Shear strength prediction method and system for rock joint wall structural surface

By combining the shear strength prediction based on XGBoost model and the improved sparrow search algorithm, the complexity of shear strength acquisition in the existing technology is solved, and the rapid and accurate shear strength prediction of rock joint wall structural surface is achieved, reducing the experiment Dependency and overfitting risk.

CN118981958AInactive Publication Date: 2024-11-19GANNAN UNIV OF SCI & TECH

Patent Information

Application Number
CN202411071560.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-06
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The acquisition of shear strength in the prior art requires complicated experiments, which leads to complex and time-consuming process, making it difficult to achieve fast and accurate predictions.

Method used

A shear strength prediction model is established based on the XGBoost model, and the hyperparameters are optimized through the improved sparrow search algorithm, combining transfer learning and Levy flight strategy to achieve shear strength prediction of rock joint wall structural surfaces.

Benefits of technology

The shear strength prediction based on conventional rock mechanical properties is realized, which avoids the need for additional experiments, has the characteristics of rapid convergence and high precision, reduces the risk of overfitting, and improves the global search ability of simulation optimization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981958A_ABST
    Figure CN118981958A_ABST
Patent Text Reader

Abstract

The invention relates to the field of information processing, in particular to a shear strength prediction method and system for a rock joint wall structural surface. The invention discloses a shear strength prediction system for a rock joint wall structural surface. The shear strength prediction system comprises a rock property data set acquisition module and a shear strength prediction module. According to the method, the shear strength of the rock joint wall structural surface is predicted through the shear strength prediction model established based on the XGBoost model, the shear strength of the rock joint wall structural surface can be predicted based on conventional rock mechanical attribute input, and extra complicated test experiments do not need to be carried out; and the hyper-parameters of the shear strength prediction model are obtained through simulation and optimization of an improved sparrow search algorithm, and the method has the characteristics of high convergence speed, high prediction result precision and the like.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of information processing, and in particular to a method and system for predicting the shear strength of a rock joint wall structural surface. Background Art

[0002] Joint wall structural surfaces are naturally formed cracks or fracture surfaces in rocks. They are widely present in geological and engineering practices and have a profound impact on the stability and mechanical behavior of rock masses. Effectively predicting the shear strength of joint wall structural surfaces can not only guide the safe construction and support design of underground excavation projects, but also optimize rock quality assessment, geological disaster risk analysis, and determination of rock mechanical parameters. However, in the existing technology, the acquisition of shear strength requires complicated experiments. Summary of the invention

[0003] The present invention predicts the shear strength of the rock joint wall structural surface by using a shear strength prediction model established based on the XGBoost model. It can predict the shear strength of the rock joint wall structural surface based on conventional rock mechanical property inputs without the need for additional complicated test experiments. Moreover, the hyperparameters of the shear strength prediction model are obtained by simulation optimization through an improved sparrow search algorithm, which has the characteristics of fast convergence speed and high prediction result accuracy.

[0004] A method for predicting the shear strength of a rock joint wall structure surface, comprising:

[0005] Acquire rock property data sets;

[0006] The rock property dataset is sent to the shear strength prediction model for processing, and the shear strength of the rock joint wall structural surface is output. The shear strength prediction model is established based on the XGBoost model, and the hyperparameters in the shear strength prediction model are simulated and optimized through the improved sparrow search algorithm.

[0007] Preferably, the shear strength prediction model is trained, specifically comprising the following steps:

[0008] A plurality of shear strength prediction samples are obtained, wherein the shear strength prediction samples include a rock property data set and the shear strength of the corresponding rock joint wall structural surface, and all shear strength prediction samples are divided into a shear strength prediction training set and a shear strength prediction test set; a shear strength prediction model is constructed and trained based on the shear strength prediction training set, and a learning rate of the shear strength prediction model training is adjusted based on the shear strength prediction test set to determine whether the training conditions are met; if the training conditions are met, the prediction accuracy of the shear strength prediction model is calculated through the shear strength prediction test set to determine whether the prediction accuracy of the shear strength prediction model is higher than a preset value; if the prediction accuracy of the shear strength prediction model is higher than the preset value, the trained shear strength prediction model is output; if the prediction accuracy of the shear strength prediction model is not higher than the preset value, the shear strength prediction model is continued to be constructed and trained based on the shear strength prediction training set; if the training conditions are not met, the shear strength prediction model is continued to be constructed and trained based on the shear strength prediction training set.

[0009] Preferably, adjusting the learning rate of the shear strength prediction model training based on the shear strength prediction test set specifically includes the following steps:

[0010] Each time the construction of the decision tree in the shear strength prediction model is completed, it is considered that an iteration of the shear strength prediction model training is completed, and the prediction error of the shear strength prediction model after the current iteration is calculated based on the shear strength prediction test set;

[0011] The learning rate of the shear strength prediction model training is iteratively updated using the following formula:

[0012] q(g+1)=q(g)·e η[δ(g-1)-δ(g)]

[0013] Among them, q(g) is the learning rate of the shear strength prediction model in the current iteration, q(g+1) is the learning rate of the shear strength prediction model in the next iteration, δ(g) is the prediction error of the shear strength prediction model after the current iteration, δ(g-1) is the prediction error of the shear strength prediction model after the previous iteration, and η is the adjustment coefficient.

[0014] Preferably, the hyperparameters in the shear strength prediction model are obtained by simulation optimization using the improved sparrow search algorithm, which specifically includes the following steps:

[0015] Step A1: Construct N sparrow individuals H n , where n = 1, 2, 3, ..., N, sparrow individuals H n Included in {x n,1 , x n,2 , x n,3 , …, x n,j , …, x n,d}, where j = 1, 2, 3, ..., d, d is the total number of hyperparameters, x n,j represents the nth sparrow individual H n The jth hyperparameter in; N sparrow individuals H n Form a population set and set the maximum number of iterations T;

[0016] Step A2: Let t = 1, where t is used to record the number of iterations;

[0017] Step A3: Calculate H for each sparrow in the population n The corresponding fitness f n , all sparrow individuals H in the population set n According to the corresponding fitness f n Arrange them from large to small and set the fitness f n The largest one is recorded as sparrow individual H g , the fitness f n The smallest one is called sparrow individual H worst , and then sort all the sparrow individuals H according to the division ratio n Divide the former into a discoverer set and the latter into a follower set;

[0018] Step A4: For all sparrow individuals H in the discoverer set n Update, the update formula is as follows:

[0019]

[0020] in, is the nth sparrow individual H in the tth iteration n is the jth hyperparameter in, α is any random number in (0,1); R2 is the warning value, which is in the interval [0,1]; ST is the safety value, which is in the interval [0.5,1]; Q is a random number that follows a normal distribution; L is a 1×d matrix with all elements being 1;

[0021] Step A5: Calculate all sparrow individuals H in the updated discoverer set n The fitness f n , the fitness value f in the discoverer set n The largest sparrow n Denoted as sparrow individual H p , for all sparrow individuals H in the follower set n Update, the update formula is as follows:

[0022]

[0023] in, For sparrow individual Hp The jth hyperparameter in , x worst For sparrow individual H worst is the jth hyperparameter in ; s is the step size, which is determined based on the Levy flight strategy; A+ is a 1×d matrix, which randomly contains two elements -1 and 1; k is the number of sparrow individuals H in the follower set n According to the fitness f n The corresponding serial number value after arranging from large to small, K is all the sparrow individuals H in the follower set n Total number of;

[0024] Step A6: Select sparrow individuals H from the population set one by one n , for the selected sparrow individual H n , execute the following content, generate a random number ζ between 0 and 1 through a random function, and judge "ζ>P c "Is it true? c is the transformation threshold, if "ζ>P c " is not true, no operation; otherwise, the selected sparrow individual H is calculated by the following formula n To update:

[0025]

[0026] Where U is a random number between -1 and 1; β is the step size control parameter; f g For sparrow individual H g The corresponding fitness, f worst For sparrow individual H worst The corresponding fitness, x g For sparrow individual H worst The jth hyperparameter in ;

[0027] Step A7: Calculate H for each sparrow in the population n The corresponding fitness f n , all sparrow individuals H in the population set n According to the corresponding fitness f n Arrange them from large to small and set the fitness f n The largest one is recorded as sparrow individual H g , and the sparrow individual H g Store it in the optimal sparrow individual database; set the fitness f n The smallest one is called sparrow individual H worst , and then sort all the sparrow individuals H according to the division ratio n Divide them; the former after division form the discoverer set, and the latter after division form the follower set, and enter A8:

[0028] Step A8: Determine whether "t<T" holds. If so, assign t+1 to t and return to A4; otherwise, select the sparrow individual H with the largest fitness value in the optimal sparrow individual database. n As a hyperparameter in the shear strength prediction model.

[0029] Preferably, the sparrow individual H n The corresponding fitness f n The calculation method is as follows:

[0030] Based on sparrow individual H n The hyperparameters in the shear strength prediction model are set, and the accuracy of the shear strength prediction model with the set hyperparameters is obtained through the shear strength prediction test set. The accuracy of the shear strength prediction model with the set hyperparameters is used as the sparrow individual H n The corresponding fitness f n .

[0031] Preferably, the step length s is determined based on the Levy flight strategy, including the following formula:

[0032] s=μ / |v| 1 / β ,β=3 / 2…(1)

[0033]

[0034] The step size s is a random number determined by formula 1-3.

[0035] Preferably, the shear strength prediction model is initialized by transfer learning.

[0036] The present invention also provides a shear strength prediction system for rock joint wall structural surface, comprising:

[0037] A rock property data set acquisition module, used to acquire a rock property data set;

[0038] The shear strength prediction module is used to send the rock property data set into the shear strength prediction model for processing, and output the shear strength of the rock joint wall structural surface. The shear strength prediction model is established based on the XGBoost model, and the hyperparameters in the shear strength prediction model are simulated and optimized through the improved sparrow search algorithm.

[0039] The present invention has the following advantages:

[0040] 1. The present invention predicts the shear strength of the rock joint wall structural surface through a shear strength prediction model established based on the XGBoost model. It can predict the shear strength of the rock joint wall structural surface based on conventional rock mechanical property input without the need for additional complicated test experiments; and the hyperparameters of the shear strength prediction model are obtained by simulation optimization through the improved sparrow search algorithm, which has the characteristics of fast convergence speed and high prediction result accuracy.

[0041] 2. The present invention adjusts the learning rate of the shear strength prediction model, so that the learning rate can be reduced when the prediction error increases, reducing the risk of overfitting, thereby stabilizing the training process. Conversely, if the prediction error decreases, increasing the learning rate can speed up the training process.

[0042] 3. The present invention improves the sparrow search algorithm by introducing the Lévy flight strategy, which can improve the global search capability of the simulation optimization algorithm, thereby converging to the global optimal solution more quickly. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 The present invention is a schematic structural diagram of a system for predicting the shear strength of a rock joint wall structure surface used in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0045] Embodiment 1, a method for predicting the shear strength of a rock joint wall structure surface, comprising:

[0046] Obtain rock property data sets, including the pre-processed joint wall compressive strength ratio, dynamic interface joint strength shear strength, JRC (joint roughness) and normal stress, etc. These data play an important role in the prediction of the shear strength of the rock joint wall structural surface;

[0047] The rock property data set is sent to the shear strength prediction model for processing, and the shear strength of the rock joint wall structure surface is output. The shear strength prediction model is established based on the XGBoost model, and the hyperparameters in the shear strength prediction model are simulated and optimized by the improved sparrow search algorithm.

[0048] It should be noted that the shear strength prediction model is a data-driven model, and its effect depends largely on the determination of hyperparameters. Once the hyperparameters of the shear strength prediction model are set reasonably, the prediction performance of the shear strength prediction model can be significantly improved and the computer computing resources occupied will be effectively reduced.

[0049] This application predicts the shear strength of the rock joint wall structural surface through a shear strength prediction model established based on the XGBoost model. It can predict the shear strength of the rock joint wall structural surface based on conventional rock mechanical property inputs without the need for additional complicated test experiments; and the hyperparameters of the shear strength prediction model are obtained through simulation optimization using an improved sparrow search algorithm, which has the characteristics of fast convergence speed and high prediction result accuracy.

[0050] Training the shear strength prediction model includes the following steps:

[0051] Several shear strength prediction samples are obtained, and the shear strength prediction samples include a rock property data set and the shear strength of the corresponding rock joint wall structural surface. The data in the rock property data set vary in a large range and can reflect the different changes of the rock joint wall structural surface. All shear strength prediction samples are divided into a shear strength prediction training set and a shear strength prediction test set. A shear strength prediction model is constructed and trained based on the shear strength prediction training set. The shear strength prediction model is initialized by transfer learning. The construction steps are to continuously add decision tree models. During the training, the residual between the model prediction value and the actual value is used as the target, and the learning rate of the shear strength prediction model training is adjusted based on the shear strength prediction test set. Determine whether the training conditions are met. The training conditions can be that the number of decision tree models reaches a preset number or the prediction error is within a preset range. If the training conditions are met, the prediction accuracy of the shear strength prediction model is calculated through the shear strength prediction test set to determine whether the prediction accuracy of the shear strength prediction model is higher than the preset value. The preset value is set by the developer. If the prediction accuracy of the shear strength prediction model is higher than the preset value, the trained shear strength prediction model is output. If the prediction accuracy of the shear strength prediction model is not higher than the preset value, the shear strength prediction model is continued to be constructed and trained based on the shear strength prediction training set; if the training conditions are not met, the shear strength prediction model is continued to be constructed and trained based on the shear strength prediction training set.

[0052] Adjust the learning rate of shear strength prediction model training based on the shear strength prediction test set, which specifically includes the following steps:

[0053] Each time the construction of the decision tree in the shear strength prediction model is completed, it is considered that an iteration of the shear strength prediction model training is completed. The prediction error of the shear strength prediction model after the current iteration is calculated based on the shear strength prediction test set. The prediction error here is the difference between the model prediction value and the actual value;

[0054] The learning rate of the shear strength prediction model training is iteratively updated using the following formula:

[0055] q(g+1)=q(g)·e η[δ(g-1)-δ(g)]

[0056] Wherein, q(g) is the learning rate of the shear strength prediction model in the current iteration, q(g+1) is the learning rate of the shear strength prediction model in the next iteration, δ(g) is the prediction error of the shear strength prediction model after the current iteration, δ(g-1) is the prediction error of the shear strength prediction model after the previous iteration, and η is the adjustment coefficient, which is determined a priori by the change in the prediction error of the shear strength prediction model;

[0057] By adjusting the learning rate of the shear strength prediction model through the above formula, it is possible to reduce the learning rate when the prediction error increases, reduce the risk of overfitting, and thus stabilize the training process. Conversely, if the prediction error decreases, increasing the learning rate can speed up the training process.

[0058] The hyperparameters in the shear strength prediction model are obtained by simulation optimization using the improved sparrow search algorithm, which specifically includes the following steps:

[0059] Step A1: Construct N sparrow individuals H n , where n = 1, 2, 3, ..., N, sparrow individuals H n Included in {x n,1 , x n,2 , x n,3 , …, x n,j , …, x n,d}, where j = 1, 2, 3, ..., d, d is the total number of hyperparameters, x n,j represents the nth sparrow individual H n The jth hyperparameter in the shear strength prediction model includes the preset number of decision trees and the learning rate. n It should be noted that when constructing the sparrow individual H n In the process, random values ​​are taken within the value range corresponding to each hyperparameter; the maximum number of iterations T is set;

[0060] Step A2: Let t = 1, where t is used to record the number of iterations;

[0061] Step A3: Calculate H for each sparrow in the population n The corresponding fitness f n , all sparrow individuals H in the population set n According to the corresponding fitness f n Arrange them from large to small and set the fitness f n The largest one is recorded as sparrow individual H g , the fitness f n The smallest one is called sparrow individual H worst, and then sort all the sparrow individuals H according to the division ratio n Divide and set the division ratio artificially, usually 2:8; the former after division form the discoverer set, and the latter after division form the follower set;

[0062] Sparrow individual H n The corresponding fitness f n The calculation method is as follows:

[0063] Based on sparrow individual H n The hyperparameters in the shear strength prediction model are set, and the accuracy of the shear strength prediction model with the set hyperparameters is obtained through the shear strength prediction test set. The accuracy of the shear strength prediction model with the set hyperparameters is used as the sparrow individual H n The corresponding fitness f n ;

[0064] Step A4: For all sparrow individuals H in the discoverer set n Update, the update formula is as follows:

[0065]

[0066] in, is the nth sparrow individual H in the tth iteration n is the jth hyperparameter in, α is any random number in (0,1); R2 is the warning value, which is in the interval [0,1]; ST is the safety value, which is in the interval [0.5,1]; Q is a random number that follows a normal distribution; L is a 1×d matrix with all elements being 1;

[0067] Step A5: Calculate all sparrow individuals H in the updated discoverer set n The fitness f n , the fitness value f in the discoverer set n The largest sparrow n Denoted as sparrow individual H p , for all sparrow individuals H in the follower set n Update, the update formula is as follows:

[0068]

[0069] in, For sparrow individual H p The jth hyperparameter in , x worst For sparrow individual H worst is the jth hyperparameter in ; s is the step size, which is determined based on the Levy flight strategy; A+ is a 1×d matrix, which randomly contains two elements -1 and 1; k is the number of sparrow individuals H in the follower set nAccording to the fitness f n The corresponding serial number value after arranging from large to small, K is all the sparrow individuals H in the follower set n Total number of;

[0070] Step A6: Select sparrow individuals H from the population set one by one n , for the selected sparrow individual H n , execute the following content, generate a random number ζ between 0 and 1 through a random function, and judge "ζ>P c "Is it true? c is the transformation threshold, which is set by the user and is generally 0.7. c " is not true, no operation; otherwise, the selected sparrow individual H is calculated by the following formula n To update:

[0071]

[0072] Where U is a random number between -1 and 1; β is the step size control parameter; f g For sparrow individual H g The corresponding fitness, f worst For sparrow individual H worst The corresponding fitness, x g For sparrow individual H worst The jth hyperparameter in ;

[0073] Step A7: Calculate H for each sparrow in the population n The corresponding fitness f n , all sparrow individuals H in the population set n According to the corresponding fitness f n Arrange them from large to small and set the fitness f n The largest one is recorded as sparrow individual H g , and the sparrow individual H g Store it in the optimal sparrow individual database; set the fitness f n The smallest one is called sparrow individual H worst , and then sort all the sparrow individuals H according to the division ratio n Divide and set the division ratio manually, usually 2:8; the former after division form the discoverer set, and the latter after division form the follower set, and enter A8:

[0074] Step A8: Determine whether "t<T" holds. If so, it means that the maximum number of iterations has not been reached. Assign t+1 to t and return to A4. Otherwise, select the sparrow individual H with the largest fitness value in the optimal sparrow individual database. n As a hyperparameter in the shear strength prediction model.

[0075] The step length s is determined based on the Levy flight strategy, including the following formula:

[0076] s=μ / |v| 1 / β ,β=3 / 2…(1)

[0077]

[0078] The step length s is a random number determined by formula 1-3. Since the convergence method of the sparrow search algorithm is to jump directly to the vicinity of the current optimal solution, it is easy to fall into the local optimum. By introducing the step length s of the Lévy flight strategy, the global search capability of the simulation optimization algorithm is improved, so that it can converge to the global optimal solution faster.

[0079] Embodiment 2, a shear strength prediction system for rock joint wall structure surface, such as Figure 1 As shown, including:

[0080] The rock property data set acquisition module is used to acquire the rock property data set, which includes the pre-processed joint wall compressive strength ratio, the shear strength of the dynamic interface joint strength, JRC (joint roughness) and normal stress, etc. These data play an important role in the prediction of the shear strength of the rock joint wall structural surface;

[0081] The shear strength prediction module is used to send the rock property data set into the shear strength prediction model for processing, and output the shear strength of the rock joint wall structural surface. The shear strength prediction model is established based on the XGBoost model, and the hyperparameters in the shear strength prediction model are simulated and optimized through the improved sparrow search algorithm.

[0082] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A method for predicting the shear strength of rock joint wall structural surfaces. It is characterized in that include: Acquire rock property data sets; The rock property data set is sent to the shear strength prediction model for processing, and the shear strength of the rock joint wall structural surface is output. The shear strength prediction model is established based on the XGBoost model, and the hyperparameters in the shear strength prediction model are simulated and optimized through the improved sparrow search algorithm.

2. A method for predicting the shear strength of a rock joint wall structure surface according to claim 1, It is characterized in that Training the shear strength prediction model includes the following steps: A plurality of shear strength prediction samples are obtained, wherein the shear strength prediction samples include a rock property data set and the shear strength of the corresponding rock joint wall structural surface, and all shear strength prediction samples are divided into a shear strength prediction training set and a shear strength prediction test set; a shear strength prediction model is constructed and trained based on the shear strength prediction training set, and a learning rate of the shear strength prediction model training is adjusted based on the shear strength prediction test set to determine whether the training conditions are met; if the training conditions are met, the prediction accuracy of the shear strength prediction model is calculated through the shear strength prediction test set to determine whether the prediction accuracy of the shear strength prediction model is higher than a preset value; if the prediction accuracy of the shear strength prediction model is higher than the preset value, the trained shear strength prediction model is output; if the prediction accuracy of the shear strength prediction model is not higher than the preset value, the shear strength prediction model is continued to be constructed and trained based on the shear strength prediction training set; if the training conditions are not met, the shear strength prediction model is continued to be constructed and trained based on the shear strength prediction training set.

3. A method for predicting the shear strength of a rock joint wall structure surface according to claim 2, It is characterized in that The learning rate of shear strength prediction model training is adjusted based on the shear strength prediction test set, which specifically includes the following steps: Each time the construction of the decision tree in the shear strength prediction model is completed, it is considered that an iteration of the shear strength prediction model training is completed, and the prediction error of the shear strength prediction model after the current iteration is calculated based on the shear strength prediction test set; The learning rate of the shear strength prediction model training is iteratively updated using the following formula: q(g+1)=q(g)·e η[δ(g-1)-δ(g)] Among them, q(g) is the learning rate of the shear strength prediction model in the current iteration, q(g+1) is the learning rate of the shear strength prediction model in the next iteration, δ(g) is the prediction error of the shear strength prediction model after the current iteration, δ(g-1) is the prediction error of the shear strength prediction model after the previous iteration, and η is the adjustment coefficient.

4. A method for predicting the shear strength of a rock joint wall structure surface according to claim 3, It is characterized in that The hyperparameters in the shear strength prediction model are obtained by simulation optimization using the improved sparrow search algorithm, which specifically includes the following steps: Step A1: Construct N sparrow individuals H n , where n = 1, 2, 3, ..., N, sparrow individuals H n Include {x n,1 , x n,2 , x n,3 , …, x n,j , …, x n,d }, where j = 1, 2, 3, ..., d, d is the total number of hyperparameters, x n,j represents the nth sparrow individual H n The jth hyperparameter in; N sparrow individuals H n Form a population set and set the maximum number of iterations T; Step A2: Let t = 1, where t is used to record the number of iterations; Step A3: Calculate H for each sparrow in the population n The corresponding fitness f n , all sparrow individuals H in the population set n According to the corresponding fitness f n Arrange them from large to small and set the fitness f n The largest one is recorded as sparrow individual H g , the fitness f n The smallest one is called sparrow individual H worst , and then sort all the sparrow individuals H according to the division ratio n Divide the former into a discoverer set and the latter into a follower set; Step A4: For all sparrow individuals H in the discoverer set n Update, the update formula is as follows: in, is the nth sparrow individual H in the tth iteration n is the jth hyperparameter in, α is any random number in (0, 1); R2 is the warning value, which is in the interval [0, 1]; ST is the safety value, which is in the interval [0.5, 1]; Q is a random number that follows a normal distribution; L is a 1×d matrix with all elements being 1; Step A5: Calculate all sparrow individuals H in the updated discoverer set n The fitness f n , the fitness value f in the discoverer set n The largest sparrow n Denoted as sparrow individual H p , for all sparrow individuals H in the follower set n Update, the update formula is as follows: in, For sparrow individual H p The jth hyperparameter in x worst For sparrow individual H worst is the jth hyperparameter in ; s is the step size, which is determined based on the Levy flight strategy; A+ is a 1×d matrix, which randomly contains two elements -1 and 1; k is the number of sparrow individuals H in the follower set n According to the fitness f n The corresponding serial number value after arranging from large to small, K is all the sparrow individuals H in the follower set n Total number of; Step A6: Select sparrow individuals H from the population set one by one n , for the selected sparrow individual H n , execute the following content, generate a random number ζ between 0 and 1 through a random function, and judge "ζ>P c "Is it true? c is the transformation threshold, if "ζ>P c " is not true, no operation; otherwise, the selected sparrow individual H is calculated by the following formula n To update: Where U is a random number between -1 and 1; β is the step size control parameter; f g For sparrow individual H g The corresponding fitness, f worst For sparrow individual H worst The corresponding fitness, x g For sparrow individual H worst The jth hyperparameter in ; Step A7: Calculate H for each sparrow in the population n The corresponding fitness f n , all sparrow individuals H in the population set n According to the corresponding fitness f n Arrange them from large to small and set the fitness f n The largest one is recorded as sparrow individual H g , and the sparrow individual H g Store it in the optimal sparrow individual database; set the fitness f n The smallest one is called sparrow individual H worst , and then sort all the sparrow individuals H according to the division ratio n Divide them; the former after division form the discoverer set, and the latter after division form the follower set, and enter A8: Step A8: Determine whether "t < T" holds. If "t < T" holds, assign t + 1 to t and return to A4; otherwise, select the sparrow individual H with the largest fitness value in the optimal sparrow individual library n as a hyperparameter in the shear strength prediction model.

5. A method for predicting the shear strength of a rock joint wall structure surface according to claim 4, It is characterized in that Sparrow individual H n The corresponding fitness f n The calculation method is as follows: Based on sparrow individual H n The hyperparameters in the shear strength prediction model are set, and the accuracy of the shear strength prediction model with the set hyperparameters is obtained through the shear strength prediction test set. The accuracy of the shear strength prediction model with the set hyperparameters is used as the sparrow individual H n The corresponding fitness f n .

6. A method for predicting the shear strength of a rock joint wall structure surface according to claim 5, It is characterized in that The step length s is determined based on the Levy flight strategy, including the following formula: s=μ / |v| 1 / β ,β=3 / 2…(1) The step size s is a random number determined by formula 1-3.

7. A method for predicting the shear strength of a rock joint wall structure surface according to claim 6, It is characterized in that The shear strength prediction model is initialized by transfer learning.

8. A shear strength prediction system for rock joint wall structural surfaces. It is characterized in that The system uses a method for predicting the shear strength of a rock joint wall structure surface according to any one of claims 1 to 7, comprising: A rock property data set acquisition module, used to acquire a rock property data set; The shear strength prediction module is used to send the rock property data set into the shear strength prediction model for processing, and output the shear strength of the rock joint wall structural surface. The shear strength prediction model is established based on the XGBoost model, and the hyperparameters in the shear strength prediction model are simulated and optimized through the improved sparrow search algorithm.

Citation Information

Patent Citations

  • New and old concrete interface shear strength calculation method based on XGBoost

    CN113723004A

  • Transverse wave time difference prediction method based on SSA-ELM algorithm

    CN114488311A

  • Air quality index prediction method based on improved sparrow search algorithm optimization

    CN115374689A

Cited By

  • Rock shear strength parameter machine learning prediction method and system

    CN121524831A