Underground anchor rod anchoring strength prediction method and system based on support vector machine method

Abstract

The invention belongs to the technical field of anchoring strength detection, and discloses an underground anchor rod anchoring strength prediction method and system based on a support vector machine method. The method comprises the steps of cleaning data, processing missing values and standardizing features; using correlation analysis or other feature selection methods to select features with high correlation with anchoring strength; performing model training by using a support vector machine method SVM, and optimizing hyper-parameters by using grid search; completing the construction of a support vector machine model; and carrying out model evaluation and result visualization. Key parameters in the mixing process are monitored in real time through the sensor network, the system can obtain anchoring strength data in real time, the detection period is greatly shortened, and the construction efficiency is improved. The support vector machine algorithm is adopted to analyze and predict the collected data, personal errors can be eliminated to a greater extent, and the accuracy and reliability of anchoring strength detection are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of anchoring strength detection, and in particular relates to a method and system for predicting the anchoring strength of underground anchor rods based on a support vector machine method. Background Art

[0002] With the rapid development of modern construction and mining engineering, the safety and efficiency of underground operations have received increasing attention. Among them, anchoring technology, as an important means of supporting and reinforcing underground structures, plays a key role in preventing ground settlement, landslides and rock fractures.

[0003] Anchoring agent is an important material used to enhance the anchoring effect. Its mixing ratio, construction process and environmental conditions will directly affect the anchoring strength. However, the traditional anchoring agent mixing and strength testing methods have some shortcomings, such as long testing cycle, inaccurate results, and difficulty in reflecting the on-site construction situation in real time. This not only increases the risk of the project, but also leads to a waste of resources.

[0004] Furthermore, the necessity of underground anchoring is an important technology to ensure the stability of underground projects (such as mines, tunnels, and underground storage facilities). Anchoring agents enhance the bearing capacity and stability of structures by combining with surrounding rocks or soil. The curing strength of anchoring agents directly affects the safety of the project, so its strength needs to be monitored in real time.

[0005] The curing process of the anchor usually involves the reaction of cement, polymer and other chemical additives. The surrounding rock properties, anchor length, burial depth and other factors will affect the anchor strength. Therefore, accurate monitoring of these factors is the key to achieve intelligent management.

[0006] Support vector machine (SVM) technology, support vector machine is a supervised learning method, widely used in classification and regression problems. Its basic principle is to separate data points of different categories by constructing a hyperplane in a high-dimensional space. The intelligent monitoring system for anchor strength after mixing of underground anchoring agents based on the support vector machine method combines modern sensor technology with machine learning methods, can monitor and predict anchor strength in real time, and provides strong support for the safety management of underground projects. This technology has broad application prospects and will promote the intelligent development of underground projects.

[0007] In underground engineering, the performance of anchoring agents is directly related to the safety and stability of the project. Traditional methods of mixing and testing anchoring agents face the following major technical problems: Low detection efficiency: Existing methods usually rely on manual sampling and laboratory testing, which leads to long detection cycles and inability to obtain anchor strength data in real time. This delay may lead to safety hazards during construction.

[0008] Insufficient data accuracy: Traditional detection methods are greatly affected by human operation and environmental factors, and often fail to reflect the actual mixing effect, which may lead to misjudgment and unnecessary waste of resources.

[0009] Poor adaptability: Different underground conditions (such as soil type, humidity, temperature, etc.) have a significant impact on the performance of the anchoring agent. Traditional methods are difficult to effectively adjust the mixing ratio and construction process in complex underground environments.

[0010] Lack of intelligent management: Existing systems are mostly manually monitored and controlled, lacking real-time data analysis and intelligent decision-making capabilities, and are unable to meet the needs of modern engineering for efficiency, safety, and intelligence. Summary of the invention

[0011] In order to overcome the problems existing in the related art, the disclosed embodiment of the present invention provides a method and system for predicting the anchor strength of underground anchor rods based on support vector machine method. The present invention particularly relates to the field of underground engineering, specifically underground anchoring technology and intelligent monitoring system.

[0012] The technical solution is as follows: A method for predicting the anchor strength of underground anchor bolts based on a support vector machine method, comprising: S1, data preprocessing: cleaning data, handling missing values, and standardizing features; S2, feature selection: feature selection methods are used to select features with high correlation with anchor strength; S3, model training: Use support vector machine method SVM for model training, use grid search to optimize hyperparameters, and complete the construction of support vector machine model; S4, model evaluation: evaluate the performance of the support vector machine model on the training set and test set by calculating the mean absolute error MAE and mean square error MSE indicators; S5, Result Visualization: Use visualization tools to show the comparison between model prediction results and actual values.

[0013] In step S1, the normalized features are normalized using Z-score.

[0014] In step S2, feature selection includes: (2.1) Data reading: Use Pandas’ read_csv function to read the dataset from the CSV file; (2.2) Separation of features and target variables: The data set contains features and target variables. The target variable is separated from the feature matrix through the drop function to form the feature matrix X and the target variable vector y_actual, and the data set is randomized. (2.3) Re-obtain the randomized features and target variables, and re-obtain the feature matrix X_shuffled and the target variable vector y_actual_shuffled from the randomized dataset.

[0015] In step (2.1), the file is read using GBK encoding; In step (2.2), the data set is randomized, including: using the sample method to randomize the data set, using frac=1 to retain all data, and random_state=50 to set the random seed.

[0016] In step S3, the support vector machine method SVM is used for model training, and the hyperparameters are optimized using grid search, including: (1) Define the parameter grid. When building a machine learning model, systematically traverse all parameter combinations to find the best parameters. Start the grid search process by calling the fit method. After the grid search is completed, check the best parameter combination and retrain the model using the best parameter combination: (2) Training set and test set division: The processed data set is randomly divided into 80% training set and 20% test set.

[0017] In step (1), a grid search is performed using RBF radial basis function or POIY polynomial kernel, and POIY polynomial kernel is used for polynomial feature mapping; Use RBF radial basis function technology to solve the interpolation problem, select Basis function training data, the form of each basis function is: ; In the formula, is a nonlinear function, and the training data points yes The center of the circle, the basis function is the data in the input space With Center The distance is used as the independent variable of the function. Since the distance is radially homogeneous, the function is called a radial basis function. The difference function based on the radial basis function technology is defined as a linear combination of basis functions, and the expression is: ; In the formula, In a two-dimensional plane, is the center of the circle, The distance between the data and the center of the circle is independent of the position and size of the data, and the distance between the points on the circle and the center of the circle is equal.

[0018] Further, the hyperparameters include: The C parameter is a regularization parameter in SVM, which is used to balance the trade-off between margin and misclassified samples; Gamma parameter, used to define the influence range of a single training sample on the model; The Epsilon parameter is used to control the error tolerance of the support vector machine model, the allowable error range, and define the support vector machine model's tolerance for noise in the training data.

[0019] In step S4, the calculation formulas for the mean absolute error MAE and mean square error MSE indicators are: ; ; In the formula, is the mean absolute error, is the root mean square error; is the predicted value of corrosion rate, in mm / a; It is the actual value of corrosion rate, the unit is mm / a.

[0020] Another object of the present invention is to provide a downhole anchor bolt anchoring strength prediction system based on support vector machine method, the system implements the downhole anchor bolt anchoring strength prediction method based on support vector machine method, and the hardware part of the system includes: The data acquisition system is provided with a data acquisition card for converting analog signals from sensors into digital signals for real-time monitoring and recording; Embedded processing unit, responsible for data preprocessing, feature extraction and real-time calculation of support vector machine (SVM) model; Communications systems to enable remote transmission of real-time data for easy monitoring and analysis; The battery system includes a battery pack for providing power to the underground environment; a battery management module for providing voltage stabilization and overload protection; Display and control interface, including a display screen for real-time display of monitoring data, system status and alarm information; Control panel for setting monitoring parameters and system calibration.

[0021] The software part of the system includes: Data acquisition module, used to collect the anchor strength data after the anchoring agent is mixed in real time; A data processing module, using a support vector machine (SVM) method to analyze the anchor strength data; The monitoring interface uses a graphical interface to display the anchor strength monitoring results and their changing trends; The data collected by the data collection module includes sensor data, where sensors are deployed to collect physical and chemical property data of the anchoring agent; Ultrasonic data to measure mixing uniformity; Environmental data, including temperature, humidity, and pH; Strength test data, anchor strength data obtained through laboratory tests; The data processing module also includes: a data preprocessing function submodule, which is used to remove noise and abnormal values.

[0022] Combined with all the above technical solutions, the beneficial effects of the present invention are as follows: the present invention collects data on the mean anchoring force under different surrounding rocks and anchoring lengths, and uses the support vector machine algorithm to model and predict the mean anchoring force. First, the borehole depth is measured using an infrared rangefinder to obtain relevant data, and the data is transmitted to an embedded processing unit. Then, the raw data is converted into a format suitable for SVM model training through data preprocessing and feature extraction. Next, the data is trained using the SVM model, and the model performance is evaluated using a training set and a test set. The prediction results use the mean anchoring force to show the anchoring strength, and the difference between the predicted mean downhole anchoring force and the actual mean anchoring force of the established SVM model is within 0.1KN. This shows that the model has good performance. The test after the hardware is complete shows that the system can effectively improve the accuracy, real-time and possibility of anchoring strength monitoring, and provide a strong guarantee for the safety of underground work. In the future, the system will further integrate more sensors and data analysis algorithms to enhance the capabilities and scope of application of intelligent monitoring.

[0023] Compared with the prior art, the advantages of the present invention further include: Improve monitoring efficiency: By real-time monitoring of key parameters in the mixing process through a sensor network, the system can instantly obtain anchor strength data, significantly shorten the detection cycle, and improve construction efficiency.

[0024] Enhanced data accuracy: Using the support vector machine (SVM) algorithm to analyze and predict the collected data can eliminate human errors to a greater extent and improve the accuracy and reliability of anchor strength detection.

[0025] Adapt to complex environments: The system can automatically adjust parameters and optimize mixing solutions according to different construction environments and material characteristics, thereby adapting to changing underground engineering conditions.

[0026] Realize intelligent decision-making: Through real-time data analysis, the system can provide intelligent decision-making support for construction and timely warn of potential risks, thereby effectively improving project safety.

[0027] Promote industry development: The successful implementation of this invention can not only improve the existing anchoring technology, but also provide new ideas for the application of intelligent monitoring and control technology in underground engineering, and promote technological progress and innovation in related fields.

[0028] The present invention will bring significant technological progress and economic benefits to the field of anchor strength detection. The method ensures high data quality by cleaning data, processing missing values ​​and standardizing features, thereby improving the effectiveness of model training. In the feature selection stage, correlation analysis and other methods are used to select features closely related to anchor strength, further optimizing the performance of the model. The accuracy and stability of the model are ensured by using the support vector machine algorithm (SVM) for model training and combining grid search to optimize hyperparameters. The system can monitor key parameters in real time through a sensor network and quickly obtain anchor strength data, thereby greatly shortening the time required for traditional detection methods and significantly improving construction efficiency.

[0029] The present invention provides an innovative detection solution for the engineering construction industry. With the continuous improvement of construction safety and efficiency requirements, the market demand for efficient and reliable anchor strength detection technology is growing. Through real-time monitoring and accurate prediction, the present invention can not only reduce the risks that may occur during the construction process, but also reduce the safety hazards caused by human errors, thereby improving the overall safety and reliability of the project. In addition, reducing the detection cycle and improving construction efficiency will directly reduce project costs and increase the return on investment. With the promotion and application of this technology, it is expected to attract more companies to adopt it, expand market share, and further promote the development of the entire industry towards intelligence and automation.

[0030] The present invention promotes the development of intelligent detection technology. Traditional anchor strength detection methods often rely on manual operation, which has certain time delays and human errors, and the pull-out test will cause certain damage to the anchor structure, which is difficult to meet the high requirements of modern engineering for efficiency and safety. The present invention adopts a prediction method based on support vector machines, combined with real-time sensor data acquisition, to achieve a rapid and accurate assessment of anchor strength. This method not only improves the automation level of data processing, but also enhances the reliability of the results, providing solid technical support for engineering construction.

[0031] The innovative application of the present invention in feature selection and model optimization provides a more scientific and systematic solution for anchor strength detection. Through correlation analysis and hyperparameter optimization, the model can better adapt to changes under different engineering conditions and ensure the accuracy of the test results. The promotion and application of this technology will help to improve the technical standards of the entire industry, promote related companies to occupy a dominant position in market competition, and further promote the progress of engineering safety management and the sustainable development of the industry. Overall, the present invention not only fills the technical gap, but also lays the foundation for promoting the intelligent and refined development of anchor strength detection technology.

[0032] The present invention solves the problem of long detection cycle and low efficiency: Traditional anchor strength detection often relies on manual detection or periodic inspection, which is not only time-consuming but also unable to provide real-time feedback on changes in anchor strength. The present invention can obtain anchor strength data in real time through real-time monitoring of the sensor network, greatly shorten the detection cycle, and improve construction efficiency. Real-time monitoring and prediction enable construction personnel to take timely measures when key parameters change during the mixing process, reducing potential risks.

[0033] The present invention solves the problems of human error and data reliability: traditional methods are easily affected by human factors, and manual operations may lead to errors in data collection and analysis, which in turn affect the assessment results of anchor strength. By adopting an automated analysis method based on support vector machine (SVM), the present invention effectively eliminates human errors and utilizes the high-precision characteristics of machine learning algorithms to improve the accuracy and reliability of data analysis and prediction. The present invention solves the problems of model adaptability and optimization: traditional anchor strength prediction methods lack sufficient adaptability and are difficult to effectively adjust according to different engineering environments and parameters. The present invention uses techniques such as feature selection and grid search to optimize hyperparameters to ensure that the support vector machine model can make accurate predictions based on different actual conditions, thereby improving the adaptability and universality of the model.

[0034] In the field of anchor strength testing, existing technologies are mostly focused on traditional physical pull-out test or prediction methods based on simple linear regression, which are difficult to cope with complex construction environments and real-time requirements. The introduction of the support vector machine (SVM) method provides an innovative solution for this field, filling the gap in existing technologies and promoting the industry to move towards intelligence and automation. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings herein are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description, serve to explain the principles of the present disclosure; Figure 1 Schematic diagram of the hardware part of the downhole anchor bolt anchoring strength prediction system based on the support vector machine method provided by an embodiment of the present invention; Figure 2 It is a flow chart of a method for predicting the anchoring strength of underground anchor bolts based on a support vector machine method provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of the software part of the downhole anchor bolt anchoring strength prediction system based on the support vector machine method provided by an embodiment of the present invention; Figure 4 This is the correlation analysis diagram of each feature; Figure 5 Fitting diagram of the mean actual anchoring force and the predicted anchoring force of the training set; Figure 6This is a comparison chart of the actual and predicted corrosion rates of the test set; Figure 7 It is the difference diagram between the actual value of the mean anchoring force in the training set and the model predicted value; Figure 8 It is the difference diagram between the actual value of the mean anchoring force after the anchoring agent acts on the test group and the value predicted by the model; Fig. 9 This is the application principle diagram of the downhole anchor bolt anchor strength prediction system based on the support vector machine method; Fig.10 A feature correlation analysis histogram provided by an embodiment of the present invention; In the figure: 1. Data acquisition system; 101. Data acquisition card; 102. Embedded processing unit; 2. Communication system; 3. Battery system; 301. Battery pack; 302. Battery management module; 4. Display and control interface; 401. Display screen; 402. Control panel; 5. Data acquisition module; 6. Data processing module; 7. Monitoring interface. DETAILED DESCRIPTION

[0036] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of 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 violating the connotation of the present invention, so the present invention is not limited by the specific implementation disclosed below.

[0037] The innovation of the present invention is that the present invention proposes a method and system for predicting the anchor strength of underground anchor bolts based on real-time monitoring of sensor networks, data processing and support vector machine (SVM) algorithm. The accuracy and quality of the data are ensured by real-time acquisition of key parameters in the mixing process and data cleaning, missing value processing and feature standardization. Correlation analysis and feature selection techniques are used to screen out features that are highly correlated with anchor strength, reduce the computational burden and improve the training efficiency and prediction accuracy of the model. The core innovation is the introduction of the SVM algorithm, which can accurately capture the nonlinear relationship between anchor strength and multiple factors, and optimize hyperparameters through grid search, further improving the prediction ability of the model. Combined with model evaluation and result visualization, users can intuitively understand the model performance and make optimization improvements. At the same time, the method effectively eliminates human errors and improves the accuracy and reliability of detection. The overall solution integrates real-time monitoring, intelligent algorithms, data processing and visualization, forming a systematic and intelligent solution with strong practicality and flexibility to meet the needs of complex construction environments.

[0038] The present invention combines sensor technology with machine learning methods, especially the support vector machine (SVM) method, to form an intelligent anchor strength monitoring system. The system can collect key parameters in the mixing process in real time, and provide more accurate anchor strength assessment through data analysis and model prediction. The innovation of the present invention lies in its integration of advanced data processing and analysis capabilities, which can achieve real-time monitoring and optimization of anchor performance in complex underground environments, providing reliable support for improving engineering safety and construction efficiency.

[0039] Embodiment 1, as Figure 1 As shown, the hardware part of the downhole anchor bolt anchoring strength prediction system based on the support vector machine method provided by the embodiment of the present invention includes a data acquisition system 1, a communication system 2, a battery system 3, and a display and control interface 4; The data acquisition system 1 comprises: Data acquisition card 101, function: convert analog signals from sensors into digital signals for real-time monitoring and recording.

[0040] Selection criteria: Choose a data acquisition card that supports multiple channels and high sampling rates (such as more than 1,000 times per second) to ensure high-precision data acquisition. Choose National Instruments PXIe-4322, which is a high-performance PXI Express data acquisition and signal generator module that is widely used in various tests and practices.

[0041] Embedded processing unit 102, function: responsible for data preprocessing, feature extraction and real-time calculation of support vector machine (SVM) model.

[0042] Selection criteria: Use an embedded device with strong computing power and storage space. This invention uses Raspberry Pi, which can quickly process sensor data.

[0043] The communication system 2 has the following functions: to realize the remote transmission of real-time data, so as to facilitate monitoring and analysis.

[0044] Selection criteria: Choose low-power, long-distance transmission wireless technologies, such as LoRa, Zigbee or NB-IoT, to adapt to the complex environment underground.

[0045] The battery system 3 includes a battery pack 301 , and the selection criteria are: using high temperature resistant, long life lithium batteries to ensure stable operation in an underground environment.

[0046] Battery management module 302, function: providing voltage stabilization and overload protection to ensure the safe operation of the device under extreme conditions.

[0047] The display and control interface 4 includes a display screen 401, function: real-time display of monitoring data, system status and alarm information. Selection criteria: select a high temperature resistant and waterproof LCD or LED display screen to ensure good readability in the underground environment.

[0048] Control panel 402, function: used to set monitoring parameters and system calibration to ensure convenient operation.

[0049] Design standards: It should be simple and easy to understand, so that operators can quickly get started.

[0050] This underground anchor bolt anchor strength prediction system (intelligent monitoring system) based on the support vector machine method can comprehensively monitor the strength and performance of underground anchoring agents by selecting appropriate sensors and hardware components, thereby improving the safety and reliability of underground projects. The various parts of the system work closely together to achieve real-time data collection and analysis, providing strong support for engineering decision-making.

[0051] Embodiment 2, as Figure 2 As shown, the method for predicting the anchoring strength of underground anchor bolts based on the support vector machine method provided in an embodiment of the present invention includes: S1, data preprocessing: cleaning data, handling missing values, and standardizing features; To ensure the effectiveness of the subsequent support vector machine model training; S2, feature selection: feature selection methods are used to select features with high correlation with anchor strength; Select the features with high correlation with anchor strength. Correlation analysis needs to be obtained after a series of processing. Figure 4 This is the result of correlation analysis; S3, model training: Use support vector machine method SVM for model training, use grid search to optimize hyperparameters, and complete the construction of support vector machine model; Support Vector Machine (SVM) is a machine learning method based on geometric principles, used to solve problems such as classification, regression, and anomaly detection. Its core concept is based on maximizing the classification boundary interval, and the goal is to find an optimal hyperplane to maximize the separation of data points of different categories.

[0052] (1) (Grid search steps to optimize hyperparameters) Define parameter grid; When building a machine learning model, choosing the right hyperparameters is crucial to the performance of the model. Define a parameter grid param_grid, which contains the following hyperparameter options: Python param_grid = { 'svr__C': [1, 10, 100, 1000], # Regularization parameter 'svr__gamma': ['scale', 'auto'], # coefficient of kernel function 'svr__epsilon': [0.01, 0.1, 0.2], # tolerance 'svr__kernel': ['rbf', 'poly']# Kernel function type } C: Regularization parameter that controls the complexity of the model. A smaller C value will result in a simpler model (tends to underfit), while a larger C value will make the model more complex (tends to overfit).

[0053] Gamma: controls the range of influence of the Gaussian kernel (RBF).

[0054] epsilon: In regression problems, SVR allows for some margin of error. This parameter controls the "tolerance" of the model when making predictions.

[0055] Kernel: Select different kernel functions (such as rbf and poly) to determine how the model maps input features.

[0056] Set up a grid search; the grid search systematically traverses all possible parameter combinations to find the best parameters. The code is: Python from sklearn.model_selection import GridSearchCV grid_search = GridSearchCV(pipeline, param_grid, cv=5, scoring='neg_mean_absolute_error') pipeline: This is the defined pipeline, which contains data preprocessing and models.

[0057] param_grid: The parameter grid passed in.

[0058] cv=5: specifies the use of 5-fold cross validation, which means that the data will be divided into 5 parts, and 1 part will be selected as the validation set each time, and the remaining 4 parts will be used as the training set. This process is repeated 5 times to ensure that each part is used as a validation set.

[0059] scoring='neg_mean_absolute_error': specifies the evaluation metric as negative mean absolute error. The negative value is because GridSearchCV requires a "bigger is better" scoring standard, so using negative values ​​makes a smaller MAE (indicating a better model) become a larger scoring value.

[0060] Perform a grid search; start the grid search process by calling the fit method: Python grid_search.fit(X_train,y_train) In this step: Training model: For each set of parameter combinations, use the training set for training.

[0061] Evaluate performance: Evaluate model performance on the validation set and record the score for each set of parameters.

[0062] Selecting the best parameters: After evaluating all combinations, GridSearchCV will automatically select the best performing parameter combination.

[0063] Check the optimal parameters; After the grid search is complete, view the best parameter combination: Python print(f"Best parameters:{grid_search.best_params_}") grid_search.best_params_ returns the parameter combination that performed best in the cross validation.

[0064] Use the best model; retrain the model using the best parameter combination: best_svr_model = grid_search.best_estimator_ best_svr_model.fit(X_train, y_train) This will use the optimal parameter configuration to train the final SVR model.

[0065] (2) Training set and test set division: randomly divide the processed data set into 80% training set and 20% test set; In the step (1), a grid search is performed using an RBF radial basis function or a POIY polynomial kernel; RBF radial basis function is suitable for nonlinear problems to capture complex relationships; The POIY polynomial kernel is used for polynomial feature mapping and is suitable for data with polynomial relationships; The method of using RBF radial basis function technology to solve the interpolation problem includes: selecting P basis functions and training data, and the form of each basis function is: ; In the formula, is a nonlinear function, and the training data points yes The center of the circle, the basis function is the data in the input space With Center The distance is used as the independent variable of the function. Since the distance is radially homogeneous, the function is called a radial basis function. The difference function based on the radial basis function technology is defined as a linear combination of basis functions, and the expression is: ; In the formula, In a two-dimensional plane, is the center of the circle, The distance between the data and the center of the circle is independent of the position and size of the data, and the distance between the points on the circle and the center of the circle is equal.

[0066] The hyperparameters include: The C parameter is a regularization parameter in SVM, which is used to balance the trade-off between margin and misclassified samples; Gamma parameter, used to define the influence range of a single training sample on the model; The Epsilon parameter is used to control the error tolerance of the support vector machine model, the allowable error range, and define the support vector machine model's tolerance for noise in the training data.

[0067] S4, model evaluation: evaluate the performance of the support vector machine model on the training set and test set by calculating the mean absolute error MAE and mean square error MSE indicators; S5, Result Visualization: Use visualization tools to show the comparison between model prediction results and actual values.

[0068] Example 3, as another embodiment of the present invention, Figure 3 As shown in the figure, the software part of the intelligent prediction system for underground anchor bolt anchoring strength based on the support vector machine method includes: The data acquisition module 5 is used to collect the anchoring strength data after the anchoring agent is mixed in real time; Data processing module 6, using support vector machine (SVM) method to analyze the anchor strength data; Monitoring interface 7, used to display anchor strength monitoring results and their changing trends; The data acquisition module 5 includes: an infrared rangefinder, a drilling depth measurement, etc.

[0069] The data processing module 6 also includes: a data preprocessing function submodule for removing noise and abnormal values.

[0070] The monitoring interface 7 adopts a graphical interface design so that users can intuitively understand the monitoring data.

[0071] Exemplarily, the data collected by the data collection module 1 include: sensor data: deploying sensors to collect physical and chemical property data of the anchoring agent, such as pressure; Ultrasonic data: used to measure mixing uniformity.

[0072] Environmental data: such as temperature, humidity, pH value, etc.

[0073] Strength test data: Anchorage strength data obtained through laboratory tests.

[0074] Correlation analysis: In this paper, correlation analysis is an important step for evaluating the linear relationship between features and target variables (in this case, "mean anchoring force / kN"). Correlation analysis can help us understand which input features have a strong influence on the target variable, so as to select features that are useful for the prediction task and eliminate redundant features.

[0075] The specific process is as follows: 1. Data preprocessing: Filling missing values: First, we fill missing values ​​for the numerical features in the data. By calculating the mean of each numerical feature and using it to fill missing values, we can ensure the integrity of the data and avoid the impact of missing values ​​on the analysis results.

[0076] One-hot encoding: For non-numeric features (such as categorical variables), we perform one-hot encoding to convert them into numerical features. Non-numeric data is processed by the pd.get_dummies() function, where drop_first=1 is used to avoid the dummy variable trap.

[0077] 2. Calculate the correlation matrix: Correlation calculation: Pearson correlation coefficient is used to measure the linear relationship between each pair of features in the data set. The value of Pearson correlation coefficient ranges from -1 to 1, where: Values ​​close to 1 indicate a positive correlation, with eigenvalues ​​increasing together.

[0078] Values ​​close to -1 indicate negative correlation and the eigenvalues ​​decrease together.

[0079] Values ​​close to 0 indicate no linear relationship.

[0080] We calculated the correlation matrix between all features of the dataset and visualized the results in the form of a heat map. The heat map can intuitively show the strength and direction of the correlation between features.

[0081] 3. Select features that are highly correlated with the target variable: Setting a threshold: We set a correlation threshold and select features whose absolute value of the correlation coefficient with the target variable (“mean anchor force / kN”) is greater than the threshold. These features are input variables that have a strong influence on the target variable.

[0082] Manual addition of key features: Based on domain knowledge, features that may have important impact on anchoring force prediction (such as “surrounding rock” and “anchorage length”) were manually added to the list of high-relevance features.

[0083] 4. Evaluate the impact of each feature on the target variable: The absolute value of the correlation coefficient between the selected feature and the target variable is calculated and used as the influence coefficient of each feature. The higher the influence coefficient, the more important the feature is for predicting the mean anchoring force.

[0084] These features and their impact coefficients are organized into a data frame (impact_df) and sorted by impact coefficient to highlight the features that have the greatest impact on the target variable.

[0085] 5. Result visualization: In order to view the correlation more intuitively, the present invention uses the seaborn library to draw a correlation heatmap. The heatmap shows the correlation relationship between different features, where the color depth represents the strength of the correlation, and the labeled value shows the correlation coefficient between each pair of features.

[0086] To ensure that Chinese characters are displayed correctly, we set the font configuration of matplotlib and adjusted the image size and layout to make the heat map clearer and easier to understand.

[0087] according to Figure 4 The correlation analysis chart of each feature can intuitively show the correlation between each data in the data set and the mean anchoring force. This data set uses the mean anchoring force to reflect the anchoring strength. The higher the grayscale, the higher the correlation. Figure 4 The data show that the surrounding rock characteristics of the anchoring section have a high correlation, among which the correlation between "sandstone layer" and "5# coal" is particularly strong.

[0088] The specific contents of selecting the features with high correlation with anchoring strength are described as follows: After the above correlation analysis of the data, Fig.10It can be seen intuitively that the correlation analysis of each feature on the final anchoring force mean value is performed, and the features with high correlation are selected. Since the data text data in the example of the present invention uses one-hot encoding, all the columns of the surrounding rock of the anchoring section show correlation. In fact, it is a correlation analysis of one column of data. It can be seen from the figure that the correlations of the buried depth / m, the surrounding rock of the anchoring section, and the anchoring length / cm are all high. Therefore, the features with high correlation are still the three columns of buried depth / m, the surrounding rock of the anchoring section, and the anchoring length / cm, and they are used in the subsequent construction of the support vector machine model and model training.

[0089] In the data processing module, the environment for constructing a support vector machine (SVM) model includes: Based on the Python programming language and the Jupyter Notebook platform. This environment provides an interactive interface that facilitates experiments and development for data scientists and machine learning engineers. As a powerful programming language, Python has a rich set of machine learning libraries (such as scikit-learn, NumPy, and pandas), making it easy and efficient to implement support vector machine models. Jupyter Notebook provides a web-based interface that allows users to execute code step by step in cells and view results in real time, making it easy to debug and visualize data. The flexibility and scalability of the overall environment, combined with powerful data processing and analysis capabilities, enable users to efficiently build, train, and evaluate SVM models.

[0090] Exemplarily, constructing a support vector machine (SVM) model includes the following contents.

[0091] (1) Data cleaning refers to processing incomplete, inaccurate or inconsistent parts of a data set for subsequent analysis and modeling. Common data cleaning operations include processing missing values, processing outliers, and removing duplicate data.

[0092] Code example: python import pandas as pd # Read the dataset data = pd.read_csv(r'C:\Users\Song668\Desktop\data1.csv', encoding='GBK') # Handling missing values data.dropna(inplace=1) # remove rows containing missing values # Remove duplicate data data.drop_duplicates(inplace=1) Exemplary, related principles include, handling missing values: missing values ​​may affect the performance of the model, so they need to be handled. Common methods include deleting rows or columns containing missing values, filling missing values ​​with statistical values ​​(such as mean, median), or using machine learning models to fill them.

[0093] Handling outliers: Outliers may be caused by errors in the data collection process or special circumstances, and need to be handled according to the specific situation. They can be detected and handled based on statistical methods (such as the 3σ principle), box plots, etc.

[0094] Remove duplicate data: Duplicate data may affect the training and evaluation of the model, so it is necessary to remove duplicate data. You can use the drop_duplicates() method to remove duplicate rows.

[0095] (2) Feature Engineering. The feature engineering part of the present invention mainly involves the following steps: (2.1) Data reading: Use Pandas' read_csv function to read the data set from the CSV file. The file path is C:\Users\Song668\Desktop\data1.csv, and GBK encoding is used for reading.

[0096] data = pd.read_csv(r'C:\Users\Song668\Desktop\data1.csv', encoding='GBK'); (2.2) Separation of features and target variables: Assuming that the dataset contains features and target variables, the target variable (anchor strength) is separated from the feature matrix through the drop function to form the feature matrix X and the target variable vector y_actual.

[0097] In the code example, Python: X = data.drop('Anchorage strength (MPa)',axis=1); y_actual = data['anchor strength (MPa)']3; Then randomize the data set: The sample method is used to randomize the data set to avoid the impact of data order on model training. Here, frac=1 is used to retain all data, and random_state=50 is used to set the random seed to ensure the repeatability of the results.

[0098] data_shuffled=data.sample(frac=1,random_state=50); Although the sample method is very practical, there are some cases where you may need to improve it or consider other solutions: 1. Stratified sampling: In some cases, it may be necessary to perform stratified sampling on the data to ensure that different categories of data are evenly distributed. Stratified sampling can be achieved by combining the groupby and sample methods.

[0099] 2. Weighted sampling: If some samples are more important than other samples, you can use weights for sampling. Although the Pandas sample method does not directly support weighted sampling, it can be achieved by modifying the weight distribution of the dataset.

[0100] 3. Memory efficiency: For very large data sets, using frac=1 may take up a lot of memory. You can consider using the frac parameter to sample a portion of the data first, or use a batch processing method.

[0101] 4. Combine with other methods: In some cases, data augmentation or other preprocessing steps may be required after random sampling to improve model performance.

[0102] (2.3) Re-obtain the randomized features and target variables: Re-obtain the feature matrix X_shuffled and target variable vector y_actual_shuffled from the randomized dataset.

[0103] In the code example, Python: X_shuffled = data_shuffled.drop('Anchorage strength (MPa)',axis=1); y_actual_shuffled = data_shuffled['anchor strength (MPa)']; Data normalization is a commonly used data preprocessing technique that can eliminate dimensionality effects, accelerate model convergence, improve model performance, and enhance model interpretability. Z-score normalization is mainly used in the present invention.

[0104] Code example: Python from sklearn.preprocessing import StandardScaler # Initialize the normalizer scaler = StandardScaler() # Standardize the data data_scaled = scaler.fit_transform(data); Z-score standardization: Standardize the data according to its mean and standard deviation so that the mean of the standardized data is 0 and the standard deviation is 1.

[0105] ; In the formula, is the original data, is the mean, is the standard deviation.

[0106] Z Score is a method often used for data standardization. In a multi-indicator evaluation system, due to the different nature of each evaluation indicator, they usually have different orders of magnitude and units. If the original data is used directly, the role of indicators with higher values ​​in the analysis will be highlighted, and the role of indicators with lower values ​​will be relatively weakened.

[0107] Therefore, in order to ensure the reliability of the results, the original data needs to be standardized.

[0108] For the sample sequence To standardize, the expression is: ; The new sequence generated It is dimensionless data with mean 0 and variance 1.

[0109] Its python code is: #Use Z-score method to standardize data Python: scaler = StandardScaler() X_scaled = scaler.fit_transform(X_shuffled); In this way, the feature engineering part completes the operations of reading data, separating features and target variables, randomizing data sets, etc. The next step is model training, prediction, and performance evaluation.

[0110] (3) Feature selection: The code does include the model training process. Specifically, GridSearchCV will perform multiple model trainings based on the provided hyperparameter grid and automatically select the optimal hyperparameter combination. Therefore, grid_search.fit(X_train_poly, y_train) is a key step in the training process, which adjusts the model parameters based on the given training data X_train_poly and y_train.

[0111] Initialize the SVR model and perform grid search to adjust parameters; In the code example, python: param_grid = { 'C': [1, 10, 100, 1000], 'gamma': ['scale', 'auto'], 'epsilon': [0.01, 0.1, 0.2], 'kernel': ['rbf', 'poly'] } svr = SVR() grid_search = GridSearchCV(svr, param_grid, cv=5, scoring='neg_mean_absolute_error') grid_search.fit(X_train_poly,y_train); The above code defines a parameter grid and uses cross validation (cv=5) to perform a grid search to find the optimal parameter combination. The scoring uses the negative mean absolute error (MAE) as the evaluation criterion.

[0112] The kernel function is a function used in support vector machines (SVM) to map data into a high-dimensional space. In the code, two possible kernel functions are selected: RBF (Radial Basis Function): A commonly used kernel function that is suitable for nonlinear problems and can capture complex relationships.

[0113] POIY (Polynomial Kernel): used for polynomial feature mapping, suitable for data with polynomial relationships.

[0114] The grid search will test these two kernel functions and finally select the one that performs best. The results of the best kernel function are shown in the optimal kernel function selection and parameter setting, which are as follows: Best parameters:{'svr_C':1000, 'svr_epsilon':0.01, 'svr_gamma':'scale','svr_kernel':'rbf'} It can be seen that the best kernel function is the RBF kernel function. The best parameter combination returned by the grid search {'svr__C':1000,'svr__epsilon':0.01,'svr__gamma':'scale','svr__kernel':'rbf'} indicates that the optimal SVR model is a regression model using the RBF kernel, which has a low tolerance for errors (C=1000) and a certain tolerance for the error between the predicted value and the actual value (epsilon=0.01). In addition, gamma='scale' is selected as the parameter of the RBF kernel to balance the complexity of the model by automatically calculating the optimal gamma value. These parameter combinations indicate that the model selects a more complex configuration during training, adapts to the nonlinear characteristics of the data, and avoids overfitting as much as possible.

[0115] Kernel: SVM can be applied to nonlinear problems with the kernel technique. The kernel function is used to map data from the original feature space to a higher-dimensional feature space, making the data easier to segment. Commonly used kernel functions include linear kernel functions, polynomial kernel functions, and radial basis function (RBF) kernel functions.

[0116] The RBF kernel function is one of the most commonly used kernel functions. It can realize complex nonlinear mapping in high-dimensional space and is therefore suitable for data with complex structures.

[0117] The method of using radial basis function technology to solve the interpolation problem is to choose Basis Functions training data, the form of each basis function is: ; In the formula, is a nonlinear function, and the training data points yes The center of the circle, the basis function is the data in the input space With Center The distance is used as the independent variable of the function. Since the distance is radially homogeneous, the function is called a radial basis function. The difference function based on the radial basis function technology is defined as a linear combination of basis functions, and the expression is: ; In the formula, In a two-dimensional plane, is the center of the circle, The distance between the data and the center of the circle is independent of the position and size of the data. The distance from the point on the circle to the center of the circle is equal, which is radial. Substituting it into the mapping function is the radial basis function.

[0118] C parameter: Parameter C is a regularization parameter in SVM, which is used to balance the trade-off between margin and misclassified samples. A larger C value means a more severe penalty for errors, which will lead to a more complex model.

[0119] When the C value is small, the model tends to choose a larger interval and tolerates more misclassified samples; when the C value is large, the model will more strictly pursue zero training error, which may lead to overfitting.

[0120] Gamma parameter: The parameter gamma is a parameter of the RBF kernel function, which defines the range of influence of a single training sample on the model. A larger gamma value will make the decision boundary more complex and may cause the model to overfit.

[0121] In SVM, the larger the gamma is, the more details the SVM model will remember for each sample in the training data, and the decision boundary will be more distorted and irregular.

[0122] Epsilon parameter: The parameter epsilon controls the error tolerance of the support vector machine model, that is, the allowable error range. It defines the tolerance of the support vector machine model to noise in the training data.

[0123] A smaller epsilon value means that the support vector machine model has a lower tolerance for errors. The support vector machine model will try harder to fit each sample in the training data, which may lead to overfitting. A larger epsilon value allows the support vector machine model to have more errors, making the support vector machine model more suitable.

[0124] The training set and test set division include,data segmentation: randomly divide the processed data set into a training set (80%) and a test set (20%) for the subsequent training and verification of the support vector machine model.

[0125] Exemplarily, monitoring interface 7 visualizes the anchor strength monitoring results and their changing trends; The result visualization includes: 1) Model performance analysis, including: Training group and test group division: In the code example, Python: from sklearn.model_selection import train_test_split X_train, X_test, y_train, y_test = train_test_split(X_scaled, y_actual, test_size=0.2, random_state=42); X_scaled: feature matrix, feature data after standardization. y_actual: target variable, i.e., mean anchor force. test_size: the proportion of the test set, here set to 0.2, means that 20% of the data set is divided into the test set, and the remaining 80% is used as the training set. random_state: random seed, used to control the random division of data. Setting the same random seed can ensure that the results of each division are consistent. Here it is set to 50. To evaluate the model performance of support vector machine (SVM) for corrosion rate prediction and safety assessment of buried x70 steel pipelines, some evaluation indicators can be used to measure the accuracy and generalization ability of the model.

[0126] The mean absolute error and root mean square error are used as two evaluation indexes to analyze the prediction performance of the model. The calculation formula is: ; ; In the formula, is the mean absolute error, is the root mean square error; is the predicted value of corrosion rate, in mm / a; It is the actual value of corrosion rate, the unit is mm / a.

[0127] Root Mean Squared Error (RMSE): The square root of MSE, which is used to measure the average error between the predicted value and the true value. The smaller the RMSE, the higher the prediction accuracy of the model.

[0128] Mean Absolute Error (MAPE): Calculates the average absolute error between the predicted value and the true value. The smaller the MAE, the higher the prediction accuracy of the model.

[0129] Model performance of the training group: mean absolute error: 0.0097; mean square error: 0.00096; Test group model performance: mean absolute error: 0.0088; mean square error: 0.00084; 2) Training set results: The code example is: y_train_pred = best_svr_model.predict(X_train) # Make predictions on the training set # Performance indicators on the training set train_mae = mean_absolute_error(y_train, y_train_pred) train_mse = mean_squared_error(y_train, y_train_pred) print("Training set mean absolute error:", train_mae) print("Training set mean square error:",train_mse; exist Figure 5 The fitting diagram of the actual mean anchoring force and the predicted mean anchoring force of the training set shows the relationship between the actual value of the mean anchoring force and the predicted value of the SVM model. The blue points represent the actual values ​​and the red points represent the predicted values. Figure 5 The fitting of each training set can be intuitively seen in the figure. There is a small error in each training result. At the same time, the actual value and the predicted value of each group are consistent, which shows that this prediction model has a certain accuracy and lays a foundation for the prediction of the test set.

[0130] 3) Test set results.

[0131] The code instructions are: y_pred = svm_model.predict(X_test_scaled) # Make predictions on the test set; like Figure 6 The comparison of the actual and predicted values ​​of the corrosion rate in the test set shows that the fitting of the actual mean anchoring force and the predicted mean anchoring force in the 12 test sets can be seen. Preliminary observations show that this prediction model has good accuracy. Figure 6 There are still small errors in the prediction, and the predicted value and the actual value tend to be consistent overall.

[0132] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0133] Prediction result analysis: In order to more accurately and clearly see whether the prediction model is successful, the present invention continues to conduct a more in-depth study on the training set results and the test set results, mainly analyzing the distribution of the difference between the predicted value and the actual value.

[0134] The difference between the actual value of the mean anchoring force in the training set and the model predicted value is as follows: Figure 7 As shown, from Figure 7 It can be seen that the difference has positive and negative phenomena, indicating that the model error has a certain floating phenomenon, which further shows the relative accuracy of the model. The vertical axis represents the difference between the actual anchoring force mean and the predicted anchoring force mean. It can be clearly seen that the difference of each group in the training set is less than 0.1KN. Therefore, it can be concluded that the training set of the SVM model of the present invention is successful.

[0135] Figure 8 What is shown in the figure is the difference between the actual value of the mean anchoring force of the test group after the anchoring agent acts and the model predicted value. The bar chart clearly shows the difference of each test group. According to the vertical axis, the difference of each group is less than 0.1KN, which shows that the SVM model of the present invention has good performance and can accurately predict the mean anchoring force based on the provided anchoring strength data. Specifically including: Good performance: If both the training and test set errors are low, it indicates that the model has learned the patterns in the data well and can generalize effectively to unseen data.

[0136] Overfitting: If the error between the training set and the test set is very different, it indicates overfitting, that is, the model performs well on the training data but performs poorly on unseen data.

[0137] Underfitting: If both the training and test set errors are high, it may indicate underfitting, that is, the model is too simple and cannot capture the patterns in the data.

[0138] Model optimization: If overfitting is detected, consider reducing model complexity (e.g., regularization) or collecting more diverse data. If underfitting is detected, consider using a more complex model or engineering more features.

[0139] From the analysis of the prediction results, it can be concluded that the SVM model of the present invention has learned the data pattern very well and can be effectively generalized to unseen data.

[0140] Experiment 2, input model.

[0141] 1) Ensure the environment is ready.

[0142] Choose the right embedded platform: Make sure the embedded device supports the Python environment; Install Python and libraries: Install Python and required libraries on the embedded device.

[0143] For example: pip install numpy pandas matplotlib scikit-learn; 2) Data storage.

[0144] Data source: Embedded devices usually have different storage methods. Data can be stored in an SD card, USB drive, or in the device's memory. Make sure to read the CSV file.

[0145] Data reading: The data reading method in the sample code needs to ensure that the path is correct. For example: data=pd.read_csv(' / path / to / data1.csv',encoding='utf-8'); 3) Optimize and simplify the code.

[0146] Dealing with resource limitations: Embedded devices may not have strong computing power, so you can consider simplifying the model, reducing the size of the data, using fewer features, or choosing a lighter-weight machine learning algorithm.

[0147] Reduce printed information: In an embedded environment, excessive output will take up resources, so you can appropriately reduce log information.

[0148] Experiment 3, establishment of interactive interface.

[0149] Data reading and processing: Read CSV files from the specified path, and parse the data in GBK encoding. Use the drop method to separate the target variable (anchor strength) from the feature, and use the mean of the feature and target variable to fill the missing values ​​to ensure data integrity.

[0150] Data shuffling and partitioning: The data is randomly shuffled using the sample method, and then the data is divided into a training set and a test set using train_test_split, with the test set accounting for 20%. This partitioning helps the evaluation and generalization ability of the model.

[0151] Create a machine learning pipeline: Use Pipeline to standardize the data (StandardScaler), generate polynomial features (PolynomialFeatures), and apply support vector regression (SVR). This modular design simplifies the machine learning process.

[0152] Hyperparameter optimization: GridSearchCV was used to perform grid search on the hyperparameters of the SVR model. Multiple parameter combinations were set and 5-fold cross validation (cv=10) was used to evaluate the performance of the model. The goal was to minimize the prediction error (neg_mean_absolute_error).

[0153] Best model training: After selecting the best hyperparameter combination, retrain the model to obtain the final predictive ability.

[0154] Graphical User Interface (GUI): A user interface was created using tkinter, allowing the user to input feature data and click buttons to predict anchor strength. The interface also includes a button to display a graph of anchor strength versus time, enhancing user experience and interactivity.

[0155] Experiment 4, overall system analysis.

[0156] Environmental adaptability: All components must meet the high temperature, high humidity and high pressure requirements of the underground working environment.

[0157] Power supply: Consider using explosion-proof power supply to ensure stable operation of the system underground.

[0158] System integration: Ensure the compatibility of sensors, communication modules and display control systems, and develop a good system architecture to facilitate subsequent maintenance and upgrades.

[0159] In the intelligent detection system of anchor strength in the well, the selection of sensors should give priority to the characteristics of high pressure resistance, corrosion resistance and high temperature resistance; the communication system should choose the anti-interference and long-distance transmission scheme; and the display and control system should ensure easy operation and normal operation in harsh environments. Through reasonable system integration, real-time monitoring and intelligent analysis of anchor strength can be effectively realized.

[0160] By selecting appropriate sensors and hardware components, the intelligent monitoring system can comprehensively monitor the strength and performance of downhole anchoring agents and improve the safety and reliability of underground projects. The various parts of the system work closely together to achieve real-time data collection and analysis, providing strong support for engineering decision-making.

[0161] like Fig. 9 , which is an application scenario diagram of a downhole anchor bolt anchoring strength prediction system based on a support vector machine method provided in an embodiment of the present invention.

[0162] Experiments have shown that existing manual inspection technologies rely on workers' experience and judgment, are easily affected by subjective factors, and have low accuracy.

[0163] Existing technologies have poor real-time performance: Traditional methods cannot provide real-time data and are unable to quickly respond to problems during construction.

[0164] Existing technology data processing is complex: Traditional monitoring methods often require complex mathematical models and a lot of calculations, resulting in low efficiency.

[0165] Insufficient adaptability of existing technologies: Existing technologies are difficult to adapt to changing geological conditions and environmental factors.

[0166] The present invention innovatively introduces the support vector machine algorithm: the system uses the support vector machine as the core algorithm, which can effectively process high-dimensional data, perform nonlinear classification, and improve the analysis ability of monitoring data.

[0167] The present invention proposes automated monitoring: the system can realize automated and intelligent monitoring of the anchoring strength after the anchoring agent is mixed, thereby reducing manual intervention and human errors.

[0168] The model proposed in the present invention has high adaptability: the support vector machine algorithm has good generalization ability, can adapt to different geological conditions and anchoring agent ratios, and improve the accuracy and reliability of monitoring.

[0169] The present invention has the following features in terms of practicality: Real-time data feedback: The system can collect anchor strength data in real time and provide timely feedback on construction status, allowing construction personnel to make quick decisions and ensure project safety.

[0170] Easy to integrate: The monitoring system can be seamlessly integrated with existing underground engineering monitoring equipment, making it easy to promote and apply.

[0171] Reduce costs: By improving monitoring efficiency and accuracy, it can effectively reduce subsequent maintenance costs and reduce the risk of accidents caused by insufficient anchoring strength.

[0172] Application example 1. Mining industry safety monitoring: In mine operations, anchoring agents are used to reinforce rock formations to improve mine safety. The intelligent monitoring system can monitor the anchoring strength in real time and promptly warn of potential risks.

[0173] Application Example 2. Construction Optimization: By analyzing the anchor strength data, the use ratio of the anchoring agent can be optimized to improve the construction efficiency and quality.

[0174] Application example 3. Monitoring during tunnel construction: During tunnel construction, the strength of the anchor directly affects the stability of the tunnel. The system can monitor the performance of the anchor to ensure the safety and smooth progress of the construction process.

[0175] Application example 4. Civil engineering foundation reinforcement: In infrastructure construction, anchoring technology is often used to reinforce soil and rock structures. The intelligent monitoring system can provide accurate data support to help engineers make scientific decisions.

[0176] Application Example 5. Geological disaster monitoring and risk assessment: By monitoring the strength changes of anchoring agents, the system can help predict and assess geological disaster risks and provide data support for disaster prevention and mitigation.

[0177] Application Example 6. Downhole stability in the oil and gas industry: Anchoring agents are also widely used in the drilling and production of oil and gas wells. The monitoring system can ensure the stability of the well wall and reduce the risk of accidents during the production process.

[0178] The commercial advantages of the present invention include: Improve safety: Through real-time monitoring and early warning, the safety of mines, tunnels and other projects can be significantly improved, the accident rate can be reduced, thereby reducing losses and protecting the lives of workers.

[0179] Reduce costs: By optimizing the use and mixing ratio of anchoring agents, material waste can be reduced and construction costs can be reduced. At the same time, preventive maintenance can reduce the cost of subsequent repairs and accident handling.

[0180] Improve efficiency: Real-time data monitoring and analysis can help engineers make faster decisions, improve construction efficiency, and shorten project cycles.

[0181] Market competitiveness: In modern engineering, safety and efficiency are the key to competition. The introduction of intelligent monitoring systems will provide enterprises with technological advantages and enhance market competitiveness.

[0182] In addition, with the continued growth of infrastructure construction and mining development, the demand for safety monitoring systems will continue to rise, and the market potential is huge.

[0183] With the development of the Internet of Things (IoT), big data and artificial intelligence technologies, the technical level of intelligent monitoring systems will continue to improve, providing the market with more innovative solution technical support.

[0184] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for predicting the anchor strength of underground anchor bolts based on support vector machine method, characterized in that: The method includes: S1, data preprocessing: cleaning data, handling missing values, and standardizing features; S2, feature selection: feature selection methods are used to select features with high correlation with anchor strength; S3, model training: Use support vector machine method SVM for model training, use grid search to optimize hyperparameters, and complete the construction of support vector machine model; S4, model evaluation: evaluate the performance of the support vector machine model on the training set and test set by calculating the mean absolute error MAE and mean square error MSE indicators; S5, Result Visualization: Use visualization tools to show the comparison between model prediction results and actual values.

2. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 1, characterized in that: In step S1, the normalized features are normalized using Z-score.

3. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 1, characterized in that: In step S2, feature selection includes: (2.1) Data reading: Use Pandas’ read_csv function to read the dataset from the CSV file; (2.2) Separation of features and target variables: The data set contains features and target variables. The target variable is separated from the feature matrix through the drop function to form the feature matrix X and the target variable vector y_actual, and the data set is randomized. (2.3) Re-obtain the randomized features and target variables, and re-obtain the feature matrix X_shuffled and the target variable vector y_actual_shuffled from the randomized dataset.

4. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 3 is characterized in that: In step (2.1), the file is read using GBK encoding; In step (2.2), the data set is randomized, including: using the sample method to randomize the data set, using frac=1 to retain all data, and random_state=50 to set the random seed.

5. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 1, characterized in that: In step S3, the support vector machine method SVM is used for model training, and the hyperparameters are optimized using grid search, including: (1) Define the parameter grid. When building a machine learning model, systematically traverse all parameter combinations to find the best parameters. Start the grid search process by calling the fit method. After the grid search is completed, check the best parameter combination and retrain the model using the best parameter combination: (2) Training set and test set division: The processed data set is randomly divided into 80% training set and 20% test set.

6. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 5, characterized in that: In step (1), a grid search is performed using RBF radial basis function or POIY polynomial kernel, and POIY polynomial kernel is used for polynomial feature mapping; Use RBF radial basis function technology to solve the interpolation problem, select Basis function training data, the form of each basis function is: ； In the formula, is a nonlinear function, and the training data points yes The center of the circle, the basis function is the data of the input space With Center The distance is used as the independent variable of the function. Since the distance is radially homogeneous, the function is called a radial basis function. The difference function based on the radial basis function technology is defined as a linear combination of basis functions, and the expression is: ； In the formula, In a two-dimensional plane, is the center of the circle, The distance between the data and the center of the circle is independent of the position and size of the data, and the distance between the points on the circle and the center of the circle is equal.

7. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 5, characterized in that: Hyperparameters include: The C parameter is a regularization parameter in SVM, which is used to balance the trade-off between margin and misclassified samples; Gamma parameter, used to define the influence range of a single training sample on the model; The Epsilon parameter is used to control the error tolerance of the support vector machine model, the allowable error range, and define the support vector machine model's tolerance for noise in the training data.

8. The method for predicting the anchoring strength of underground bolts based on support vector machine method according to claim 1, characterized in that: In step S4, the calculation formulas for the mean absolute error MAE and mean square error MSE indicators are: ； ； In the formula, is the mean absolute error, is the root mean square error; is the predicted value of corrosion rate, in mm / a; It is the actual value of corrosion rate, in units of mm / a.

9. A downhole anchor bolt anchoring strength prediction system based on support vector machine method, characterized in that: The system implements the method for predicting the anchoring strength of underground bolts based on the support vector machine method as described in any one of claims 1 to 8, and the hardware part of the system includes: A data acquisition system (1) is provided with a data acquisition card (101) for converting analog signals from sensors into digital signals for real-time monitoring and recording; An embedded processing unit (102), used for data preprocessing, feature extraction and real-time calculation of a support vector machine (SVM) model; Communication system (2) for remote transmission of real-time data to facilitate monitoring and analysis; A battery system (3) is provided with a battery pack (301) and a battery management module (302), wherein the battery pack (301) is used to provide power for an underground environment, and the battery management module (302) is used to provide voltage stabilization and overload protection; A display and control interface (4), provided with a display screen (401) for displaying monitoring data, system status and alarm information in real time; The control panel (402) is used to set monitoring parameters and system calibration.

10. The downhole anchor bolt anchoring strength prediction system based on support vector machine method according to claim 9, characterized in that: The software part of the system includes: A data acquisition module (5) is used to collect anchor strength data after the anchoring agent is mixed in real time; A data processing module (6) is used to analyze the anchor strength data using a support vector machine (SVM) method; The monitoring interface (7) adopts a graphical interface for displaying the anchor strength monitoring results and their changing trends.

Images