Shield tunneling machine tunneling parameter optimization method and system based on machine learning algorithm

Abstract

The invention belongs to the technical field of tunneling construction, and provides a shield tunneling machine tunneling parameter optimization method and system based on a machine learning algorithm, and the method comprises the steps: firstly determining correlation coefficients between different parameters in current shield tunneling data, synchronous grouting monitoring data and ground surface settlement data and the importance of each variable; then determining an optimal parameter set according to correlation coefficients among different parameters and importance of each variable; and finally, according to the optimal parameter set and a preset prediction model, taking minimum ground surface settlement, maximum tunneling speed and minimum tunneling specific energy as optimization objectives to obtain optimized tunneling parameters for tunneling. On the basis of data optimization, intelligent optimization of the tunneling parameters can be realized by constructing a deep neural network model and deeply excavating a complex mapping relation between the tunneling parameters in the multi-source data and the ground surface settlement, so that the ground surface settlement is effectively controlled, and the tunnel construction safety and the surrounding environment stability are guaranteed.

Description

Technical Field

[0001] This invention belongs to the field of tunnel excavation construction technology, and in particular relates to a method and system for optimizing tunneling parameters of a tunnel boring machine based on machine learning algorithms. Background Technology

[0002] During tunnel excavation, surface settlement is one of the key factors affecting construction safety and the stability of the surrounding environment. Inappropriate excavation parameters, such as cutterhead speed, advance speed, grouting pressure, and soil chamber pressure, can lead to excessive soil disturbance, resulting in excessive surface settlement and potentially causing serious consequences such as tilting of surrounding buildings and rupture of underground pipelines.

[0003] With the development of artificial intelligence technology, some machine learning-based methods have begun to be applied to land subsidence prediction and tunneling parameter optimization. However, most existing methods only consider a single or a few influencing factors and fail to fully explore the complex correlations between multi-source data, resulting in limited generalization ability and optimization effect of the models. Summary of the Invention

[0004] To address the aforementioned problems, this invention proposes a method and system for optimizing tunneling parameters of a tunnel boring machine based on machine learning algorithms. By constructing a deep neural network model, this invention deeply mines the complex mapping relationship between tunneling parameters and surface settlement in multi-source data, enabling intelligent optimization of tunneling parameters, thereby effectively controlling surface settlement and ensuring tunnel construction safety and the stability of the surrounding environment.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: In a first aspect, the present invention provides a method for optimizing tunneling parameters of a tunnel boring machine based on machine learning algorithms, comprising: Acquire current shield tunneling data, synchronous grouting monitoring data, and surface settlement data; Determine the correlation coefficients and importance of each variable among different parameters in the current shield tunneling data, synchronous grouting monitoring data, and surface settlement data; The optimal parameter set is determined based on the correlation coefficients between different parameters and the importance of each variable; Based on the optimal parameter set and the preset prediction model, the optimized tunneling parameters for tunnel excavation are obtained with the optimization objectives of minimizing surface subsidence, maximizing tunneling speed, and minimizing tunneling specific energy. The prediction model is a hybrid deep neural network model including a long short-term memory network and a Transformer, and the prediction model explores the mapping relationship between tunneling parameters and surface subsidence.

[0006] Furthermore, the shield tunneling data includes cutterhead rotation speed, cutterhead torque, penetration depth, screw conveyor rotation speed, propulsion speed, and total propulsion force; the synchronous grouting monitoring data includes wave velocity, dominant frequency, average deviation, and thickness; and the surface settlement data is the surface settlement caused by shield tunnel construction estimated using the Peck formula.

[0007] Furthermore, the surface subsidence is relative to the distance from the tunnel centerline. Surface subsidence at the location : ; in, This represents the maximum surface subsidence above the tunnel centerline. This is the settling trough width coefficient.

[0008] Furthermore, the Pearson correlation coefficient is used to measure the linear correlation between the parameters. for: ; in, It is a variable and variables The covariance between the two; and These are the averages of the two variables; and These represent the deviations between the two variables.

[0009] Furthermore, when the correlation coefficient When the correlation coefficient is high, it indicates a positive correlation between the two variables; when the correlation coefficient is low, it indicates a positive correlation between the two variables. When the time is right, it indicates a negative correlation between the two variables; when the time is right... When the value is between 0 and the first preset value, it indicates that the two parameters are unrelated; when... When the value is between the first and second preset values, it indicates that there is a low correlation between the two parameters; when When the value is between the first and third preset values, it indicates that the two parameters have a moderate correlation; when... When the value is between the first preset value and 1, it indicates that there is a significant correlation between the two parameters; the first preset value is less than the second preset value, and the second preset value is less than the third preset value.

[0010] Furthermore, the random forest algorithm is used to evaluate the impact of changes in feature variables on the Gini index, thereby measuring the importance of the variables: the random forest model is used to predict OOB (Out of Birth), resulting in... The mean squared residuals of the out-of-bag data are used to generate the regression decision tree. Feature parameters are randomly selected for decision tree splitting, and the parameters are randomly permuted in multiple out-of-bag data samples to form a new OOB test set. The established random forest regression model is used to predict the new test set again to obtain new OOB residuals. The corresponding rows of the error matrix are subtracted from the out-of-bag data, the average of the subtraction results is taken, and then divided by the standard error to obtain the importance score of the feature variable.

[0011] Furthermore, the prediction model mines the mapping relationship between tunneling parameters and surface subsidence, including: a long short-term memory network is used to extract local features of the data, process tunneling data and surface subsidence data with time series characteristics, and dynamically correlate the tunneling parameters and subsidence over time; the Transformer mines the global dependencies of the data through a self-attention mechanism.

[0012] Secondly, the present invention also provides a shield tunneling parameter optimization system based on machine learning algorithms, comprising: The data acquisition module is configured to acquire current shield tunneling data, synchronous grouting monitoring data, and surface settlement data. The data processing module is configured to: determine the correlation coefficients and importance of each variable among different parameters in the current shield tunneling data, synchronous grouting monitoring data, and surface settlement data; The data optimization module is configured to determine the optimal parameter set based on the correlation coefficients between different parameters and the importance of each variable. The optimization module is configured to: obtain optimized tunneling parameters for tunnel excavation based on the optimal parameter set and a preset prediction model, with the optimization objectives of minimizing surface subsidence, maximizing tunneling speed, and minimizing tunneling specific energy; wherein, the prediction model is a hybrid deep neural network model including a long short-term memory network and a Transformer, and the prediction model explores the mapping relationship between tunneling parameters and surface subsidence.

[0013] Thirdly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the shield tunneling parameter optimization method based on machine learning algorithms described in the first aspect.

[0014] Fourthly, the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the program to implement the steps of the shield tunneling parameter optimization method based on machine learning algorithm described in the first aspect.

[0015] Fifthly, the present invention also provides a computer program product, the computer program product comprising a computer program, which, when executed by a processor, implements the steps of the shield tunneling parameter optimization method based on machine learning algorithm described in the first aspect.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention first determines the correlation coefficients and importance of each variable among different parameters in current shield tunneling data, synchronous grouting monitoring data, and surface settlement data. Then, based on the correlation coefficients and importance of each variable, an optimal parameter set is determined. Finally, based on the optimal parameter set and a pre-defined prediction model, optimized tunneling parameters for tunnel excavation are obtained with the optimization objectives of minimizing surface settlement, maximizing tunneling speed, and minimizing tunneling specific energy. The prediction model is a hybrid deep neural network model including a Long Short-Term Memory network and a Transformer, which uncovers the mapping relationship between tunneling parameters and surface settlement. Based on data optimization, by constructing a deep neural network model, the complex mapping relationship between tunneling parameters and surface settlement in multi-source data is deeply explored, enabling intelligent optimization of tunneling parameters, thereby effectively controlling surface settlement and ensuring tunnel construction safety and the stability of the surrounding environment. Attached Figure Description

[0017] The accompanying drawings, which form part of this embodiment, are used to provide a further understanding of this embodiment. The illustrative embodiments and their descriptions are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0018] Figure 1 This is the interface for exporting synchronous grouting monitoring data in Embodiment 1 of the present invention; Figure 2 This is a flowchart illustrating the optimal parameter set generation process in Embodiment 1 of the present invention. Figure 3 This is the LSTM-Transformer model structure of Embodiment 1 of the present invention; Figure 4 This illustrates the effect of hyperparameters on the loss value in Embodiment 1 of the present invention. Figure 5 This is a flowchart of the construction parameter optimization process in Embodiment 1 of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0020] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0021] Example 1: During tunnel excavation, surface settlement is one of the key factors affecting construction safety and the stability of the surrounding environment. Inappropriate excavation parameters, such as cutterhead speed, advance speed, grouting pressure, and soil chamber pressure, can lead to excessive soil disturbance, resulting in excessive surface settlement and potentially causing serious consequences such as tilting of surrounding buildings and rupture of underground pipelines.

[0022] Currently, surface settlement control mainly relies on traditional experience-based methods and simple monitoring and feedback mechanisms. Construction workers set initial tunneling parameters based on their own experience and limited project types, and then manually adjust them during construction based on surface settlement monitoring data. However, this approach has many drawbacks: on the one hand, experience-based judgment lacks scientific rigor and universality, making it difficult to adapt to complex and changing geological conditions and construction scenarios; on the other hand, the timeliness and accuracy of manual adjustments are insufficient, often intervening only after significant anomalies in surface settlement have occurred, resulting in poor control effectiveness.

[0023] As described in the background section, some studies have attempted to use mathematical models to analyze the relationship between tunneling parameters and surface settlement. However, due to the complex mechanisms involved in tunnel construction, including soil mechanics and fluid mechanics, and the high uncertainty of geological conditions, traditional mathematical models struggle to accurately describe actual engineering problems, resulting in significant discrepancies between predicted and actual outcomes. With the development of artificial intelligence, some machine learning-based methods have begun to be applied to surface settlement prediction and tunneling parameter optimization. However, most existing methods only consider a single or a few influencing factors, failing to fully explore the complex correlations between multi-source data, thus limiting the model's generalization ability and optimization effectiveness.

[0024] To address at least one of the aforementioned problems, this embodiment provides a method for optimizing tunneling parameters of a tunnel boring machine (TBM) based on machine learning algorithms. This method comprehensively utilizes multi-source data and precisely optimizes tunneling parameters to effectively control surface settlement. By constructing a deep neural network model, this method deeply mines the complex mapping relationship between tunneling parameters and surface settlement in multi-source data, enabling intelligent optimization of tunneling parameters. This effectively controls surface settlement, ensuring tunnel construction safety and the stability of the surrounding environment. The method includes: S1. Data Preparation and Preprocessing: S1.1 Data Acquisition: Optionally, this embodiment relies on data collected from actual engineering projects, such as tunnel boring machines or ultra-large diameter tunnel boring machines. The data mainly includes tunnel boring data, synchronous grouting monitoring data, and surface settlement data.

[0025] like Figure 1As shown, the tunneling data is exported from a CSV file via a link to the tunnel's intelligent control system. It mainly includes key tunneling parameters such as cutterhead rotation speed, cutterhead torque, penetration depth, screw conveyor speed, propulsion speed, and total propulsion force. In this embodiment, the average value of the operating parameters for each ring is calculated to explore the nonlinear relationship between ground grouting and control parameters.

[0026] The synchronous grouting monitoring data is collected on-site by elastic wave detection equipment and system. It mainly uses a single-component high-precision sensor (signal acquisition device) and a high-bandwidth longitudinal wave signal acquisition control motherboard to acquire longitudinal wave signal data. The data is exported as a TXT file and mainly includes key monitoring data such as wave velocity, main frequency, average deviation and thickness.

[0027] In this embodiment, the surface settlement data is calculated using a combination of empirical formulas and field data. High-precision total stations and levels are used at the construction site to periodically measure the elevation changes of surface settlement points, obtaining settlement data for different time periods. Earth pressure cells and other sensors are used to collect soil stress and strain data during construction. Regarding empirical formulas, the Peck formula is used to estimate surface settlement caused by shield tunnel construction, and its specific expression is as follows: (1) in, Distance from the tunnel centerline The amount of surface subsidence at the location; This represents the maximum surface subsidence above the tunnel centerline. This is the settling trough width coefficient.

[0028] Optionally, parameters can be inverted using field data to retrieve settlement data from multiple settlement points obtained through field monitoring. and the corresponding distance Substituting into the formula, the maximum surface subsidence is solved by inversion using optimization algorithms such as the least squares method. and settling trough width coefficient The determined empirical formula parameters and the compiled field data are substituted into the empirical formula to calculate the surface settlement at different locations and construction stages. According to the construction progress, the surface settlement increment caused by each construction step is calculated in stages, and the total surface settlement is obtained by accumulating the results. For example, in shield tunnel construction, after each ring is advanced, the surface settlement distribution caused by the construction of that ring is calculated using the Peck formula based on the construction parameters and geological conditions of that ring. Then, the settlement calculation results of each ring are superimposed to obtain the surface settlement situation under the current construction progress.

[0029] S1.2 Redundant parameter filtering and feature selection: During the operation of a tunnel boring machine (TBM) system, there are complex coupling relationships between different construction parameters (data). Some parameters may be highly correlated, and their data have similar information and effects. The simultaneous existence of such a set of parameters may lead to a decrease in the accuracy of the model, so further analysis is necessary.

[0030] like Figure 2 As shown, to further analyze the relationship between different parameters, the Pearson correlation coefficient (PCC) is used to measure the linear correlation between the parameters. The formula for calculating the Pearson correlation coefficient is expressed as: (2) in, It is a variable and variables The covariance between the two; and These are the averages of the two variables; and These represent the deviations between the two variables. Correlation coefficient. The value is between [-1, 1], according to the correlation coefficient The magnitude and sign of the correlation coefficient can be used to analyze and judge the correlation between parameters. When the correlation coefficient is high, it indicates a positive correlation between the two variables; when the correlation coefficient is low, it indicates a positive correlation between the two variables. When the two variables are negatively correlated, it indicates a negative correlation between them. When the value is between [0, 0.09], it indicates that the two parameters are unrelated; when... When the value is between [0.09, 0.3], it indicates a low correlation between the two parameters; when... When the value is between [0.3, 0.5], it indicates a moderate correlation between the two parameters; when... When the value is between [0.5, 1], it indicates that there is a significant correlation between the two parameters.

[0031] To retain as much useful information as possible, the discrimination interval for redundant parameters is set to [0.8, 1]. When the absolute value of the correlation coefficient between two parameters is greater than 0.8, the two parameters are highly correlated and can be considered to have a high degree of similarity in their data information. Only one parameter needs to be retained.

[0032] The Random Forest (RF) algorithm is used to evaluate the impact of changes in feature variables on the Gini index, thereby measuring the importance of variables and screening out basic parameters with a high impact on land subsidence. The RF algorithm primarily evaluates the importance of parameters based on the Oriented Bounding Box (OBB) error. After randomly permuting the parameters, it measures the reduction in the model's mean square residual or the reduction in model accuracy to determine the importance of feature parameters. The calculation steps are as follows: Building a regression decision tree: Using a random forest model to predict OOB (Out-of-Body) errors, we can obtain... Mean squared residuals of out-of-bag data; Generating the error matrix: When generating the regression decision tree, feature parameters are randomly selected for decision tree splitting, and the parameters are... Random permutations are performed on the out-of-bag (OOB) data samples to form a new OOB test set. The established random forest regression model is then used to predict the new test set again to obtain the new OOB residuals. Importance scoring: using out-of-bag data , , ..., Subtracting the corresponding row from the error matrix, averaging the results, and then dividing by the standard error yields the average decrease in the mean squared residual of the variable, which is the importance score of the feature variable.

[0033] The Gini index is a type of correlation, so this sentence can be rewritten as follows: To further filter redundant parameters, random forest is used to perform importance analysis and feature selection on the input parameters to determine the optimal parameter set. In Python, random forest is used to evaluate the importance of the input parameters, with the number of leaf nodes set to 60, the tree depth set to 12, and land subsidence as the output variable. The importance scores of the remaining parameters are calculated and ranked.

[0034] After Pearson correlation coefficient analysis, parameters with high correlation between pairs of parameters are removed. Then, random forest feature selection is used to select the parameters that are highly correlated with the output variable from the remaining parameters, which is the optimal parameter input set.

[0035] S2. Deep Neural Network Model Construction: S2.1 Model Design: like Figure 3As shown, a hybrid deep neural network model consisting of a Long Short-Term Memory (LSTM) network and a Transformer is constructed. The LSTM network is used to extract local features of the data, process tunneling data and surface subsidence data with time-series characteristics, and mine the dynamic correlation between parameters and subsidence over time. The Transformer uses a self-attention mechanism to mine global dependencies in the data, comprehensively considering the complex relationships between different construction stages and various types of data. The Transformer model is improved by using fully connected layers and LSTM modules as the decoder.

[0036] S2.2 Parameter Initialization: like Figure 4 As shown, the key hyperparameters of the LSTM-Transformer model's internal structure include the decoder optimizer learning rate (lr2), the encoder optimizer learning rate (lr1), the dropout probability parameter, the number of encoder layers (numlayers), and the hidden layer parameter (hidden_size). This embodiment uses the optimization tool Optuna and Bayesian optimization to optimize these hyperparameters, ultimately finding values ​​that minimize the model's loss. Bayesian optimization is a probabilistic modeling and mathematical optimization technique for global optimization. Its goal is to find the globally optimal solution to an unknown objective function within a finite number of iterations. Applying this idea to hyperparameter tuning results in the Bayesian optimizer, which integrates mathematical modeling, probability theory, and optimization algorithms, becoming a powerful tool, particularly suitable for machine learning hyperparameter tuning problems that require finding optimal solutions in the parameter space. Optuna (Optimization Tuning) is an open-source Python library for hyperparameter optimization. Its main goal is to provide a flexible, lightweight, and easy-to-use tool to help machine learning engineers and researchers effectively search for optimal combinations of hyperparameters when training models. Using the Optuna tool to implement Bayesian optimization mainly includes: 1) defining the hyperparameter search space; 2) unpacking the hyperparameters and creating the model; 3) defining the objective function (core); and 4) starting to optimize the key hyperparameters.

[0037] S2.3 Model Training and Optimization: The preprocessed dataset was divided into training, validation, and test sets in a 7:2:1 ratio. The training set was used for learning and tuning model parameters; the validation set was used to evaluate model performance during training by monitoring the loss function and evaluation metrics (such as mean squared error MSE and mean absolute error MAE) on the validation set, and adjusting the model's hyperparameters (such as learning rate, number of network layers, number of neurons, etc.) to prevent overfitting; the test set was used to objectively evaluate the finally trained model and obtain its true performance on unknown data.

[0038] S2.4, Layered Training: First, the weight optimization layer is locked, and the neural network composed of LSTM and Transformer is trained using the training set. An intelligent optimization algorithm automatically adjusts the learning rate to accelerate model convergence in the early stages of training and avoid excessive oscillations in the later stages. L2 regularization is used to constrain the weight parameters of the neural network to prevent overfitting. Through continuous iterative training, the neural network parameters are optimized based on metrics such as mean squared error (MSE) and mean absolute error (MAE) until the model's performance on the validation set reaches stability or optimality. Next, the weight optimization layer is unlocked, and the neural network layer is locked, and the weight optimization layer is trained using the training set. Based on real-time construction data and surface settlement monitoring feedback, the weight values ​​in the weight optimization matrix are dynamically adjusted through an adaptive weight adjustment mechanism. For example, when surface settlement exceeds the warning threshold, the adjustment of the weights corresponding to key tunneling parameters affecting settlement (such as grouting pressure and advance speed) is increased to further optimize the model's prediction results for tunneling parameters.

[0039] like Figure 5 As shown, the trained deep neural network model is treated as an agent, and the tunnel excavation process is treated as a sequential decision-making problem. The agent selects a set of excavation parameters as actions based on the current construction state. After executing these actions, the optimization objectives are based on evaluation indicators such as minimizing surface settlement, maximizing excavation speed, and minimizing excavation specific energy. The Proximal Policy Optimization (PPO) algorithm is employed. Through continuous trial and error, the agent learns how to select the optimal excavation parameters under different construction states to maximize long-term cumulative rewards, achieving dynamic optimization of the excavation parameters and ensuring effective control of surface settlement.

[0040] S2.5, Continuous Optimization: As construction progresses, newly collected construction data and surface settlement monitoring data are continuously added to the training dataset. The model is periodically retrained and optimized, and model parameters are adjusted based on new data to ensure that the model continuously adapts to changing construction conditions, geological conditions, and surface settlement requirements. This guarantees the accuracy and effectiveness of tunneling parameter optimization and enables long-term stable control of surface settlement.

[0041] Example 2: This embodiment provides a shield tunneling parameter optimization system based on machine learning algorithms, including: The data acquisition module is configured to acquire current shield tunneling data, synchronous grouting monitoring data, and surface settlement data. The data processing module is configured to: determine the correlation coefficients and importance of each variable among different parameters in the current shield tunneling data, synchronous grouting monitoring data, and surface settlement data; The data optimization module is configured to determine the optimal parameter set based on the correlation coefficients between different parameters and the importance of each variable. The optimization module is configured to: obtain optimized tunneling parameters for tunnel excavation based on the optimal parameter set and a preset prediction model, with the optimization objectives of minimizing surface subsidence, maximizing tunneling speed, and minimizing tunneling specific energy; wherein, the prediction model is a hybrid deep neural network model including a long short-term memory network and a Transformer, and the prediction model explores the mapping relationship between tunneling parameters and surface subsidence.

[0042] The working method of the system is the same as the shield tunneling parameter optimization method based on machine learning algorithm in Example 1, and will not be repeated here.

[0043] Example 3: This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the shield tunneling parameter optimization method based on machine learning algorithm described in Embodiment 1.

[0044] Example 4: This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor. When the processor executes the program, it implements the steps of the shield tunneling parameter optimization method based on machine learning algorithm described in Embodiment 1.

[0045] Example 5: This embodiment provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the shield tunneling parameter optimization method based on machine learning algorithm described in Embodiment 1.

[0046] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A method for optimizing tunneling parameters of a shield tunneling machine based on a machine learning algorithm, characterized in that, The method comprises: acquiring current shield tunneling data, synchronous grouting monitoring data and ground settlement data; determining the correlation coefficients between different parameters and the importance of each variable in the current shield tunneling data, the synchronous grouting monitoring data and the ground settlement data; determining an optimal parameter set according to the correlation coefficients between different parameters and the importance of each variable; obtaining optimized tunneling parameters for tunneling according to the optimal parameter set and a preset prediction model, with the minimum ground settlement, the maximum tunneling speed and the minimum tunneling specific energy as the optimization objectives; wherein the prediction model is a hybrid deep neural network model comprising a long short-term memory network and a Transformer, and the prediction model mines the mapping relationship between the tunneling parameters and the ground settlement.

2. The shield tunneling parameter optimization method based on a machine learning algorithm according to claim 1, wherein the shield tunneling data comprises cutterhead rotation speed, cutterhead torque, penetration, screw rotation speed, pushing speed and total pushing force; the synchronous grouting monitoring data comprises wave speed, main frequency, average deviation and thickness; and the ground settlement data is the ground settlement caused by shield tunneling construction estimated by a Peck formula.

3. The machine learning algorithm based tunneling parameter optimization method for a tunneling machine according to claim 2, wherein, The ground settlement is an amount of ground settlement at a distance from a tunnel center line :​ ； wherein, is the maximum ground settlement above the tunnel centerline; is the settlement trough width factor.

4. The machine learning algorithm based tunneling parameter optimization method for a tunneling machine according to claim 1, wherein, The linear correlation between parameters is measured using the Pearson correlation coefficient, which is denoted by and is given by: ； where is a variable and a variable the covariance between the two variables; and are the mean values of the two variables, respectively; and are the deviations of the two variables, respectively.

5. The machine learning algorithm based tunneling parameter optimization method for a tunneling machine according to claim 4, wherein, When the correlation coefficient When the correlation coefficient is high, it indicates a positive correlation between the two variables; when the correlation coefficient is low, it indicates a positive correlation between the two variables. When the time is right, it indicates a negative correlation between the two variables; when the time is right... When the value is between 0 and the first preset value, it indicates that the two parameters are unrelated; when... When the value is between the first and second preset values, it indicates that there is a low correlation between the two parameters; when When the value is between the first and third preset values, it indicates that the two parameters have a moderate correlation; when... When the value is between the first preset value and 1, it indicates that there is a significant correlation between the two parameters; the first preset value is less than the second preset value, and the second preset value is less than the third preset value.

6. The machine learning algorithm based tunneling parameter optimization method for a tunneling machine of claim 1, wherein, The importance of the variable is measured by evaluating the impact of the change of the characteristic variable on the Gini index using the random forest algorithm: the random forest model is used to predict the OOB to obtain the mean square residual of the out-of-bag data; when generating the regression decision tree, the characteristic parameters are randomly selected for decision tree splitting, the parameters are randomly replaced in multiple out-of-bag data samples to form a new OOB test set, the established random forest regression model is used to predict the new test set again to obtain a new OOB residual; the importance score of the characteristic variable is obtained by subtracting the corresponding row of the error matrix from the out-of-bag data, averaging the result after subtraction, and then dividing by the standard error.

7. The machine learning algorithm based tunneling parameter optimization method for a tunneling machine of claim 1, wherein, mining the mapping relationship between the tunneling parameters and the ground settlement by the prediction model comprises: the long short-term memory network is used to extract local features of data, process tunneling data and ground settlement data with time series characteristics, and mine dynamic correlations of parameters and settlement with time; and the Transformer mines global dependency relationships of data through a self-attention mechanism.

8. A machine learning algorithm based tunneling parameter optimization system for a tunneling machine, characterized in that, The method comprises: a data acquisition module configured to acquire current shield tunneling data, synchronous grouting monitoring data and ground settlement data; a data processing module configured to determine the correlation coefficients between different parameters and the importance of each variable in the current shield tunneling data, the synchronous grouting monitoring data and the ground settlement data; a data optimization module configured to determine an optimal parameter set according to the correlation coefficients between different parameters and the importance of each variable; an optimization module configured to obtain optimized tunneling parameters for tunneling according to the optimal parameter set and a preset prediction model, with the minimum ground settlement, the maximum tunneling speed and the minimum tunneling specific energy as the optimization objectives; wherein the prediction model is a hybrid deep neural network model comprising a long short-term memory network and a Transformer, and the prediction model mines the mapping relationship between the tunneling parameters and the ground settlement.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and capable of running on the processor, characterized in that, The processor implements the steps of the shield tunneling parameter optimization method based on a machine learning algorithm according to any one of claims 1-7 when executing the program.

10. A computer program product, characterised in that, The computer program product comprises a computer program, and the computer program implements the steps of the shield tunneling parameter optimization method based on a machine learning algorithm according to any one of claims 1-7 when executed by a processor.

Images