A method for predicting blasting vibration peak value based on improved SVM algorithm

Abstract

The invention discloses a blasting vibration peak prediction method based on an improved SVM algorithm, relates to the field of machine learning and engineering blasting vibration safety technology, and specifically aims to solve the problem that the existing prediction accuracy and reliability in the background technology still need to be improved. A blasting vibration peak prediction method based on an improved SVM algorithm is set up, including: S1. constructing an original data set; S2. preprocessing the original data set; S3. cleaning and feature extraction of the processed data set; S4. obtaining a final improved SVM prediction model; S5. measuring the performance of the final improved SVM prediction model; the method improves the prediction ability of the SVM network so that it can accurately predict the vibration intensity, thereby providing a reference for engineering decision-making.

Description

Technical Field

[0001] The invention relates to the field of machine learning and engineering blasting vibration safety technology, and in particular to a blasting vibration peak prediction method based on an improved SVM algorithm. Background Art

[0002] Blasting vibration velocity prediction is based on the need for vibration control and safety assessment in blasting engineering. During blasting operations, the release of explosive energy will generate strong vibration waves, which can propagate into the surrounding medium and affect surrounding buildings, structures and the natural environment. In order to ensure the safety of blasting operations and reduce the impact on the surrounding environment, it is necessary to monitor and predict the vibration velocity generated by blasting.

[0003] The prediction of blasting vibration velocity has the following important significance for the safe construction of the project:

[0004] Ensure safety: During blasting operations, accurate prediction of vibration velocity is crucial to ensure the safety of the surrounding environment. Through prediction, the potential impact of seismic waves generated by blasting on surrounding buildings, structures and personnel can be evaluated, so that corresponding safety protection measures can be taken to avoid or reduce damage caused by blasting vibration.

[0005] Improve blasting efficiency: Accurately predicting blasting vibration velocity helps improve the efficiency of blasting operations, reduce additional safety precautions caused by inaccurate predictions, reduce operating costs, and improve the reliability and economy of blasting operations.

[0006] Blasting vibration control: Predicting the blasting vibration speed can help the construction party formulate a reasonable blasting plan and control the vibration intensity within a safe range by monitoring and adjusting the blasting parameters.

[0007] Environmental protection: The prediction of blasting vibration velocity helps to evaluate the potential impact of blasting on the environment, so as to take environmental protection measures, such as reducing the blasting vibration intensity, optimizing the blasting time, etc., to reduce interference and damage to the environment.

[0008] Data support and decision-making: Through the prediction of blasting vibration velocity, a large amount of field data can be accumulated. These data are of great value for the improvement of blasting technology, the research and development of new technologies, and the formulation of relevant specifications. At the same time, they provide scientific basis and decision-making support for blasting management.

[0009] The prediction of blasting vibration velocity is not only a technical issue, but also an important link related to the safety, efficiency, cost and environment of blasting operations. Therefore, the prediction of blasting vibration velocity is an indispensable part of blasting engineering, which is of great significance for ensuring the safety of blasting operations and reducing the impact on the surrounding environment.

[0010] The application of intelligent algorithms in predicting blasting vibration velocity has made some progress, and with the continuous development of machine learning and data science technology, the application of these algorithms is gradually increasing; machine learning models, such as random forest (RF), gradient boosting machine (GBM) and artificial neural network (ANN), have been used to predict blasting vibration velocity. These models are able to process a large amount of input data and learn the laws of predicting vibration velocity.

[0011] However, using the above model for prediction will overfit the training data, resulting in a decrease in generalization ability, ignoring the correlation between features, and thus missing the opportunity to improve model performance by combining features, ultimately making the prediction results inaccurate and the prediction effect poor, and further research and technical development are needed; therefore, the present invention proposes a blasting vibration peak prediction method based on an improved SVM algorithm, aiming to improve the algorithm accuracy, capture high-level features in the data, and further improve the accuracy and reliability of the prediction. Summary of the invention

[0012] In view of the defects of the above-mentioned prior art, the present invention aims to provide a method for predicting blasting vibration peak value based on an improved SVM algorithm. The method improves the prediction ability of the SVM network so that it can accurately predict the vibration intensity and provide a reference for engineering decision-making. For example, blasting parameters can be adjusted to prevent blasting vibration from affecting surface buildings.

[0013] The present invention aims to improve the limitations of neural networks in predicting blasting vibration velocity, and aims to improve the accuracy of the prediction system while ensuring the prediction ability of the system in practical applications.

[0014] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0015] A method for predicting blasting vibration peak value based on improved SVM algorithm, comprising:

[0016] S1. Obtain data uploaded by the data acquisition and transmission system and construct the original data set;

[0017] S2. Preprocessing the original data set in step S1 to obtain a processed data set;

[0018] S3. Cleaning and feature extraction of the data set processed in step S2 to obtain a feature data set;

[0019] S4. Based on the feature data set of step S3 and combined with the traditional SVM model, a final improved SVM prediction model is obtained;

[0020] S5. Use the final improved SVM prediction model in step S4 to make predictions and measure the performance of the final improved SVM prediction model.

[0021] Preferably, the process of obtaining the final improved SVM prediction model includes:

[0022] S401. Based on the traditional SVM model, a preliminary improved SVM model is constructed in combination with the feature data set in step S3;

[0023] S402. Enable the preliminary improved SVM model to learn.

[0024] Preferably, in step S401, the process of constructing a preliminary improved SVM model includes:

[0025] S401-1. Select one feature data from the feature data set in step S3 to construct a new optimization problem, and replace the basic optimization problem in the traditional SVM model with the new optimization problem;

[0026] S401-2. Select the sigmoid kernel function as the kernel function of the traditional SVM model to obtain a preliminary improved SVM model.

[0027] Preferably, the new optimization problem expression in step S401-1 is:

[0028]

[0029] sty i (ω·x i +b)≥1-ξ i ,ξ i ≥0

[0030] Where ω is the weight vector, b is the bias term, ||ω|| 2 is the square of the norm of the weight vector, used for regularization, C is the regularization parameter, ξ i is the slack variable, y i is the label of the ith data point, x i is the corresponding eigenvector, ω·x i Represents the dot product between the weight vector and the feature vector, n is the total number, and st is subject to.

[0031] Preferably, in step S402, the learning process of preliminarily improving the SVM model includes:

[0032] S402-1. Optimize the hyperparameters of the preliminary improved SVM model, update its optimal weight vector ω and bias term b, and obtain the trained improved SVM model;

[0033] S402-2. Based on the mean absolute error and the root mean square error, the values ​​of the correlation coefficient R and the relative error σ are obtained, and R and σ are used to evaluate the performance of the improved SVM model after training in step S402-1. At the same time, the complexity of the improved SVM model after training is controlled by adjusting the regularization parameter C;

[0034] S402-3. Apply the improved SVM model evaluated in step S402-2 to the new data in the feature data set in step 3, and calculate ω·x in the improved SVM model after evaluation i +b value, the data is assigned to the category with a calculated value greater than 0.05, and the improved SVM model after classification is obtained;

[0035] S402-4. Repeat step S401 and step 402-1 to step S402-3 for other data in the feature data set of step 3, output and update the weights and biases of the preliminary improved SVM model, complete the preliminary improved SVM model learning, and obtain the final improved SVM prediction model. Preferably, in step S402-1, the process of performing hyperparameter optimization on the preliminary improved SVM model includes:

[0036] S402-1A. Randomly select multiple hyperparameters in the preliminary improved SVM model to form a first generation data set;

[0037] S402-1B. The fitness function is used to calculate the fitness of each individual in the first generation data set, and it is determined whether the individual satisfies the termination condition at this time. If so, it goes to step S402-1G;

[0038] S402-1C. Calculate the probability of each individual being selected in the next selection according to the fitness degree obtained in step S402-1B and the probability formula of individual i being selected;

[0039] S402-1D. According to the probability calculated in step S402-1C, select M individuals with high probability and form the second generation data set; wherein, a certain data can be repeatedly selected, and M <N;

[0040] S402-1E. The data in the second-generation data set is based on the mating probability pc, and a "crossover" operation is performed to generate NM data, and then the generated NM data is subjected to a "mutation" operation to obtain a mutated NM data set; the data in the second-generation data set is then "copied" to generate L data, and the L data is merged with the mutated NM data set to generate a third-generation data set;

[0041] S402-1F: Return to step S402-1B based on the third-generation data set. Step S402-1B determines whether the individual satisfies the termination condition at this time. If so, go to S402-1G;

[0042] S402-1G: Output and update the optimal weight vector ω and bias term b in the preliminary improved SVM model, realize the hyperparameter optimization of the preliminary improved SVM model, and obtain the trained improved SVM model.

[0043] Preferably, in step S402-1B, the fitness function expression is:

[0044]

[0045] Among them, f(i) is the fitness of the i-th individual, F(xi) is the objective function value of the i-th individual, and N is the population size;

[0046] In step S402-1C, the probability formula for individual i being selected is:

[0047] The probability of individual i being selected = the fitness of individual i ÷ the sum of the fitness of all individuals.

[0048] Preferably, the performance method of the final improved SVM prediction model includes: calculating the correlation coefficient R and the relative error σ based on the mean absolute error and the root mean square error, and measuring the performance of the final improved SVM prediction model by the correlation coefficient R and the relative error σ.

[0049] Preferably, the expression of mean absolute error is:

[0050]

[0051] The expression of the root mean square error is:

[0052]

[0053] The calculation formula of the correlation coefficient R is:

[0054]

[0055] The calculation formula of relative error σ is:

[0056]

[0057] In the formula, y i and y0 represent the actual value and predicted value respectively. and Represent the mean of the actual value and the predicted value, N represents the number of samples, i represents the number of data; R reflects the correlation coefficient. A blasting vibration peak prediction system based on an improved SVM algorithm includes:

[0058] The data acquisition module is used to acquire the data uploaded by the data acquisition and transmission system and construct the original data set;

[0059] The data preprocessing module is used to preprocess the original data set of the data acquisition module to obtain a processed data set; the original data set is preprocessed to obtain a processed data set;

[0060] A feature extraction module is used to clean and extract features from the data set processed by the data preprocessing module to obtain a feature data set;

[0061] The model acquisition module, based on the feature data set of the feature extraction module and combined with the traditional SVM model, obtains the final improved SVM prediction model;

[0062] The testing module uses the final improved SVM prediction model of the model acquisition module to make predictions, and then measures the performance of the final improved SVM prediction model.

[0063] Wherein, the prediction system is implemented based on the method described in any one of claims 1-9.

[0064] The beneficial effects of the present invention are as follows: the present invention discloses a method for predicting blasting vibration peak value based on an improved SVM algorithm. Compared with the prior art, the improvements of the present invention are as follows:

[0065] (1) In order to improve the accuracy and efficiency of the prediction model, the present invention improves the hyperparameter optimization part of the SVM algorithm, so that the model can more accurately predict future values, while making full use of resources and improving the accuracy and effectiveness of the prediction through optimization.

[0066] (2) The present invention optimizes the parameter selection process of the SVM model when processing numerical data, thereby optimizing the SVM model.

[0067] (3) The present invention applies the improved SVM model to new data sets and evaluates the performance of the model by comparing the differences between the predicted results and the actual observed values. The model can also be used to predict blasting vibrations to provide a reference for engineering decision-making. For example, blasting parameters can be adjusted to prevent surface settlement caused by blasting vibrations. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 is a flow chart of the method of the present invention; DETAILED DESCRIPTION

[0069] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0070] Embodiment 1:

[0071] See attached Figure 1 A method for predicting blasting vibration peak value based on an improved SVM algorithm is shown, comprising the steps of:

[0072] S1. Obtain the data uploaded by the data acquisition and transmission system and construct the original data set;

[0073] Specifically, vibration sensors are installed at key parts of the tunnel (such as the arch waist and the arch of the tunnel), and the vibration sensors are connected to the data acquisition and transmission system. The data acquisition and transmission system monitors the relevant data of the detonator section and the surrounding environment conditions in real time during each blasting operation; the relevant data of the detonator section in each blasting operation includes the data uploaded by the vibration sensor;

[0074] The data monitored by the vibration sensor mainly involves the blasting vibration speed and frequency of the x, y and z axes. Through these data, the synthetic speed, that is, the synthetic effect of the vibration speed, can be calculated; the synthetic speed is an important parameter for evaluating the impact of blasting vibration on the surrounding environment and structure. It helps to determine whether the blasting vibration is within the safe allowable range, thereby ensuring the stability of the tunnel structure and the safety of the surrounding environment.

[0075] S2. Preprocessing the original data set in step S1 to obtain a processed data set;

[0076] The original data set containing noise, missing values ​​and outliers in step S1 is preprocessed by normalization, standardization and normalization to improve the quality of the original data set and pave the way for subsequent data analysis and model building;

[0077] Specifically, check and remove invalid or erroneous records in the original data set, unify the data format into a standard format, standardize the data scale, etc., to ensure the quality of the processed data, ensure its completeness, accuracy and consistency, obtain the processed data set, and store the processed data set in preparation for subsequent analysis.

[0078] S3. Cleaning and feature extraction of the data set processed in step S2 to obtain a feature data set;

[0079] The data set in step S2 is processed by methods such as data cleaning and feature extraction. When performing data set cleaning and feature extraction, it is necessary to comprehensively consider the characteristics of the data and the analysis objectives, and flexibly select and adjust the threshold to achieve the best data set preprocessing effect;

[0080] Specifically, duplicate data or erroneous data that may be caused by human errors or errors in the automation process in the data set processed in step S2 are deleted to obtain a cleaned data set; the cleaned data set is then subjected to processes such as selecting important features, constructing new features, and scaling features, and features that are beneficial to model building are extracted, and the extracted features are then used to form a feature data set;

[0081] S4. Based on the result of step S3 and combined with the traditional SVM model, a final improved SVM prediction model is obtained;

[0082] S401. Based on the traditional SVM model and combined with the feature data set in step 3, a preliminary improved SVM model is constructed;

[0083] The specific process of constructing the preliminary improved SVM model is:

[0084] 401-1. Modify the basic optimization problem of the traditional SVM; select a feature data in the feature data set of step 3 to construct a new optimization problem, and replace the basic optimization problem in the traditional SVM model with the new optimization problem;

[0085] The new optimization problem is expressed as:

[0086]

[0087] sty i (ω·x i +·b)≥1-ξ i ,ξ i ≥0

[0088] Where ω is the weight vector, b is the bias term, ||ω|| 2 is the square of the norm of the weight vector, used for regularization, C is the regularization parameter, ξ i is the slack variable, y i is the label of the ith data point, x i is the corresponding eigenvector, ω·x i represents the dot product between the weight vector and the feature vector; n is the total number; st refers to subject to, which is restricted to...;

[0089] S401-2. Select the kernel function of the traditional SVM; specifically, based on the preliminary SVM model, select the sigmoid kernel function as its kernel function to obtain a preliminary improved SVM model; wherein, the sigmoid kernel function can map the feature data in step 3 to a high-dimensional space, so that the feature data is linearly separable in this space.

[0090] S402. Learning the preliminary improved SVM model to obtain an improved SVM model;

[0091] The learning process of the preliminary improved SVM model includes:

[0092] S402-1. Training the preliminary improved SVM model; specifically, performing hyperparameter optimization on the preliminary improved SVM model, thereby minimizing the preliminary improved SVM model, updating its optimal weight vector ω and bias term b, and obtaining the trained improved SVM model;

[0093] The process of initial improvement of SVM model hyperparameter optimization is:

[0094] S402-1A. Randomly select multiple hyperparameters in the preliminary improved SVM model to form a first generation data set;

[0095] S402-1B. The fitness function is used to calculate the fitness of each individual in the first generation data set, and it is determined whether the individual satisfies the termination condition at this time. If so, it goes to step S402-1G;

[0096] The fitness function expression is:

[0097]

[0098] Among them, f(i) is the fitness of the i-th individual, F(xi) is the objective function value of the i-th individual, and N is the population size;

[0099] S402-1C. Calculate the probability of each individual being selected in the next selection according to the fitness degree obtained in step S402-1B and the probability formula of individual i being selected;

[0100] The probability formula for individual i being selected is:

[0101] The probability of individual i being selected = the fitness of individual i ÷ the sum of the fitness of all individuals;

[0102] S402-1D. According to the probability calculated in step S402-1C, select M individuals with high probability and form the second generation data set; wherein, a certain data can be repeatedly selected, and M <N;

[0103] S402-1E. The data in the second-generation data set is based on the mating probability pc, and a "crossover" operation is performed to generate NM data, and then the generated NM data is subjected to a "mutation" operation to obtain a mutated NM data set; the data in the second-generation data set is "copied" to generate L data, and the L data is merged with the mutated NM data set to generate a third-generation data set;

[0104] The crossover operation can ensure that each evolution retains excellent features. After a piece of data is generated through crossover, it is necessary to randomly select several data on the new data set and then randomly modify the data value, thereby introducing new data into the existing data set, breaking through the current search limitations and making it easier for the algorithm to find the global optimal solution.

[0105] S402-1F: Return to step S402-1B based on the third-generation data set. Step S402-1B determines whether the individual satisfies the termination condition at this time. If so, go to S402-1G;

[0106] S402-1G: terminate the calculation and output the optimal result, that is, output and update the optimal weight vector ω and bias term b in the preliminary improved SVM model, realize the hyperparameter optimization of the preliminary improved SVM model, and obtain the improved SVM model after training;

[0107] S402-2. Evaluate the improved SVM model after training in step S402-1; specifically, based on the mean absolute error (MAE) and the root mean square error (MSE), obtain the values ​​of the correlation coefficient R and the relative error σ, and use the values ​​to evaluate the performance of the improved SVM model after training. At the same time, adjust the regularization parameter C to control the complexity of the improved SVM model after training to avoid overfitting or underfitting, and obtain the improved SVM model after evaluation;

[0108] S402-3. Application of the improved SVM model after evaluation; the improved SVM model evaluated in step S402-2 is used for new data in the feature data set in step 3, and the new data is classified, specifically, by calculating the ω·x in the improved SVM model after evaluation i +b value, assign data to calculate ω·x i + The categories with b value greater than 0.05, the improved SVM model after classification was obtained;

[0109] S402-4. Repeat steps 401 and 402-1 to 402-3 for the other data in the feature data set of step 3, output and update the weights and biases of the preliminary improved SVM model, complete the preliminary improved SVM model learning, and obtain the final improved SVM prediction model;

[0110] Specifically, the feature data corresponding to each time step t in the feature data set of step 3 repeats step 401 and step 402-1 to step 402-3, outputs and updates the weights and biases of the preliminary improved SVM model, completes the preliminary improved SVM model learning, and obtains the final improved SVM prediction model;

[0111] S5. Use the final improved SVM prediction model of step S4 to make predictions, and measure the performance of the final improved SVM prediction model;

[0112] Model evaluation: Use the final improved SVM prediction model of step S4 for prediction, and use the mean absolute error (MAE), root mean square error (MSE), correlation coefficient R and relative error σ to evaluate the prediction results. R reflects the correlation coefficient between the predicted value and the true value. The closer the value is to 1, the better the fit of the final improved SVM prediction model and the better the model prediction performance. σ reflects the relative error between the predicted value and the true value, indicating the degree to which the predicted value deviates from the true value. The closer the value is to 0, the smaller the error is, indicating that the prediction effect of the final improved SVM prediction model is better.

[0113] Specifically, based on the mean absolute error (MAE) and the root mean square error (MSE), the values ​​of the correlation coefficient R and the relative error σ are calculated. The performance of the final improved SVM prediction model is measured by the values ​​of the correlation coefficient R and the relative error σ. The specific expression is as follows:

[0114]

[0115]

[0116]

[0117]

[0118] In the formula, y i and y0 represent the actual value and predicted value respectively. and Represent the means of the actual value and the predicted value respectively, N represents the number of samples, i represents the number of data, and R reflects the correlation coefficient. The closer the value is to 1, the better the model fit is and the better the model prediction performance is.

[0119] The above shows and describes the basic principles, main features and advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for predicting blasting vibration peak value based on improved SVM algorithm, characterized by: include: S1. Obtain data uploaded by the data acquisition and transmission system and construct the original data set; S2. Preprocessing the original data set in step S1 to obtain a processed data set; S3. Cleaning and feature extraction of the data set processed in step S2 to obtain a feature data set; S4. Based on the feature data set of step S3 and combined with the traditional SVM model, a final improved SVM prediction model is obtained; S5. Use the final improved SVM prediction model in step S4 to make predictions and measure the performance of the final improved SVM prediction model; The process of obtaining the final improved SVM prediction model includes: S401. Based on the traditional SVM model, a preliminary improved SVM model is constructed in combination with the feature data set in step S3; S402. Make the preliminary improved SVM model learn; In step S401, the process of constructing a preliminary improved SVM model includes: S401-1. Select one feature data from the feature data set in step S3 to construct a new optimization problem, and replace the basic optimization problem in the traditional SVM model with the new optimization problem; S401-2. Select the sigmoid kernel function as the kernel function of the traditional SVM model to obtain a preliminary improved SVM model; In step S402, the learning process of the preliminary improved SVM model includes: S402-1. Optimize the hyperparameters of the preliminary improved SVM model, update its optimal weight vector ω and bias term b, and obtain the trained improved SVM model; S402-2. Based on the mean absolute error and the root mean square error, the values ​​of the correlation coefficient R and the relative error σ are obtained, and R and σ are used to evaluate the performance of the improved SVM model after training in step S402-1. At the same time, the complexity of the improved SVM model after training is controlled by adjusting the regularization parameter C; S402-3. Apply the improved SVM model evaluated in step S402-2 to the new data in the feature data set in step 3, and calculate ω·x in the improved SVM model after evaluation i +b value, the data is assigned to the category with a calculated value greater than 0.05, and the improved SVM model after classification is obtained; S402-4. Repeat steps S401 and 402-1-3 for other data in the feature data set of step 3, output and update the weights and biases of the preliminary improved SVM model, complete the preliminary improved SVM model learning, and obtain the final improved SVM prediction model; In step S402-1, the process of performing hyperparameter optimization on the preliminary improved SVM model includes: S402-1A. Randomly select multiple hyperparameters in the preliminary improved SVM model to form a first generation data set; S402-1B. The fitness function is used to calculate the fitness of each individual in the first generation data set, and it is determined whether the individual satisfies the termination condition at this time. If so, it goes to step S402-1G; S402-1C. Calculate the probability of each individual being selected in the next selection according to the fitness degree obtained in step S402-1B and the probability formula of individual i being selected; S402-1D. According to the probability calculated in step S402-1C, select M individuals with high probability and form the second generation data set; wherein, a certain data can be repeatedly selected, and M <N; S402-1E. The data in the second-generation data set is based on the mating probability pc, and a "crossover" operation is performed to generate NM data, and then the generated NM data is subjected to a "mutation" operation to obtain a mutated NM data set; the data in the second-generation data set is then "copied" to generate L data, and the L data is merged with the mutated NM data set to generate a third-generation data set; S402-1F: Return to step S402-1B based on the third-generation data set. Step S402-1B determines whether the individual satisfies the termination condition at this time. If so, go to S402-1G; S402-1G: Output and update the optimal weight vector ω and bias term b in the preliminary improved SVM model, realize the hyperparameter optimization of the preliminary improved SVM model, and obtain the trained improved SVM model; The method for measuring the performance of the final improved SVM prediction model includes: calculating the value of the correlation coefficient R and the relative error σ based on the mean absolute error and the root mean square error, and measuring the performance of the final improved SVM prediction model through the value of the correlation coefficient R and the relative error σ; The expression of mean absolute error is: The expression of the root mean square error is: The calculation formula of the correlation coefficient R is: The calculation formula of relative error σ is: In the formula, y i and y0 represent the actual value and predicted value respectively. and They represent the means of the actual value and the predicted value respectively, N represents the number of samples, i represents the number of data; R reflects the correlation coefficient.

2. The method for predicting blasting vibration peak value based on improved SVM algorithm according to claim 1 is characterized in that: The new optimization problem expression in step S401-1 is: style i (ω·x i +b)≥1-ξ i ,x i ≥0 Where ω is the weight vector, b is the bias term, ||ω|| 2 is the square of the norm of the weight vector, used for regularization, C is the regularization parameter, ξ i is the slack variable, y i is the label of the ith data point, x i is the corresponding eigenvector, ω·x i Represents the dot product between the weight vector and the feature vector, n is the total number, and st is subject to.

3. The method for predicting blasting vibration peak value based on improved SVM algorithm according to claim 1, characterized in that: In step S402-1B, the fitness function expression is: Among them, f (i) is the fitness of the ith individual, F(xi) is the objective function value of the ith individual, and N is the population size; In step S402-1C, the probability formula for individual i being selected is: The probability of individual i being selected = the fitness of individual i ÷ the sum of the fitness of all individuals.

4. A blasting vibration peak prediction system based on an improved SVM algorithm, characterized in that: include: The data acquisition module is used to acquire the data uploaded by the data acquisition and transmission system and construct the original data set; A data preprocessing module is used to preprocess the original data set of the data acquisition module to obtain a processed data set; The original data set is preprocessed to obtain a processed data set; A feature extraction module is used to clean and extract features from the data set processed by the data preprocessing module to obtain a feature data set; The model acquisition module, based on the feature data set of the feature extraction module and combined with the traditional SVM model, obtains the final improved SVM prediction model; The testing module uses the final improved SVM prediction model of the model acquisition module to make predictions, thereby measuring the performance of the final improved SVM prediction model; Wherein, the prediction system is implemented based on the method described in any one of claims 1-3.

Images