation threshold to obtain a second tunneling parameter data set; According to the data in the first tunneling parameter data set, perform feature construction to obtain a tunneling performance feature data set, wherein the tunneling performance feature data set includes thrust penetration index, torque penetration index and specific penetration; Merge the second tunneling parameter data set and the tunneling performance feature data set to obtain the model training data set.

3. The method of claim 2, wherein According to the data in the first tunneling parameter data set, perform feature construction to obtain a tunneling performance feature data set, wherein the tunneling performance feature data set includes thrust penetration index, torque penetration index and specific penetration, comprising: According to the formula The characteristic construction is carried out, and the field thrust penetration index is obtained , the represents the thrust, the represents the number of cutters, and the represents the penetration; According to the formula The characteristic construction is carried out, and the field torque penetration index is obtained , wherein represents the cutter torque, and the cutter diameter is represented by ; According to the formula The characteristic construction is carried out, and the specific penetration is obtained.

4. The method of constructing a rock mass condition prediction model according to any one of claims 1 to 3, characterized in that, According to the model performance of the plurality of machine learning models obtained by training, the machine learning model with the best model performance is used as the rock mass condition prediction model, comprising: The accuracy and recall rate of the machine learning model obtained by training, the accuracy, and the area under the receiver operating characteristic curve are accumulated to obtain the model performance index accumulation of the machine learning model; The machine learning model with the largest model performance index accumulation is used as the rock mass condition prediction model.

5. A device for constructing a rock mass condition prediction model, characterized in that, The method comprises the following steps: An acquisition module acquires original tunneling data of a tunnel boring machine and a rock mass state grade corresponding to the original tunneling data as original data; A data processing module is configured to perform feature screening and feature construction based on the original data to form a model training data set; A model training module is configured to perform model training based on the model training data set using multiple machine learning model training algorithms, to optimize hyperparameters in the model training process using a Northern Goshawk optimization algorithm improved through a vertical-horizontal cross strategy, and to select a machine learning model with the best model performance as a rock mass condition prediction model based on model performances of the multiple machine learning models obtained through training; In the model training process, the Northern Goshawk optimization algorithm improved through the vertical-horizontal cross strategy is used to optimize the hyperparameters in the model training process, including: Step 1, initializing a northern goshawk population , , the represents a population matrix of northern goshawks, represents the position of the th northern goshawk, the represents the position of the th northern goshawk in the th dimension, and the position of each dimension of the northern goshawk represents the value of a hyperparameter; Step 2: determining a value of a certain model performance indicator of the machine learning model corresponding to the Northern Goshawk according to the position of the Northern Goshawk as a fitness value; the model performance indicator includes any one of a weighted average of accuracy and recall rate, accuracy, and area under the curve of a receiver operating characteristic curve; Step 3, use the formula , , To identify and attack prey; Indicates the first After the first Northern Goshawk identified and attacked its prey, the second... Wei's new position; Indicates the first Fitness value of a Northern Goshawk; Indicates the first The location of the prey of the northern eagle; Indicates the first The first of the northern goshawk's prey The position of the dimension; Indicates the first The fitness value of the prey of the Northern Goshawk; Indicates the first Target location update during prey identification and attack operations by a Northern Goshawk; Indicates the first step in prey identification and attack operations. The fitness value of the target location of a Northern Goshawk is updated. for Random numbers within; It is a random number, either 1 or 2; Step 4, the pursuit and escape operation is performed using the formula , , , wherein represents the new position of the th northern goshawk after the pursuit and escape operation in the th dimension; represents the target updated position of the th northern goshawk when performing the pursuit and escape operation; represents the fitness value of the target updated position of the th northern goshawk when performing the pursuit and escape operation; represents the current iteration number; represents the maximum iteration number; Step 5: performing a horizontal cross operation on the position of the Northern Goshawk in Step 2 and the position of the Northern Goshawk obtained in Step 4; Step 6: performing a vertical cross operation on the position of the Northern Goshawk in Step 2; Step 7: comparing the fitness values of the Northern Goshawks before and after the vertical cross operation in Step 6, and retaining the Northern Goshawk with a higher fitness value; Step 8, re-perform steps 2 to 7 until the maximum number of iterations is met When the fitness value is the highest, the coordinates of the northern goshawk are output as the hyperparameter combination of the machine learning model. The horizontal cross operation on the position of the Northern Goshawk in Step 2 and the position of the Northern Goshawk obtained in Step 4 includes: According to the formula , the position of the northern goshawk in step 2 and the position of the northern goshawk obtained in step 4 are crossbred in the first dimension; wherein, represents the first dimensional position of the first northern goshawk in step 2; represents the first dimensional position of the first northern goshawk obtained in step 4; represents the first dimensional position of the first northern goshawk in step 2 after crossbreeding; represents the first dimensional position of the first northern goshawk obtained in step 4 after crossbreeding; and is a random number between ; The vertical cross operation on the position of the Northern Goshawk in Step 2 includes: According to the formula The Northern Goshawk in step 2 The peacekeeping Dimensions are vertically intersected; among them, To bring the Northern Goshawk in step 2 The peacekeeping The offspring individuals obtained after vertical crossover of dimensions dimension.

6. An electronic device, comprising: The construction device of the rock mass condition prediction model includes the rock mass condition prediction model according to claim 5.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores at least one program code, and the program code is executed by the processor to implement the construction method of the rock mass condition prediction model according to any one of claims 1 to 4.

Images