Tunnel blasting vibration velocity prediction method and system fused with multi-factor intelligent optimization

Abstract

The invention belongs to the technical field of intelligent tunnel engineering, and discloses a tunnel blasting vibration velocity prediction method and system fusing multi-factor intelligent optimization. The method comprises the following steps: acquiring blasting parameters and historical vibration monitoring data; the method comprises the following steps: constructing a data set containing multi-factor characteristics, initializing XGBoost hyper-parameters, carrying out iterative optimization on the XGBoost hyper-parameters by using an SSA algorithm, inputting training set data to train an optimized SSA-XGBoost model, verifying model performance by using a test set, and predicting the performance by comparing with a traditional model; and dynamically adjusting single-section explosive quantity parameters in real time according to the predicted value, generating risk early warning in combination with a safety threshold, and realizing closed-loop optimization of blasting design. According to the method, real-time risk early warning is generated in combination with the safety threshold value and the SSA-XGBoost model, closed-loop optimization of blasting design is achieved by dynamically adjusting blasting parameters, and construction efficiency is improved while construction safety is ensured.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent tunnel engineering, and particularly relates to a tunnel blasting vibration velocity prediction method and system integrating multi-factor intelligent optimization. Background Technique

[0002] In the field of tunnel engineering, the control of blasting vibration velocity is the core link to ensure construction safety, the stability of surrounding building structures, and the quality of life of surrounding residents. Among them, the effective prediction of the peak vibration velocity is of utmost importance. In the past, the prediction methods mainly relied on traditional empirical formulas - the Sadovskii formula or single machine learning models, which were prone to problems such as poor parameter generalization and the inability to reflect the influence of multi-factor coupling. In addition, there was also prediction through numerical simulation calculations, which was highly complex, time-consuming in calculation, and difficult to meet the needs of actual projects. Recently, machine learning techniques have been gradually applied to the engineering field, such as BP neural networks, support vector machines, etc. However, their prediction accuracy is often limited by the single model structure, insufficient parameter optimization, especially the lack of dynamic correlation analysis under multi-factor coupling, and the failure to fully utilize intelligent algorithms to optimize the model performance, resulting in overfitting or underfitting phenomena.

[0003] The XGBoost algorithm, with its integrated advantages of gradient boosting decision trees, shows excellent prediction efficiency and accuracy in non-linear data modeling. However, the performance of this model largely depends on the setting of hyperparameters. Traditional hyperparameter optimization methods are difficult to find the global optimal solution. The Sparrow Search Algorithm (SSA), as a new type of swarm intelligence optimization algorithm, constructs a heuristic search strategy by simulating the foraging behavior of biological groups, has powerful global search capabilities and fast convergence characteristics, and can effectively optimize the hyperparameters of the XGBoost model. This fusion strategy not only significantly reduces the deviation of manual parameter tuning but also provides high-precision intelligent decision-making support for blasting vibration control. Summary of the Invention

[0004] To overcome the problems existing in the related technologies, the disclosed embodiments of the present invention provide a tunnel blasting vibration velocity prediction method and system integrating multi-factor intelligent optimization. Specifically, it relates to an intelligent prediction method and system for the peak value of tunnel blasting vibration velocity based on SSA optimization.

[0005] The technical solution is as follows: A tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization includes the following steps: S1, collecting blasting parameters and historical vibration monitoring data. The blasting parameters include the maximum charge per delay, the distance from the blast center, the drilling depth, and the uniaxial compressive strength of the rock; S2, performing normalization processing on the data to construct a data set containing multi-factor features. This data set includes a training set, a validation set, and a test set. Among them, the ratios of the training set, the validation set, and the test set are 8:1:1; S3. Initialize the XGBoost hyperparameters, and use the SSA algorithm to iteratively optimize the XGBoost hyperparameters to complete the construction of the SSA-XGBoost model; S4. Input the training set data to train the optimized SSA-XGBoost model, and use the test set to verify the model performance and compare the prediction performance of the traditional model; S5. Dynamically adjust the single-section charge parameter in real time according to the predicted value, and generate a risk warning in combination with the safety threshold to achieve the closed-loop optimization of the blasting design.

[0006] In step S1, collect the blasting parameters and historical vibration monitoring data, including: in the multi-source data collection, for the blasting vibration monitoring, use a blasting vibration meter to simultaneously measure the three components perpendicular to each other of the particle vibration, and perform vector synthesis on the vibration velocity components in the three directions to obtain the total vibration velocity , which is used as the output variable for model training, and the calculation formula is: ; In the formula, is the maximum radial component of the vibration velocity, is the maximum tangential component of the vibration velocity, is the maximum vertical component of the vibration velocity.

[0007] In step S3, initialize the XGBoost hyperparameters, including: The objective function of XGBoost consists of a loss function and a regularization term, and the formula is: ; In the formula, is the objective function of XGBoost, is the loss function of the model bias, is the actual value, is the predicted value, is the total number of samples in the dataset, is the regularization term of the th tree, is the total number of trees, is the basis function; ; In the formula, is the parameter for controlling the tree structure, is the number of leaf nodes, is the weight of the first leaf node, is the weight of the second leaf node, is the set composed of the scores of the leaf nodes of each tree, is the L1 norm, is the L2 norm; L1 regularization automatically screens the blast center distance and maximum charge characteristics in the blast parameters, and L2 regularization prevents the model from overfitting; ; In the formula, is the objective function value at the th step, is the sum of the predicted values of the first trees, is the predicted value of the regression tree.

[0008] Furthermore, the SSA algorithm is used to iteratively optimize the XGBoost hyperparameters, including: Define the parameters for optimizing the XGBoost model, including the number of iterations n_estimators, the maximum depth max_depth, and the learning rate learning_rate; Set the parameters of the sparrow search algorithm, including the number of sparrows and the maximum number of iterations, and define the fitness function; Simulate the foraging process and update the positions of the discoverers, followers, and vigilant ones; The calculation formula for updating the position of the discoverer is: ; In the formula, is the position information of the th sparrow in the th dimension, is the exponential function, is the number of iterations; is a random value, ; is the maximum number of iterations, is the early warning value, ; is a random number that follows a standard normal distribution, is a -dimensional matrix with a value of 1; is the safety threshold, ; When , it means that the discoverer is in a safe position for searching; When , it means that the discoverer has been discovered by the predator and needs to fly to other safe positions for searching.

[0009] Furthermore, the calculation formula for updating the position of the follower is: ; In the formula, is the exponential function, ​is the globally worst position, is the number of sparrows, is the best position where the discoverer is currently located; is a vector of dimension When it represents that the foraging situation of the th follower is not ideal and the position needs to be updated; When the follower can move to the best position.

[0010] Furthermore, the calculation formula for updating the position of the vigilant is: ; In the formula, is the current best position, is a random number subject to the standard normal distribution; is a random number, ; is the fitness value of the current sparrow individual, and are the current global best and worst fitness values respectively, is the minimum constant used to avoid a zero denominator.

[0011] In step S4, for the SSA-XGBoost model, the root mean square error RMSE, mean absolute error MAE, and coefficient of determination R 2 are used for evaluation.

[0012] In step S5, the safety threshold is determined by introducing geological feature variables and combining with Sadovsky's empirical formula to determine the safety threshold of blasting vibration velocity. The safety threshold formula is as follows: ; In the formula, is the vibration velocity safety threshold, is the coefficient determined according to geological conditions and blasting experience, is the maximum charge per single section, is the distance from the blast center, is the weight coefficient of geological features, is the normalization function of the fracture density geological feature; ; In the formula, is the normalization function based on rock hardness, is the normalization function based on fracture density.

[0013] In step S5, the closed-loop optimization of blasting design is realized, including: predicting the vibration velocity of the future construction section through the trained SSA-XGBoost model to optimize the blasting parameters, and simultaneously monitoring and evaluating the blasting effect.

[0014] Another object of the present invention is to provide a tunnel blasting vibration velocity prediction system integrating multi-factor intelligent optimization. This system implements the tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization, and this system includes: A multi-source data acquisition module, which is used to collect blasting parameters and historical vibration monitoring data. The blasting parameters include the maximum charge per delay, the distance from the blast center, the drilling depth, and the rock compressive strength; A data preprocessing module, which is used to perform normalization processing on the data and construct a data set containing multi-factor features. This data set includes a training set, a validation set, and a test set. Among them, the ratios of the training set, the validation set, and the test set are 8:1:1; An SSA-XGBoost model building module, which is used to iteratively optimize the XGBoost hyperparameters using the SSA algorithm to complete the construction of the SSA-XGBoost model; A model training and validation module, which is used to input the training set data to train the optimized SSA-XGBoost model, use the test set to verify the model performance and compare the prediction performance of the traditional model; An engineering application module, which is used to dynamically adjust the charge per delay parameter in real time according to the predicted value, generate a risk warning in combination with the safety threshold, and realize the closed-loop optimization of blasting design.

[0015] Combining all the above technical solutions, the beneficial effects of the present invention are as follows: The present invention combines the global optimization ability of the sparrow search algorithm (SSA) and the powerful prediction ability of XGBoost through the SSA-XGBoost model, effectively capturing the complex relationships between data. In the test set, the R² value of the SSA-XGBoost model reaches 0.973, showing a significant improvement compared with the traditional BP neural network (R² = 0.819) and BP-PSO (R² = 0.926). It proves that it can effectively improve the prediction accuracy of blasting vibration velocity.

[0016] The SSA-XGBoost model in the present invention shows good robustness under different data volumes and data distributions. Even in the case of a small data volume or uneven data distribution, it can still maintain a high prediction accuracy, while traditional models often have large errors. For example, in the test set, the MAE of SSA-XGBoost is 2.419, which is much lower than 5.112 of the traditional BP and 2.503 of BP-PSO.

[0017] The present invention forms a new safety threshold formula by improving the Sadovsky empirical formula, and makes it applicable to the vibration velocity safety control under different geological conditions by introducing a geological feature vector. Combining the safety threshold and the SSA-XGBoost model to generate real-time risk warnings, and realizing the closed-loop optimization of blasting design by dynamically adjusting blasting parameters, which not only ensures construction safety but also improves construction efficiency.

[0018] The present invention effectively improves construction safety and efficiency. Using SSA-XGBoost can achieve high-precision prediction of blasting vibration velocity, can dynamically adjust parameters such as the charge amount per section, reduce the risk of vibration velocity exceeding the limit, reduce the impact on surrounding buildings and residents, and save potential compensation costs; compared with traditional numerical simulations, the working efficiency of this system is significantly improved, the blasting design cycle can be shortened and the labor cost can be reduced; the commercial value is applicable to blasting scenarios such as tunnel engineering and mine exploitation, and there is a clear market demand. Combining intelligent warning and closed-loop optimization can develop related hardware integration systems with high added value.

[0019] In existing research, SSA-XGBoost has been tried in many fields such as mechanics and medicine. However, in the prediction of tunnel blasting vibration velocity, for the first time, it combines multi-source data (geology, charge amount, distance from the blast center, etc.) with a dynamic closed-loop optimization mechanism to solve the problem that the traditional Sadovsky formula cannot quantify the coupling effects of multiple factors; the "prediction-monitoring-feedback" closed-loop mechanism of the present invention realizes the leap from static prediction to dynamic control, filling the technical gap in engineering practice.

[0020] The Sadovsky formula relies on empirical coefficients and cannot reflect the coupling effects of multiple factors (such as geological conditions, charge amount, distance from the blast center, etc.). The present invention considers the integration of multiple factors to solve the problem of poor parameter generalization. While traditional models (such as BP neural networks) are prone to overfitting when the data is scarce (the MAE of the test set reaches 5.112), the present invention through data normalization and SSA optimization enables the model to still maintain high precision (R²>0.97) even in small samples or unbalanced data. Although SSA-XGBoost has been applied in other fields, its application in the field of tunnel blasting vibration velocity prediction is less and the effect is poor. The present invention verifies its feasibility through specialized hyperparameter optimization (such as introducing features such as distance from the blast center and rock strength), and is significantly superior to traditional models (the RMSE of the test set drops from 4.276 to 3.362). Description of the Drawings

[0021] The drawings here are incorporated into the specification and form a part of this specification, showing the embodiments in line with the present disclosure, and are used together with the specification to explain the principles of the present disclosure; Figure 1It is the schematic diagram of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 2 It is the flow chart of the tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 3 It is the density map of the training set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 4 It is the result map of the training set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 5 It is the density map of the test set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 6 It is the result map of the test set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 7 It is the density map of the test set of the BP model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization of the present invention; Figure 8 It is the density map of the test set of the BP-PSO model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization of the present invention; Figure 9 It is the density map of the test set of the XGBoost model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization of the present invention; Figure 10 It is the density result map of the training set of the BP model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization of the present invention; Figure 11 It is the density result map of the training set of the BP-PSO model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization of the present invention; Figure 12 It is the density result map of the training set of the XGBoost model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) integrating multi-factor intelligent optimization of the present invention; Figure 13 It is the schematic diagram of the tunnel blasting vibration velocity prediction system integrating multi-factor intelligent optimization provided by the embodiments of the present invention; In the figure: 1. Multi-source data acquisition module; 2. Data preprocessing module; 3. SSA-XGBoost model building module; 4. Model training and verification module; 5. Engineering application module. Specific implementation manners

[0022] To make the above objects, features, and advantages of the present invention more obvious and understandable, the specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description in order to fully understand the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific implementations disclosed below.

[0023] The innovation of the present invention lies in: the present invention integrates the Sparrow Search Algorithm (SSA) into the XGBoost model, and uses its global search ability to optimize the hyperparameters (number of iterations, learning rate, etc.) of XGBoost, thereby improving the prediction accuracy; through multi-source data acquisition (blasting parameters, geological conditions, vibration monitoring data) and normalization processing, a highly generalized dataset is constructed to solve the problem of poor parameter generalization under the influence of multi-factor coupling. The SSA-XGBoost model still maintains a low error (MAE is only 2.419) in small samples or unbalanced data, and its robustness is better than that of traditional models; geological feature variables are introduced and combined with Sadovsky's empirical formula to improve the safety threshold, and the prediction results are combined with the safety threshold to generate risk warnings in real time, dynamically adjust parameters such as the single-section charge amount, and feedback the monitoring data to iteratively optimize the model, forming a closed-loop control of blasting design - prediction - monitoring - feedback, paying attention to construction safety while improving construction efficiency.

[0024] Example 1, as Figure 1 shown, the principle of the tunnel blasting vibration velocity prediction method with multi-factor intelligent optimization provided by the embodiment of the present invention.

[0025] Specifically, as Figure 2 shown, the tunnel blasting vibration velocity prediction method with multi-factor intelligent optimization provided by the embodiment of the present invention includes the following steps: S1, collect blasting parameters and historical vibration monitoring data, and the blasting parameters include the maximum single-section charge amount, the distance from the blast center, the drilling depth, and the rock compressive strength; S2, perform normalization processing on the data to construct a dataset containing multi-factor features, and this dataset includes a training set, a validation set, and a test set. Among them, the ratios of the training set, the validation set, and the test set are 8:1:1; S3, initialize the XGBoost hyperparameters, and use the SSA algorithm to iteratively optimize the XGBoost hyperparameters to complete the construction of the SSA-XGBoost model; S4. Input the training set data to train the optimized SSA-XGBoost model, use the test set to verify the model performance and compare the prediction performance of the traditional models; For example, the traditional BP neural network, BP-PSO model and the unoptimized XGBoost), evaluate the prediction errors (MAE, RMSE, etc.) and the coefficient of determination (R 2 ); S5. Dynamically adjust the single-section charge parameters in real time according to the predicted values, generate risk warnings in combination with safety thresholds, and achieve the closed-loop optimization of blasting design.

[0026] In step S1, collect the blasting parameters and historical vibration monitoring data, including: in the multi-source data collection, for blasting vibration monitoring, use the TC-4850 blasting vibration meter to simultaneously measure three components perpendicular to each other of the particle vibration, and perform vector synthesis on the vibration velocity components in the three directions to obtain the total vibration velocity , which is used as the output variable for model training, and the calculation formula is: ; In the formula, is the maximum radial component of the vibration velocity, is the maximum tangential component of the vibration velocity, is the maximum vertical component of the vibration velocity.

[0027] Exemplarily, in step S2, perform data normalization processing, standardize the data of the training set and the test set to 0-1, eliminate the influence caused by the different magnitudes and dimensions of each variable, and the division of the data set is in the ratio of training set: validation set: test set = 8:1:1; Exemplarily, in step S3, initialize the XGBoost hyperparameters, including: The basic principle of the XGBoost hyperparameters, its objective function is composed of a loss function and a regularization term, and its algorithm can be implemented according to the following formula: ; In the formula, is the objective function of XGBoost, is the loss function of the model bias, is the actual value, is the predicted value, is the total number of samples in the data set, is the th tree regularization term, is the total number of trees, is the base function; Introduce elastic net regularization (Elastic Net), combine L1 and L2 regularization: ; In the formula, is the parameter for controlling the tree structure, is the number of leaf nodes, is the weight of the first leaf node, is the weight of the second leaf node, is the set composed of the scores of the leaf nodes of each tree, is the L1 norm, is the L2 norm; L1 regularization can automatically screen key features in the blasting parameters (such as the distance to the blast center, the maximum charge amount, etc.), and L2 regularization can effectively prevent the model from overfitting.

[0028] ; In the formula, is the objective function value at the th step, is the sum of the predicted values of the first trees, is the predicted value of the regression tree.

[0029] Exemplarily, in step S3, the SSA algorithm is used to iteratively optimize the XGBoost hyperparameters, including: Defining the parameters for optimizing the XGBoost model, such as the number of iterations (n_estimators), the maximum depth (max_depth), the learning rate (learning_rate), etc.; Setting the parameters of the sparrow search algorithm, including the number of sparrows and the maximum number of iterations, and defining the fitness function; Simulating the foraging process and updating the positions of the discoverers, followers, and vigilant ones; The calculation formula for updating the position of the discoverer is: ; In the formula, is the position information of the th sparrow in the th dimension, is the exponential function, is the number of iterations; is a random value, ; is the maximum number of iterations, is the warning value, ; is a random number that follows the standard normal distribution, is a -dimensional matrix with a value of 1; is the safety threshold, ; When it means that the discoverer is searching in a safe position; When it means that the discoverer is detected by the predator and needs to fly to other safe positions for searching.

[0030] The calculation formula for updating the position of the follower is: ; In the formula, is the exponential function, is the global worst position, is the number of sparrows, is the current best position where the discoverer is located; is a vector of dimension, and the value of each element is randomly -1 or 1; When it means that the foraging situation of the th follower is not ideal and its position needs to be updated; When the follower can move to the best position.

[0031] The calculation formula for updating the position of the vigilant is: ; In the formula, is the current best position, is a random number subject to the standard normal distribution; is a random number, ; is the fitness value of the current sparrow individual, and are the current global best and worst fitness values respectively, is the minimum constant used to avoid the denominator being zero.

[0032] Exemplarily, in the model evaluation in step S4, the calculation formulas for the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) are as follows: ; ; ; In the formula: is the predicted value; is the true value; is the average value of the true values.

[0033] The specific performance indicators of the prediction model are compared and evaluated by these 3 indicators as shown in Table 1 below. The SSA-XGBoost model shows higher R² values and lower MAE and RMSE values on both the training set and the test set, proving its effectiveness in blast vibration velocity prediction.

[0034] Table 1 Performance indicators corresponding to different models

[0035] Among them, Figure 3 is the density map in the training set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 4 is the result map of the training set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 5 is the density map of the test set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 6 is the result map of the test set of the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization provided by the embodiments of the present invention; Figure 7 is the density map of the test set of the BP model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization of the present invention; Figure 8 is the density map of the test set of the BP-PSO model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization of the present invention; Figure 9 is the density map of the test set of the XGBoost model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization of the present invention; Figure 10 is the density result map of the training set of the BP model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization of the present invention; Figure 11 is the density result map of the training set of the BP-PSO model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization of the present invention; Figure 12 is the density result map of the training set of the XGBoost model for comparison with the tunnel blasting vibration velocity prediction method (SSA-XGBoost model) that integrates multi-factor intelligent optimization of the present invention; Exemplarily, in step S5, the safety threshold determines the safety threshold of the blasting vibration velocity by introducing geological feature variables and combining with the Sadovsky empirical formula. The safety threshold formula is as follows: ; In the formula, is the vibration velocity safety threshold, is the coefficient determined according to geological conditions and blasting experience, is the maximum charge per delay, is the distance from the blast center, is the weight coefficient of geological features, is the normalization function of the fracture density geological feature (such as fracture density, etc.); ; In the formula, is the normalization function based on rock hardness, is the normalization function based on fracture density. Either one can be selected for actual calculation.

[0036] Exemplarily, the blasting closed-loop optimization in step S5 is to predict the vibration velocity of the future construction section through the trained model to optimize the blasting parameters, and at the same time monitor and evaluate the blasting effect, and feedback the evaluation result to the design stage to provide a basis for the next section of blasting design, so as to realize the real-time optimization and improvement of tunnel blasting construction.

[0037] Table 2 Vibration velocity safety thresholds of some surrounding buildings

[0038] Table 3 K values of different lithologies in the blast area, values

[0039] Example 2, as Figure 13 shown, the tunnel blasting vibration velocity prediction system with multi-factor intelligent optimization provided by the embodiment of the present invention includes: Multi-source data acquisition module 1, which is used to collect blasting parameters (maximum single-section charge amount, distance from the blast center, drilling depth, and rock compressive strength) and historical vibration monitoring data; Data preprocessing module 2, which is used to normalize the data, construct a data set containing multi-factor features, and divide the training set, validation set, and test set; SSA-XGBoost model building module 3, which is used to initialize the XGBoost hyperparameters and iteratively optimize the XGBoost hyperparameters using the SSA algorithm; Model training and validation module 4, which is used to input the training set data to train the optimized SSA-XGBoost model, use the test set to verify the model performance and compare the prediction performance of traditional models (such as: traditional BP neural network, BP-PSO model, and unoptimized XGBoost), and evaluate the prediction error (MAE, RMSE, etc.) and the coefficient of determination (R 2 ); Engineering application module 5, which is used to dynamically adjust parameters such as the single-section charge amount in real time according to the prediction value, generate risk warnings in combination with safety thresholds (such as: improved Sadovskii formula limits), and realize the closed-loop optimization of blasting design.

[0040] As described above, it is only a relatively preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention shall be covered by the protection scope of the present invention.

Claims

1. A tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization, characterized in that: The method comprises the following steps: S1, collect blasting parameters and historical vibration monitoring data. Blasting parameters include maximum single-stage charge, blasting center distance, drilling depth and rock compressive strength; S2, normalize the data and construct a data set containing multi-factor features. The data set includes a training set, a validation set, and a test set. The ratio of the training set, validation set, and test set is 8:1:

1. S3, initialize XGBoost hyperparameters, use SSA algorithm to iteratively optimize XGBoost hyperparameters, and complete the SSA-XGBoost model construction; S4, input the training set data to train the optimized SSA-XGBoost model, use the test set to verify the model performance and compare the prediction performance of the traditional model; S5, dynamically adjusts the single-stage charge parameters in real time according to the predicted values, generates risk warnings in combination with safety thresholds, and realizes closed-loop optimization of blasting design.

2. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 1 is characterized in that: In step S1, blasting parameters and historical vibration monitoring data are collected, including: blasting vibration monitoring in multi-source data collection, using a blasting vibrometer to simultaneously measure three mutually perpendicular components of particle vibration, and vector synthesizing the vibration velocity components in the three directions to obtain the total vibration velocity , as the output variable of model training, the calculation formula is: ； In the formula, is the maximum radial component of the vibration velocity, is the maximum tangential component of the vibration velocity, is the maximum vertical component of vibration velocity.

3. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 1 is characterized in that: In step S3, XGBoost hyperparameters are initialized, including: The objective function of XGBoost consists of a loss function and a regularization term, and the formula is: ； In the formula, is the objective function of XGBoost, is the loss function of the model deviation, is the actual value, is the predicted value, is the total number of samples in the data set, For the The regularization term of the tree, is the total number of trees, is the basis function; Introducing elastic network regularization Elastic Net, combined with L1 and L2 regularization, we have: ； In the formula, To control the parameters of the tree structure, is the number of leaf nodes, is the weight of the first leaf node, is the weight of the second leaf node, is the set of scores of each tree leaf node, is the L1 norm, is the L2 norm; L1 regularization automatically screens the characteristics of blasting center distance and maximum charge in blasting parameters, and L2 regularization prevents the model from overfitting; ； In the formula, For the The objective function value of the step, For the front The sum of the predicted values ​​of the trees, is the predicted value of the regression tree.

4. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 3 is characterized in that: The SSA algorithm is used to iteratively optimize XGBoost hyperparameters, including: Define the parameters for XGBoost model optimization, including the number of iterations n_estimators, the maximum depth max_depth, and the learning rate learning_rate; Set the parameters of the sparrow search algorithm, including the number of sparrows, the maximum number of iterations, and define the fitness function; Simulate the foraging process and update the positions of discoverers, followers, and sentinels; The calculation formula for the discoverer's location update is: ； In the formula, For the A sparrow in the Location information in the dimension, is an exponential function, is the number of iterations; is a random value, ; is the maximum number of iterations, is the warning value, ; is a random number that follows a standard normal distribution, The value is 1 Dimensional matrix; is the safety threshold, ; when When , it means the finder is in a safe position to search; when When , it means that the discoverer has been discovered by the predator and needs to fly to another safe location to search.

5. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 4 is characterized in that: The follower position update is calculated as: ； In the formula, is an exponential function, is the global worst position, is the number of sparrows, The best position for the finder. for dimensional vector, each element has a random value of -1 or 1.

6. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 4 is characterized in that: The calculation formula for the sentinel position update is: ； In the formula, is the current best position, is a random number that follows a standard normal distribution; is a random number, ; is the fitness value of the current sparrow individual, and are the current global best and worst fitness values, respectively. is the minimum constant.

7. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 1 is characterized in that: In step S4, for the SSA-XGBoost model, the root mean square error RMSE, mean absolute error MAE, and determination coefficient R 2 Conduct an assessment.

8. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 1 is characterized in that: In step S5, the safety threshold is determined by introducing geological characteristic variables and combining the Sadovsky empirical formula to determine the safety threshold of the blasting vibration velocity. The safety threshold formula is as follows: ； In the formula, is the vibration speed safety threshold, is a coefficient determined based on geological conditions and blasting experience. is the maximum dosage of a single stage, is the explosion center distance, is the weight coefficient of geological characteristics, is the normalized function of the geological characteristics of fracture density; ； In the formula, is a normalized function based on rock hardness, is a normalized function based on crack density.

9. The tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization according to claim 1 is characterized in that: In step S5, closed-loop optimization of blasting design is achieved, including: predicting the vibration velocity of the future construction section through the trained SSA-XGBoost model to optimize the blasting parameters, and monitoring and evaluating the blasting effect at the same time.

10. A tunnel blasting vibration velocity prediction system integrating multi-factor intelligent optimization, characterized in that: The system implements the tunnel blasting vibration velocity prediction method integrating multi-factor intelligent optimization as described in any one of claims 1 to 9, and the system comprises: A multi-source data acquisition module (1) is used to collect blasting parameters and historical vibration monitoring data, wherein the blasting parameters include the maximum single-stage charge, blasting center distance, drilling depth and rock compressive strength; A data preprocessing module (2) is used to normalize the data and construct a data set containing multi-factor features, wherein the data set includes a training set, a validation set, and a test set, wherein the ratio of the training set, the validation set, and the test set is 8:1:1; SSA-XGBoost model building module (3), used to iteratively optimize XGBoost hyperparameters using the SSA algorithm to complete the SSA-XGBoost model building; The model training and verification module (4) is used to input the training set data to train the optimized SSA-XGBoost model, and use the test set to verify the model performance and compare the prediction performance of the traditional model; The engineering application module (5) is used to dynamically adjust the single-stage charge parameters in real time according to the predicted values, generate risk warnings in combination with safety thresholds, and realize closed-loop optimization of blasting design.

Images