A method for optimizing shield tunneling parameters based on the PSO algorithm

Abstract

This invention proposes a shield tunneling parameter optimization method based on the PSO algorithm, comprising: acquiring raw tunneling data; data preprocessing: removing parameters from non-tunneling and shutdown / restart states of the shield machine, and removing data outliers; geological condition classification; establishing a prediction model: constructing a BP neural network model to build corresponding mappings, establishing the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque, and determining the fitness function; prediction stage: inputting the BP neural network model into the PSO particle swarm optimization algorithm, verifying the optimal solutions for cutterhead rotation speed and total thrust obtained through training, and outputting the shield active operation parameters of cutterhead rotation speed and total thrust with the objectives of minimizing shield energy consumption and shortening the construction period. This invention can optimize shield tunneling parameters in different strata, achieving the goals of reducing shield energy consumption and increasing tunneling speed, thereby saving construction costs and shortening the construction period.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular to a method for optimizing shield tunneling parameters based on the PSO algorithm. Background Technology

[0002] With the rapid development of underground transportation engineering in my country, large-diameter slurry shield tunneling has become a common construction method. In response to the national goals of "achieving carbon peaking and carbon neutrality as soon as possible," reducing project costs and enabling tunnels to be put into use early, minimizing energy consumption during shield tunneling and shortening the construction period have become the two most pressing concerns for all relevant parties. During shield tunneling, the setting of tunneling parameters significantly impacts the tunneling effect. Among these, the cutterhead speed and total thrust of the tunnel boring machine (TBM), manually controlled by the operator, are the main research subjects for predicting tunneling parameters.

[0003] Current prediction technologies include BP neural network prediction models that can output the total thrust of the tunnel boring machine (TBM), random forest prediction models, GA-BPNN algorithm prediction models, and prediction models that use convolutional neural networks and long short-term memory networks to predict the cutterhead rotation speed. However, most existing research and inventions are only applied to single tunnels, and their prediction results may not be applicable to other tunnels. There are few studies based on sample sets composed of data from multiple tunnels, and the general distribution patterns and ranges of TBM cutterhead rotation speed and total thrust are still unclear. Furthermore, there is a lack of research on providing corresponding operational suggestions for TBM operators based on TBM energy consumption and construction period, making it difficult to achieve these goals. Summary of the Invention

[0004] The purpose of this invention is to propose a shield tunneling parameter optimization method based on the PSO algorithm to solve the aforementioned problems existing in the prior art.

[0005] Technical solution: A method for optimizing shield tunneling parameters based on the PSO algorithm, comprising:

[0006] Step S1: Acquisition of raw tunneling data: Collect construction data of no less than 8 shield tunnels, which includes various geological conditions as raw tunneling data;

[0007] Step S2, Data Preprocessing: Remove parameters from the original tunneling data obtained in Step S1 that are in the non-tunneling state and the shutdown and restart state of the tunneling machine. Select the first 10% of the tunneling state of each ring as the rising segment and the last 90% of the data as the stable segment. Remove the parameters in the rising segment. Use the Mahalanobis distance method to process the data to remove outliers. Normalize the training set data to reduce the influence of differences in the units and scales of different parameters. Then perform statistical analysis on the tunneling parameters.

[0008] Step S3, Geological Condition Classification: Establish corresponding grading standards according to different soil types. For composite soil layers, adopt corresponding scores as parameter values ​​to obtain the range of cutterhead rotation speed and total thrust of the tunnel boring machine in different soils.

[0009] Step S4: Establish a prediction model: Establish the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque. Use total thrust and cutterhead rotation speed as initial parameters, input the initial parameters as training set into the artificial neural network model, train the neural network, and obtain the fitness formulas for optimizing shield tunneling energy consumption and construction period.

[0010] Step S5, Prediction Stage: Using the training set from Step S4, the PSO particle swarm optimization algorithm is loaded into the BP artificial neural network model, and predictions are made on the optimal hyperparameterized artificial neural network model; with the minimum shield energy consumption and the shortest construction period as decision objectives, the predicted cutterhead rotation speed and total thrust are output respectively, obtaining the shield energy consumption and tunneling speed under two cutterhead rotation speeds and total thrusts; after comparison, a set of data that meets the engineering requirements is selected.

[0011] Its specific process includes:

[0012] Step S51: Establish an optimal hyperparameterized artificial neural network model for predicting cutterhead rotation speed and total thrust. The specific process is as follows:

[0013] A BP artificial neural network model was established, and K-fold cross-validation was used to predict the shield cutterhead rotation speed and total shield thrust for the next 2 to 3 rings. Then, the corresponding mapping formula was constructed to obtain the relationship between the shield's active operating parameters (total thrust, cutterhead rotation speed) and passive parameters (cutterhead rotation speed, tunneling speed, and cutterhead torque).

[0014] f v (x n ,x F ) = v;

[0015] f T (x n ,x F ) = T;

[0016] In the formula: x n x is the rotational speed of the cutter head. F v is the total thrust, v is the tunneling speed, and T is the cutterhead torque.

[0017] Step S52: Establish the PSO optimization model. PSO particle swarm optimization is used to make decisions and optimize the shield tunneling with the goal of minimizing energy consumption and minimizing construction period, so as to obtain the optimal cutterhead speed and total thrust.

[0018] Step S521: Initialize the position and velocity of the particle swarm, setting the velocity to 0 or generating it randomly within a certain range;

[0019] Step S522: Load the relationship model between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque, and input it into the PSO optimization model;

[0020] Step S523: Calculate the fitness value of each particle using the derived fitness formula to measure the quality of the particle's position, update the historical optimal solution and global optimal solution of each particle, and update its velocity and position based on the historical optimal solution and global optimal solution.

[0021] Step S524: If the stopping condition is met, output the optimal solution; otherwise, continue to update the historical optimal solution and the global optimal solution for each particle until the maximum number of iterations is reached.

[0022] Step S53: Establish a BP neural network. The specific process includes:

[0023] Step S531: Construct a BP neural network model structure using an input layer, an intermediate layer, a Dropout layer, and an output layer, with the optimizer set to Adam; the activation function of the input layer is the sigmoid function.

[0024] Step S532: Set the number of iterations to n, and use early stopping to process the training process. When the error of the model on the validation set has not improved for 10 consecutive iterations, the training process is interrupted.

[0025] Step S533: After the minimum mean square error of the validation set is determined, the hyperparameters of the BP neural network model can be determined at this time.

[0026] According to a further improvement of the present invention, when constructing the BP neural network model, the sample set is divided into a training set and a test set in a ratio of 8:2. One-tenth of the data in the training set is selected as the validation set to determine the pairwise relationship between the data in the same group of parameters and to determine the prediction model. The test set is used to evaluate the error between the prediction model and the actual values ​​of the cutterhead speed and cutterhead torque.

[0027] According to a further improvement of the present invention, step S2 specifically includes:

[0028] Step S21: Remove data from the original tunneling data obtained in step S1, specifically data on the tunnel boring machine in non-tunneling states: Remove data from when the tunnel boring machine is stopped and during segment assembly. Set a discriminant function: f i =F i ·v i ·T i ·n i ;

[0029] In the formula, F i v represents the total thrust at the i-th time step, in kN. i T represents the tunneling speed at the i-th time step, in mm / min. i n is the cutterhead torque at the i-th time step, in MN·m; i Let be the rotational speed of the tool turret at the i-th time step, in rpm;

[0030] When any one of the total thrust, tunneling speed, cutterhead torque, and cutterhead rotation speed of the tunnel boring machine is 0, the discriminant function is 0, which means that the tunnel boring machine is in a non-tunneling state and the non-tunneling state data is removed.

[0031] Step S22, taking the steady-state phase, includes:

[0032] Step S221: Select data with a tunneling cycle duration of more than 30 minutes and reorder them;

[0033] Step S222: Arrange the tunneling status data of each ring by time, take the first 10% of the data as the rising segment, and the last 90% of the data as the stable segment;

[0034] Step S23: Remove outlier data: The Mahalanobis distance method is used to process outlier data. The Mahalanobis distance method is a data similarity measurement method that calculates the distance between data points based on the covariance matrix of the data. The specific calculation method is as follows:

[0035]

[0036] In the formula, x i Let N be a single data point in the dataset, μ be the mean of the sample, and S be the covariance matrix of the data. This method is compared with other processing methods such as box plots and the 3σ method.

[0037] Step 24: Smoothing: Use the moving average method to filter and denoise the digital signal. Select an appropriate window size and move it along the time axis of the signal. Take the average value of all samples in the window as the value of the center sample of the window. Set the window size to 5 and smooth all shield tunneling parameters. This removes most of the noise while retaining the overall trend of change in the entire shield tunneling process.

[0038] Step S25, Normalization: Before building the model, the training set data is processed using a normalization method to reduce the influence of differences in the dimensions and scales of different parameters, thereby improving the fitting accuracy and speed. The specific method is as follows:

[0039]

[0040] In the formula, x norm Here are the normalized data, and x is the original data. max and x min These are the maximum and minimum values ​​in the original data, respectively.

[0041] According to a further improvement of the present invention, step S3 specifically includes:

[0042] Step S31: Assign a score to the soil layer based on its type, plasticity index, and the proportion of different particle sizes in the soil. The score ranges from 0 to 9, and a grading standard table is obtained.

[0043] Step S32: For composite soil layers, select the weighted average of the corresponding scores as the score value. The calculation formula is as follows:

[0044]

[0045] In the formula, n is the total number of soils in each ring, and S i The percentage (%) of different soil types in each ring of soil, d i The score for each ring of soil;

[0046] Step S33: Classify different types of soil according to the data in the classification standard table in step S31, and obtain the range of cutterhead rotation speed and total thrust of the tunnel boring machine in different types of soil.

[0047] According to a further improvement of the present invention, step S4 specifically includes:

[0048] Step S41: Select the main parameters for shield tunneling construction and excavation, calculate the Pearson correlation coefficient, Spearman correlation coefficient, and feature importance based on random forest for each pair of shield tunneling parameters, and draw the corresponding heatmaps; for feature selection based on Pearson and Spearman correlation coefficients, identify parameters that have a certain correlation (|r|≥0.2) with tunneling speed, cutterhead rotation speed, total thrust, and cutterhead torque; for feature selection based on random forest feature importance, select shield tunneling parameters whose importance is accumulated from high to low up to 0.95; finally, select the intersection of the features selected by the three methods.

[0049] Step S42: Use a BP neural network model to construct the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque through forward propagation, backward propagation and gradient descent;

[0050] Step S43: Optimize the hyperparameters of the BP neural network model using a grid search method, specifying the possible range of hyperparameter values ​​and the possible values ​​of each parameter; traverse all possible combinations of hyperparameters; and use K-fold cross-validation.

[0051] The process of K-fold cross-validation includes:

[0052] The data in the training set is divided into K groups on average. One group of data is randomly selected from these groups as the validation set, and the remaining K-1 groups are used for training.

[0053] Repeat the above process K times, with each time the selected validation set is different. Take the MSE of each validation set and use the average of the K times as the final MSE.

[0054] According to a further improvement of the present invention, the shield tunneling parameter prediction process in step S5 is divided into two parts: shield tunneling energy consumption minimization decision optimization and construction period minimization decision optimization.

[0055] The specific optimization decision for minimizing shield tunneling energy consumption is as follows: Shield tunneling energy consumption refers to the work done per unit volume of soil excavated, and its calculation formula is:

[0056]

[0057] Among them, W T Work done by the cutter head torque, n is the cutter head rotation speed, T is the cutter head torque, t is time, W F The work done for the total thrust, where F is the total thrust and v is the tunneling speed;

[0058] Substituting the above relationship into this formula yields the fitness function for shield tunneling energy consumption:

[0059]

[0060] The fitness function is applied to the algorithm flow to determine the historical optimal solution p. best and the global optimal solution g best Set the relevant parameters for the particle swarm, including inertia weight w = 0.9, acceleration constants c1 = c2 = 0.5, and maximum particle velocity V. max =1, number of particles N=100, number of iterations G max =100, and after optimization calculation, the optimal cutter head speed n and total thrust F are obtained;

[0061] The specific optimization strategy for minimizing the construction period involves: increasing the tunneling speed to improve tunneling efficiency and shorten the construction period; and utilizing the time consumed by the tunnel boring machine (TBM) for tunneling. In the formula, Let v be the excavation distance and v be the tunneling speed; substituting the relationship into the equation yields the fitness function as follows:

[0062]

[0063] In the formula, L is the shield excavation distance, set to 1m; t is the tunneling time; f v (x i,n ,xi,F ) is the mapping formula for the shield tunneling speed of the i-th ring;

[0064] Wherein, the cutter head rotation speed x i,n ∈[n min ,n max Total thrust x i,F ∈[F min ,F max When t takes a smaller value, a better fitness value is obtained. When the global optimal solution is obtained, that is, when the condition of the shortest construction period is met, the optimal parameter combination is obtained.

[0065] When using the PSO algorithm to optimize tunnel boring machine (TBM) operation, various parameters are set, including inertia weight w = 1.5, acceleration constants c1 = c2 = 0.5, and maximum particle velocity V. max =1, number of particles N=200, number of iterations G max =300. After optimization calculation, the optimal cutter head speed n and total thrust F are obtained.

[0066] Beneficial effects: By combining shield tunneling construction experience with Pearson correlation coefficient, data preprocessing is used to remove abnormal data in the ascending and stable phases of shield machine operation. This helps to ensure the rationality, scientificity, and accuracy of the data, reduces data dimensionality, and improves the model's predictive efficiency.

[0067] Adding a Dropout layer to a backpropagation (BP) neural network can help avoid problems such as overfitting, vanishing gradients, and exploding gradients, thereby enhancing the model's generalization ability and improving its accuracy.

[0068] K-fold cross-validation is used to cross-validate the BP neural network model, ensuring that the model has better evaluation index scores for training and validation when the sample data is small.

[0069] The PSO optimization algorithm is used to optimize the BP neural network model, which makes the particles in the PSO optimization model continuously update their own positions, so that the individual optimal solution and the historical optimal solution move towards the target position. It requires fewer parameters, making the model relatively simple and easy to implement, which helps the model converge quickly and avoid getting trapped in local optima.

[0070] The final output parameters are fewer, which makes it easier for the tunnel boring machine operator to adjust the driving parameters more intuitively in order to achieve the goals of low energy consumption and short construction period. Attached Figure Description

[0071] Figure 1 This is the overall flowchart of the present invention.

[0072] Figure 2 This is a schematic diagram of the K-fold cross-validation principle of the present invention.

[0073] Figure 3 This is a structural diagram of the BP neural network model of the present invention.

[0074] Figure 4 This is a heat map of the Pearson correlation coefficient of the shield tunneling parameters of this invention.

[0075] Figure 5 This is a heat map of the Spearman correlation coefficients of the shield tunneling parameters of this invention.

[0076] Figure 6 This is a flowchart of the PSO particle swarm optimization algorithm of the present invention.

[0077] Figure 7 The fitness curve of the PSO algorithm in the application case of this invention. Detailed Implementation

[0078] The technical solution of the present invention will be further described in detail below through embodiments and with reference to the accompanying drawings.

[0079] To make the technical problems of the present invention clearer, embodiments of the present invention will be provided below, along with a more comprehensive description. However, the specific embodiments given herein are only for explaining the present invention and are not intended to limit the scope or application of the invention.

[0080] Example 1:

[0081] See Figure 1 This embodiment provides a method for optimizing key parameters of a tunnel boring machine (TBM) based on the PSO algorithm. More specifically, it is a method for optimizing TBM parameters based on the PSO particle swarm optimization algorithm, aiming to reduce TBM energy consumption and shorten construction period. The method includes the following steps:

[0082] Step S1: Acquisition of raw tunneling data: Collect construction data of no less than 8 shield tunnels, which includes various geological conditions as raw tunneling data;

[0083] Step S2, Data Preprocessing: Remove parameters from the original tunneling data obtained in Step S1 that are in the non-tunneling state and the shutdown and restart state of the tunneling machine. Select the first 10% of the tunneling state of each ring as the rising segment and the last 90% of the data as the stable segment. Remove the parameters in the rising segment. Use the Mahalanobis distance method to process the data to remove outliers. Normalize the training set data to reduce the influence of differences in the units and scales of different parameters. Then perform statistical analysis on the tunneling parameters.

[0084] Step S3, Geological Condition Classification: Establish corresponding grading standards according to different soil types. For composite soil layers, adopt corresponding scores as parameter values ​​to obtain the range of cutterhead rotation speed and total thrust of the tunnel boring machine in different soils.

[0085] Step S4: Establish a prediction model: Establish the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque. Use total thrust and cutterhead rotation speed as initial parameters, input the initial parameters as training set into the artificial neural network model, train the neural network, and obtain the fitness formulas for optimizing shield tunneling energy consumption and construction period.

[0086] Step S5, Prediction Stage: Using the training set from Step S4, the PSO particle swarm optimization algorithm is loaded into the BP artificial neural network model, and predictions are made on the optimal hyperparameterized artificial neural network model; with the minimum shield energy consumption and the shortest construction period as decision objectives, the predicted cutterhead rotation speed and total thrust are output respectively, obtaining the shield energy consumption and tunneling speed under two cutterhead rotation speeds and total thrusts; after comparison, a set of data that meets the engineering requirements is selected.

[0087] Its specific process includes:

[0088] Step S51: Establish an optimal hyperparameterized artificial neural network model for predicting cutterhead rotation speed and total thrust. The specific process is as follows:

[0089] A BP artificial neural network model was established, and K-fold cross-validation was used to predict the shield cutterhead rotation speed and total shield thrust for the next 2 to 3 rings. Then, a corresponding mapping was constructed to obtain the relationship between the shield's active operating parameters (total thrust, cutterhead rotation speed) and passive shield parameters (tunneling speed, cutterhead torque).

[0090] f v (x n ,x F ) = v;

[0091] f T (x n ,x F ) = T;

[0092] In the formula, x n x is the rotational speed of the cutter head; F v is the total thrust, v is the tunneling speed, and T is the cutterhead torque.

[0093] Step S52: Establish the PSO optimization model. PSO particle swarm optimization is used to make decisions and optimize the shield tunneling with the goal of minimizing energy consumption and minimizing construction period, so as to obtain the optimal cutterhead speed and total thrust.

[0094] Step S521: Initialize the position and velocity of the particle swarm, setting the velocity to 0 or generating it randomly within a certain range;

[0095] Step S522: Load the relationship model between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque, and input it into the PSO optimization model;

[0096] Step S523: Calculate the fitness value of each particle using the derived fitness formula to measure the quality of the particle's position, update the historical optimal solution and global optimal solution of each particle, and update its velocity and position based on the historical optimal solution and global optimal solution.

[0097] Step S524: If the stopping condition is met, output the optimal solution; otherwise, continue to update the historical optimal solution and the global optimal solution for each particle until the maximum number of iterations is reached.

[0098] Step S53: Establish a BP neural network. The specific process includes:

[0099] Step S531: Construct a BP neural network model structure using an input layer, an intermediate layer, a Dropout layer, and an output layer, with the optimizer set to Adam; the activation function of the input layer is the sigmoid function.

[0100] Step S532: Set the number of iterations to n, and use early stopping to process the training process. When the error of the model on the validation set has not improved for 10 consecutive iterations, the training process is interrupted.

[0101] Step S533: After the minimum mean square error of the validation set is determined, the hyperparameters of the BP neural network model can be determined at this time.

[0102] According to a further improvement of the present invention, when constructing the BP neural network model, the sample set is divided into a training set and a test set in a ratio of 8:2. One-tenth of the data in the training set is selected as the validation set to determine the pairwise relationship between the data in the same group of parameters and to determine the prediction model. The test set is used to evaluate the error between the prediction model and the actual values ​​of the cutterhead speed and cutterhead torque.

[0103] According to a further improvement of the present invention, step S2 specifically includes:

[0104] Step S21: Remove data from the original tunneling data obtained in step S1, specifically data on the tunnel boring machine in non-tunneling states: Remove data from when the tunnel boring machine is stopped and during segment assembly. Set a discriminant function: f i =F i ·v i ·T i ·n i ;

[0105] In the formula, F i v represents the total thrust at the i-th time step, in kN. i T represents the tunneling speed at the i-th time step, in mm / min. i n is the cutterhead torque at the i-th time step, in MN·m;i Let be the rotational speed of the tool turret at the i-th time step, in rpm;

[0106] When any one of the total thrust, tunneling speed, cutterhead torque, and cutterhead rotation speed of the tunnel boring machine is 0, the discriminant function is 0, which means that the tunnel boring machine is in a non-tunneling state and the non-tunneling state data is removed.

[0107] Step S22, taking the steady-state phase, includes:

[0108] Step S221: Select data with a tunneling cycle duration of more than 30 minutes and reorder them;

[0109] Step S222: Arrange the tunneling status data of each ring by time, take the first 10% of the data as the rising segment, and the last 90% of the data as the stable segment;

[0110] Step S23: Remove outlier data: The Mahalanobis distance method is used to process outlier data. The Mahalanobis distance method is a data similarity measurement method that calculates the distance between data points based on the covariance matrix of the data. The specific calculation method is as follows:

[0111]

[0112] In the formula, x i Let N be a single data point in the dataset, μ be the sample mean, and S be the covariance matrix of the data. Compared with other processing methods such as box plots and the 3σ method, the data distribution after processing is basically consistent with the box plot and 3σ method, but it can extract a larger amount of data after outliers, resulting in better prediction performance.

[0113] Step S24: Smoothing: Use the moving average method to filter and denoise the digital signal. Select an appropriate window size and move it along the time axis of the signal. Take the average value of all samples in the window as the value of the center sample of the window. Set the window size to 5 and smooth all shield tunneling parameters. While removing most of the noise, retain the overall trend of change in the entire shield tunneling process.

[0114] Step S25, Normalization: Before building the model, the training set data is processed using a normalization method to reduce the influence of differences in the dimensions and scales of different parameters, thereby improving the fitting accuracy and speed. The specific method is as follows:

[0115]

[0116] In the formula, x norm Here are the normalized data, and x is the original data. max and x minThese are the maximum and minimum values ​​in the original data, respectively.

[0117] According to a further improvement of the present invention, step S3 specifically includes:

[0118] Step S31: Assign a score to the soil layer based on its type, plasticity index, and the proportion of different particle sizes in the soil. The score range is 0-9, resulting in a grading standard table. The grading standards are as follows:

[0119] Table 1 Grading Standards

[0120]

[0121]

[0122] Step S32: For composite soil layers, select the weighted average of the corresponding scores as the score value. The calculation formula is as follows:

[0123]

[0124] In the formula, n is the total number of soils in each ring, and S i The percentage (%) of different soil types in each ring of soil, d i The score for each ring of soil;

[0125] Step S33: Classify different types of soil according to the data in the classification standard table in step S31, and obtain the range of cutterhead rotation speed and total thrust of the tunnel boring machine in different types of soil.

[0126] According to a further improvement of the present invention, step S4 specifically includes:

[0127] Step S41: Calculate the Pearson correlation coefficient, Spearman correlation coefficient, and feature importance based on random forest for each pair of shield tunneling parameters, and plot the corresponding heatmaps. For feature selection based on Pearson and Spearman correlation coefficients, identify parameters that have a certain correlation (|r|≥0.2) with tunneling speed, cutterhead rotation speed, total thrust, and cutterhead torque. For feature selection based on random forest feature importance, select shield tunneling parameters whose importance is accumulated from high to low up to 0.95. Finally, select the intersection of the features selected by the three methods.

[0128] Pearson correlation coefficient is commonly used to explore the linear relationship between two variables. In tunnel engineering, the Pearson correlation coefficient between each pair of shield tunneling parameters is calculated, and a corresponding heatmap is plotted. The darker the color in the heatmap, the stronger the correlation between the two data points. For example... Figure 4As shown, r>0 indicates a positive correlation between the data, r<0 indicates a negative correlation between the data, r=0 indicates that the two data are uncorrelated, r=1 indicates that the two data are perfectly positively correlated, and r=-1 indicates that the two data are perfectly negatively correlated.

[0129] The calculation process for the Pearson correlation coefficient is as follows:

[0130] Calculate the covariance (cov(X,Y)) and standard deviation (s) of two variables. The covariance (cov(X,Y)) indicates whether the two variables have the same trend, and the standard deviation (s) indicates the dispersion of a single variable. Substituting the covariance and standard deviation into the formula for calculating the correlation coefficient yields the formula for calculating the Pearson correlation coefficient.

[0131]

[0132] Among them, s X and s Y Let X and Y represent the standard deviations, respectively.

[0133] The Spearman correlation coefficient is a method for measuring the correlation between two variables, suitable for situations where the data exhibits ordinal or nonlinear relationships. Its calculation method is based on comparing the ranks (or grades) of the two sets of data, thus avoiding the influence of data distribution patterns and outliers on the correlation assessment.

[0134] The basic idea behind Spearman's correlation coefficient is to transform the original data into rank data and calculate the correlation between the rank data. Rank refers to sorting the data from smallest to largest, and then replacing the original data with integers from 1 to n (where n is the sample size), so that each data point corresponds to a rank. By comparing the rank data, the degree of correlation between two sets of data can be obtained.

[0135] The Spearman correlation coefficient ranges from -1 to 1. A coefficient of 1 indicates a perfect positive correlation between the two sets of data, -1 indicates a perfect negative correlation, and 0 indicates no correlation. For example... Figure 5As shown, when the absolute value of the Spearman correlation coefficient is ≥0.8, it indicates a very strong correlation between the parameters; when its absolute value is ≥0.6 and <0.8, it indicates a strong correlation; when its absolute value is ≥0.4 and <0.6, it indicates a moderate correlation; when its absolute value is ≥0.2 and <0.4, it indicates a weak correlation; and when its absolute value is <0.2, it indicates a very weak correlation. Furthermore, the Spearman correlation coefficient exhibits invariance and monotonicity, meaning that for the same set of data, regardless of the order, the calculated correlation coefficient remains the same; and for two sets of data, if one set is greater than (less than) the other, then their correlation coefficient should also be greater than (less than) the other set.

[0136] The calculation process for the Spearman correlation coefficient is as follows:

[0137] ① Sort the two sets of data separately to obtain the ranks of the two sets of data;

[0138] ② Calculate the rank difference d for each data point, i.e., d = xy;

[0139] In the formula: x and y are the ranks of the two sets of data, respectively;

[0140] ③ Calculate the sum of squares S of the rank differences, i.e.

[0141] ④ Calculate the Spearman correlation coefficient r according to the following formula. s :

[0142] It is important to note that when two sets of data have the same rank, their average rank should be used as the representative rank to calculate the rank difference.

[0143] Random forest variable importance scoring can be used for feature selection, feature engineering, and model interpretation. Unlike traditional feature selection methods, random forest-based methods can handle complex feature problems such as high dimensionality, nonlinearity, and multicollinearity. Furthermore, random forests can be used to interpret the model through feature importance scoring, further clarifying the underlying reasons for the model's predictions.

[0144] Random forest-based variable importance scoring methods primarily assess feature importance by calculating the average reduction in impurity for each feature across all decision trees in the random forest. In a random forest, each time a decision tree is split, an optimal feature is selected for the split. Therefore, the importance of a feature can be measured by calculating the average reduction in impurity that feature brings when used for splitting across all decision trees.

[0145] After each decision tree in a random forest is constructed, the importance of a variable can be assessed by utilizing the split contribution of each variable within the decision tree. In classification problems, Gini Importance or Mean Decrease Impurity (MDI) can be used to measure variable importance; in regression problems, Mean Decrease Accuracy (MDA) can be used to measure variable importance.

[0146] Gini importance measures the importance of a variable by calculating the number of times each variable is used as a splitting node. During the construction of each decision tree, the Random Forest algorithm selects the splitting method that maximizes the Gini coefficient based on different splitting variables and splitting points. Therefore, the more times a variable is used as a splitting node, the higher its Gini importance. The Gini importance is calculated as follows:

[0147]

[0148] Where M is the number of decision trees in the random forest, and T m Let v represent the m-th decision tree. t Let be the variable used for splitting at node t, j be the variable to be calculated, and Δi(t,m) be the reduction in impurity at node t. Therefore, the Gini importance is mainly calculated based on the number of times each variable is selected as a splitting node in the random forest.

[0149] Mean Impurity Reduction (MDI) measures the importance of a variable by calculating the average amount of impurity reduction used to reduce nodes in a random forest. Node impurity can be measured by the Gini coefficient, entropy, or classification error rate, among other things. The formula for calculating Mean Impurity Reduction is as follows:

[0150]

[0151] Where Δi(t,m) represents the reduction in impurity of node t in the m-th decision tree, v t These are the variables used to partition node t. The MDI (Mean Differential Indicator) is primarily calculated based on the average amount each variable contributes to reducing node impurity within the random forest. Similar to Gini importance, the more times a variable is used to reduce node impurity, the higher its MDI.

[0152] MDA (Multi-Valued Aspect) is a method used in regression problems to measure the importance of variables. It is calculated by determining the average impact of each feature on the model's accuracy. Specifically, it calculates the percentage decrease in model accuracy when a particular variable is not involved in modeling. The larger this percentage, the greater the variable's contribution to the model's accuracy, and the higher its importance. The formula for calculating MDA is as follows:

[0153]

[0154] in, This represents the prediction result of the model trained on the j-th sample after the i-th feature is excluded. This represents the average of all model predictions, and N represents the number of samples.

[0155] Based on the random forest variable importance scoring method, the importance of shield tunneling parameters on tunneling speed, cutterhead rotation speed, total thrust, and cutterhead torque in tunnels 1-9 was explored. The scores are shown in Table 2.

[0156] Table 2. Variable Importance Scores Based on Random Forest

[0157]

[0158]

[0159] Finally, based on Pearson and Spearman correlation coefficients, feature selection identified parameters with a certain correlation (|r|≥0.2) with tunneling speed, cutterhead rotation speed, total thrust, and cutterhead torque. For feature selection based on random forest feature importance, shield tunneling parameters with importance accumulated from high to low up to 0.95 were selected. Finally, the intersection of the features selected by the three methods was chosen.

[0160] Step S42: Use a BP neural network model to construct the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque through forward propagation, backward propagation and gradient descent;

[0161] Step S43: Optimize the hyperparameters of the BP neural network model using a grid search method, specifying the possible value range of each hyperparameter and the possible values ​​for each parameter; traverse all possible hyperparameter combinations; employ K-fold cross-validation; in this embodiment, k-fold cross-validation is used, with k set to 10, meaning the training set is divided into 10 groups on average, and one group is randomly selected each time as the validation set. The remaining 9 groups are used for training, and this training and validation process is repeated 10 times, with each validation set being different. Finally, the mean square error (MSE) of the 10 validation sets is taken as the final MSE. After determining the mean square error of the minimum validation set, the hyperparameters of the BP neural network model can be determined, and further, the relationship between the total thrust and cutterhead rotation speed and the tunneling speed and cutterhead torque can be determined.

[0162] The process of K-fold cross-validation includes:

[0163] The data in the training set is divided into K groups on average. One group of data is randomly selected from these groups as the validation set, and the remaining K-1 groups are used for training.

[0164] Repeat the above process K times, with each time the selected validation set is different. Take the MSE of each validation set and use the average of the K times as the final MSE.

[0165] According to a further improvement of the present invention, the shield tunneling parameter prediction process in step S5 is divided into two parts: shield tunneling energy consumption minimization decision optimization and construction period shortest decision optimization. Both parts employ a BP neural network model loaded into the PSO algorithm, using different fitness functions for decision-making and optimization. The BP neural network model has one input layer (activation function sigmoid), one hidden layer, one Dropout layer (Dropout = 0.5), and one output layer. The Dropout layer is added to avoid overfitting, gradient vanishing, and gradient exploding, enhancing its generalization ability. The optimizer is set to Adam, the learning rate is set to 0.001, and the epochs are set to 5000. Early stopping is used to process the training process; if the model's error on the validation set does not improve for 20 consecutive epochs, the training process is interrupted to ensure efficiency. After training the BP neural network, the mean squared error of the validation set is obtained, which determines the hyperparameters of the BP neural network model. The BP neural network model is then loaded into the PSO algorithm for PSO optimization with two different decision objectives.

[0166] The specific optimization decision for minimizing shield tunneling energy consumption is as follows: Shield tunneling energy consumption refers to the work done per unit volume of soil excavated, and its calculation formula is:

[0167]

[0168] Among them, W T Work done by the cutter head torque, n is the cutter head rotation speed, T is the cutter head torque, t is time, W F The work done for the total thrust, where F is the total thrust and v is the tunneling speed;

[0169] Substituting the above relationship into this formula yields the fitness function for shield tunneling energy consumption:

[0170]

[0171] In the formula, n is the rotational speed of the shield cutterhead, and f T (x n ,x F ) represents the shield cutterhead torque mapping; F represents the total shield thrust; R represents the shield cutterhead excavation radius; f v (x n ,x F () is a mapping formula for shield tunneling speed;

[0172] The fitness function is applied to the algorithm flow to determine the historical optimal solution p. best and the global optimal solution g best Set the relevant parameters for the particle swarm, including inertia weight w = 0.9, acceleration constants c1 = c2 = 0.5, and maximum particle velocity V. max =1, number of particles N=100, number of iterations G max =100. After optimization calculation, the optimal cutter head speed n and total thrust F are obtained.

[0173] Generally speaking, the optimal solution can be found after 10 iterations, such as Figure 7 As shown, it is important to note that after optimizing the two tunneling parameters, the energy consumption of the shield should be minimized while ensuring that the tunneling speed is not too low. Otherwise, the tunneling efficiency will be too low, resulting in a longer construction period, which contradicts the original intention of optimizing the shield tunneling parameters. After the prediction is completed, it should be compared with the average and predicted values ​​of the tunneling speed under different soil types in the corresponding conditions. If the tunneling speed is not too low, the predicted result should be selected. In this embodiment, after PSO optimization, the predicted tunneling speed is not greater than the rated tunneling speed of the shield machine. Except for the sand layer of three tunnels, the output tunneling speed is greater than the average tunneling speed in the corresponding strata, which meets the requirements of shield tunneling efficiency and construction period.

[0174] The specific optimization strategy for minimizing the construction period involves: increasing the tunneling speed to improve tunneling efficiency and shorten the construction period; and utilizing the time consumed by the tunnel boring machine (TBM) for tunneling. In the formula, Let v be the excavation distance and v be the tunneling speed; substituting the relationship into the equation yields the fitness function as follows:

[0175]

[0176] In the formula, L is the shield excavation distance, set to 1m; t is the tunneling time; f v (x i,n ,x i,F ) is the mapping formula for the shield tunneling speed of the i-th ring;

[0177] Wherein, the cutter head rotation speed x i,n ∈[n min ,n max Total thrust x i,F ∈[F min ,F max When t takes a smaller value, a better fitness value is obtained. When the global optimal solution is obtained, that is, when the condition of the shortest construction period is met, the optimal parameter combination is obtained.

[0178] When using the PSO algorithm to optimize tunnel boring machine (TBM) operation, various parameters are set, including inertia weight w = 1.5, acceleration constants c1 = c2 = 0.5, and maximum particle velocity V. max =1, number of particles N=200, number of iterations G max =300. After optimization calculation, the optimal cutter head speed n and total thrust F are obtained.

[0179] Generally, in the same tunnel project, the larger the particle size, the lower the cutterhead rotation speed, and the greater the required total thrust. After obtaining the optimal parameters, they are compared with the average values ​​of the strata where the tunnel boring machine is located to determine whether the PSO optimization algorithm significantly increases the tunneling speed, thereby judging the effectiveness of the PSO algorithm. In this embodiment, after PSO optimization, except for the sand layer, the optimized tunneling speed of the other strata is greater than the average tunneling speed before optimization. Although the lowest tunneling speed in the sand layer is less than the average tunneling speed, the average tunneling speed is still conservative within the optimized tunneling speed range. Therefore, the method of shortening the construction period by optimizing the operating parameters using the PSO algorithm is reliable.

[0180] It should be noted that, unlike the decision objective of minimizing construction period, when the decision objective is to minimize shield tunneling energy consumption, lower energy consumption is not always better. The optimal cutterhead speed and total thrust should be selected while ensuring construction efficiency. If the cutterhead speed or total thrust parameters do not match the actual engineering situation, adjustments should be made promptly to ensure normal construction. After optimizing the shield tunneling parameters under both decision objectives, two sets of predicted shield tunneling energy consumption and tunneling speed values ​​are output. The optimal cutterhead speed and total thrust are then selected from the given two parameters.

[0181] Example 2:

[0182] For tunnels 1 to 9, minimizing shield tunneling energy consumption was the decision objective. The PSO algorithm was used to optimize the operating parameters, and the optimal solution was found after 10 iterations. The shield tunneling energy consumption fitness values ​​differed significantly among soil, sand, and gravel layers, and increased with increasing particle size in the strata. Table 3 shows the optimal active parameters for minimizing shield tunneling energy consumption for different shield machines and strata in tunnels 1-9.

[0183] Table 3. Optimized operating parameters for minimizing shield tunneling energy consumption in different geological formations.

[0184]

[0185]

[0186] After optimizing the operating parameters, it is necessary to consider not only minimizing the energy consumption of the tunnel boring machine (TBM), but also ensuring that the tunneling speed is not too low. Otherwise, it will result in low tunneling efficiency and excessively long construction period, which would contradict the original intention of optimizing the TBM operating parameters. The output tunneling speed is obtained through the relationship between total thrust, cutterhead rotation speed, and tunneling speed using a BP neural network model. After comparison with the average tunneling speed in the corresponding strata, the optimized tunneling speed after PSO optimization is greater than the average tunneling speed in the corresponding strata. Table 4 shows a comparison between the optimized tunneling speed and the average tunneling speed in its corresponding strata.

[0187] Table 4 shows the optimized tunneling speeds for different geological formations, using the minimum energy consumption of the shield tunnel as the decision-making operational parameter.

[0188]

[0189] Example 3:

[0190] Based on the shortest construction period as the decision-making method, the PSO algorithm is used to optimize the shield tunneling operation parameters. The parameters of the PSO algorithm are set as follows: inertia weight ω is set to 1.5, acceleration constants c1 and c2 are set to 0.5, and maximum particle velocity V... max Set to 1, particle number N is 200, and iteration count G. max The number of times is 300. The optimal parameters for the shortest construction period in different strata are shown in Table 5.

[0191] Table 5. Optimized operational parameters for different geological formations, with the shortest construction period as the decision-making parameter.

[0192]

[0193] After optimization using the PSO algorithm, the tunneling speed increased significantly. Statistical analysis of the average tunneling speed before and after optimization for tunnels 1 to 9 revealed that, except for sand layers, the minimum optimized tunneling speed was greater than the average speed before optimization. Although the minimum optimized tunneling speed in sand layers was less than the average speed, the average speed was still relatively conservative within the optimized range. Therefore, the PSO algorithm optimization model, which prioritizes minimizing the construction period, significantly improved the tunneling speed and shortened the construction period after optimizing the operating parameters. A comparison of the tunneling speeds before and after optimization is shown in Table 6.

[0194] Table 6 Comparison of tunneling speeds before and after optimization in different geological formations.

[0195]

[0196] Example 4:

[0197] This is a case study of the application of the PSO algorithm-based optimization method for key parameters of tunnel boring machines in a construction project.

[0198] This embodiment takes a shield tunnel project in the coastal area of ​​China as an example, and selects the shield tunnel data of rings 761-770 from February 27, 2023 to March 2, 2023 for analysis. After preprocessing the shield data, a total of 12 tunneling rings were formed. The stratum where the shield is located is a gravel layer. The PSO algorithm was used to optimize the shield operation parameters with the decision objectives of minimizing shield energy consumption and shortening the construction period. The comparison table of operation parameters and tunneling speed before and after optimization is shown in Table 7.

[0199] Table 7 Comparison of operational parameters and tunneling speed based on minimizing shield tunneling energy consumption and shortening construction period.

[0200]

[0201] As shown in Table 7, after optimizing the operating parameters using the PSO algorithm, both the tunneling speed and shield energy consumption were significantly improved. Specifically, in the optimization with the goal of minimizing shield energy consumption, the optimized tunneling speed increased by 16.28% compared to the actual average, and the optimized shield energy consumption decreased by 39.97% compared to the actual average. In the optimization with the goal of minimizing the construction period, the optimized tunneling speed increased by 24.58% compared to the actual average, and the optimized shield energy consumption decreased by 21.91% compared to the actual average. This demonstrates that the method of optimizing shield tunneling parameters based on the PSO algorithm is feasible.

[0202] A comparison of the predicted values ​​for minimizing shield tunneling energy consumption and minimizing construction period revealed that while the tunneling speeds were not significantly different, the predicted energy consumption based on minimizing shield tunneling energy consumption was far lower than that based on minimizing construction period. If there are no strict time constraints, the operating parameters based on minimizing shield tunneling energy consumption can be used for tunneling. This approach can significantly reduce shield tunneling energy consumption, thus saving project costs, and can also shorten the overall construction period with a higher tunneling speed, making it a highly efficient and energy-saving construction solution. It should be noted that the results obtained here are only from the above example and do not indicate that the optimization algorithm based on minimizing shield tunneling energy consumption is superior to the optimization algorithm based on minimizing construction period. In real-world cases, users are required to exercise their initiative and select the tunneling parameters that best suit the actual requirements of the project after comparing the tunneling parameters obtained from the two different decision objectives.

[0203] The above descriptions are preferred embodiments of the present invention and are not intended to limit the invention. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the invention should be included within the scope of protection of the invention.

Claims

1. A method for optimizing shield tunneling parameters based on the PSO algorithm, characterized in that, Includes the following steps: Step S1: Acquisition of raw tunneling data: Collect construction data of no less than 8 shield tunnels, which includes various geological conditions as raw tunneling data; Step S2, Data Preprocessing: Remove parameters from the original tunneling data obtained in Step S1 that are in the non-tunneling state and the shutdown and restart state of the tunneling machine. Select the first 10% of the tunneling state of each ring as the rising segment and the last 90% of the data as the stationary segment. Remove the parameters in the rising segment. Use the Mahalanobis distance method to process the data to remove outliers. Normalize the training set data to reduce the impact of differences in the dimensions and scales of different parameters. Then, perform statistical analysis on the tunneling parameters and organize them into time series data. Step S3, Geological Condition Classification: Establish corresponding grading standards according to different soil types. For composite soil layers, adopt corresponding scores as parameter values ​​to obtain the range of cutterhead rotation speed and total thrust of the tunnel boring machine in different soils. Step S4: Establish a prediction model: Establish the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque. Use total thrust and cutterhead rotation speed as initial parameters, input the initial parameters as training set into the artificial neural network model, train the neural network, and obtain the fitness formulas for optimizing shield tunneling energy consumption and construction period. Step S5, Prediction Stage: Using the training set from Step S4, the PSO particle swarm optimization algorithm is loaded into the BP artificial neural network model, and predictions are made on the optimal hyperparameterized artificial neural network model; with the minimum shield energy consumption and the shortest construction period as decision objectives, the predicted cutterhead rotation speed and total thrust are output respectively, obtaining the shield energy consumption and tunneling speed under two cutterhead rotation speeds and total thrusts; after comparison, a set of data that meets the engineering requirements is selected. Its specific process includes: Step S51: Establish an optimal hyperparameterized artificial neural network model for predicting cutterhead rotation speed and total thrust. The specific process is as follows: A BP artificial neural network model was established, and K-fold cross-validation was used to predict the shield cutterhead rotation speed and total shield thrust for the next 2 to 3 rings. Then, a corresponding mapping formula was constructed to obtain the relationship between the shield's active operating parameters (total thrust and cutterhead rotation speed) and the shield's passive parameters (tunneling speed and cutterhead torque). f v (x n ,x F )=v; f T (x n ,x F )=T; In the formula: x n x is the rotational speed of the cutter head. F v is the total thrust, v is the tunneling speed, and T is the cutterhead torque. Step S52: Establish the PSO optimization model. PSO particle swarm optimization is used to make decisions and optimize the shield tunneling with the goal of minimizing energy consumption and minimizing construction period, so as to obtain the optimal cutterhead speed and total thrust. Step S521: Initialize the position and velocity of the particle swarm, setting the velocity to 0 or generating it randomly within a certain range; Step S522: Load the relationship model between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque, and input it into the PSO optimization model; Step S523: Calculate the fitness value of each particle using the derived fitness formula to measure the quality of the particle's position, update the historical optimal solution and global optimal solution of each particle, and update its velocity and position based on the historical optimal solution and global optimal solution. Step S524: If the stopping condition is met, output the optimal solution; otherwise, continue to update the historical optimal solution and the global optimal solution for each particle until the maximum number of iterations is reached. Step S53: Establish a BP neural network. The specific process includes: Step S531: Construct a BP neural network model structure using an input layer, an intermediate layer, a Dropout layer, and an output layer, with the optimizer set to Adam; the activation function of the input layer is the sigmoid function. Step S532: Set the number of iterations to n, and use early stopping to process the training process. When the error of the model on the validation set has not been improved for 10 consecutive iterations, the training process is interrupted. Step S533: After the minimum mean square error of the validation set is determined, the hyperparameters of the BP neural network model can be determined at this time.

2. The method for optimizing shield tunneling parameters based on the PSO algorithm according to claim 1, characterized in that, When constructing the BP neural network model, the sample set is divided into a training set and a test set in an 8:2 ratio. One-tenth of the data in the training set is selected as the validation set to determine the pairwise relationships between data in the same group of parameters and to determine the prediction model. The test set is used to evaluate the error between the prediction model and the actual values ​​of the cutterhead speed and cutterhead torque.

3. The method for optimizing shield tunneling parameters based on the PSO algorithm according to claim 1, characterized in that, Step S2 specifically includes: Step S21: Remove data from the original tunneling data obtained in step S1, specifically data on the tunnel boring machine in non-tunneling states: Remove data from when the tunnel boring machine is stopped and during segment assembly. Set a discriminant function: f i =F i ·v i ·T i ·n i ; In the formula, F i v represents the total thrust at the i-th time step, in kN. i T represents the tunneling speed at the i-th time step, in mm / min. i n is the cutterhead torque at the i-th time step, in MN·m; i Let be the rotational speed of the tool turret at the i-th time step, in rpm; When any one of the total thrust, tunneling speed, cutterhead torque, and cutterhead rotation speed of the tunnel boring machine is 0, the discriminant function is 0, which means that the tunnel boring machine is in a non-tunneling state and the non-tunneling state data is removed. Step S22, taking the steady-state phase, including: Step S221: Select data with a tunneling cycle duration of more than 30 minutes and reorder them; Step S222: Arrange the tunneling status data of each ring by time, take the first 10% of the data as the rising segment, and the last 90% of the data as the stable segment; Step S23: Remove outlier data: The Mahalanobis distance method is used to process outlier data. The Mahalanobis distance method is a data similarity measurement method that calculates the distance between data points based on the covariance matrix of the data. The specific calculation method is as follows: In the formula, x i Let N be a single data point in the dataset, μ be the mean of the sample, and S be the covariance matrix of the data. This method is compared with other processing methods such as box plots and the 3σ method. Step 24: Smoothing: Use the moving average method to filter and denoise the digital signal. Select an appropriate window size and move it along the time axis of the signal. Take the average value of all samples in the window as the value of the center sample of the window. Set the window size to 5 and smooth all shield tunneling parameters. This removes most of the noise while retaining the overall trend of change in the entire shield tunneling process. Step S25, Normalization: Before building the model, the training set data is processed using a normalization method to reduce the influence of differences in the dimensions and scales of different parameters, thereby improving the fitting accuracy and speed. The specific method is as follows: In the formula, x norm Here are the normalized data, and x is the original data. max and x min These are the maximum and minimum values ​​in the original data, respectively.

4. The method for optimizing shield tunneling parameters based on the PSO algorithm according to claim 1, characterized in that, Step S3 specifically includes: Step S31: Assign a score to the soil layer based on its type, plasticity index, and the proportion of different particle sizes in the soil. The score ranges from 0 to 9, and a grading standard table is obtained. Step S32: For composite soil layers, select the weighted average of the corresponding scores as the score value. The calculation formula is as follows: In the formula, n is the total number of soils in each ring, and S i d represents the percentage of different soil types in each ring of soil. i The score for each ring of soil; Step S33: Classify different types of soil according to the data in the classification standard table in step S31, and obtain the range of cutterhead rotation speed and total thrust of the tunnel boring machine in different types of soil.

5. The method for optimizing shield tunneling parameters based on the PSO algorithm according to claim 1, characterized in that, Step S4 specifically includes: Step S41: Calculate the Pearson correlation coefficient, Spearman correlation coefficient, and feature importance based on random forest for each pair of shield tunneling parameters, and plot the corresponding heatmaps. For feature selection based on Pearson and Spearman correlation coefficients, identify parameters with a correlation |r|≥0.2 with tunneling speed, cutterhead rotation speed, total thrust, and cutterhead torque. For feature selection based on random forest feature importance, select shield tunneling parameters whose importance is accumulated from high to low up to 0.

95. Finally, select the intersection of the features selected by the three methods. Step S42: Use a BP neural network model to construct the relationship between total thrust and cutterhead rotation speed and tunneling speed and cutterhead torque through forward propagation, backward propagation and gradient descent; Step S43: Optimize the hyperparameters of the BP neural network model using a grid search method, specifying the possible range of hyperparameter values ​​and the possible values ​​of each parameter; traverse all possible combinations of hyperparameters; and use K-fold cross-validation. The process of K-fold cross-validation includes: The data in the training set is divided into K groups on average. One group of data is randomly selected from these groups as the validation set, and the remaining K-1 groups are used for training. Repeat the above process K times, with each time the selected validation set is different. Take the MSE of each validation set and use the average of the K times as the final MSE.

6. The method for optimizing shield tunneling parameters based on the PSO algorithm according to claim 1, characterized in that, The shield tunneling parameter prediction process in step S5 is divided into two parts: shield tunneling energy consumption minimization decision optimization and construction period shortest period decision optimization. The specific optimization decision for minimizing shield tunneling energy consumption is as follows: Shield tunneling energy consumption refers to the work done per unit volume of soil excavated, and its calculation formula is: Among them, W T Work done by the cutter head torque, n is the cutter head rotation speed, T is the cutter head torque, t is time, W F The work done for the total thrust, where F is the total thrust and v is the tunneling speed; Substituting the above relationship into this formula yields the fitness function for shield tunneling energy consumption: In the formula, n is the rotational speed of the shield cutterhead, and f T (x n ,x F () represents the shield cutterhead torque mapping, where F is the total shield thrust, R is the shield cutterhead excavation radius, and f v (x n ,x F () is a mapping formula for shield tunneling speed; The fitness function is applied to the algorithm flow to determine the historical optimal solution p. best and the global optimal solution g best Set the relevant parameters for the particle swarm, including inertia weight w = 0.9, acceleration constants c1 = c2 = 0.5, and maximum particle velocity V. max =1, number of particles N=100, number of iterations G max =100, and after optimization calculation, the optimal cutter head speed n and total thrust F are obtained; The specific optimization strategy for minimizing the construction period involves: increasing the tunneling speed to improve tunneling efficiency and shorten the construction period; and utilizing the time consumed by the tunnel boring machine (TBM) for tunneling. In the formula, Let v be the excavation distance and v be the tunneling speed; substituting the relationship into the equation yields the fitness function as follows: In the formula, L is the shield excavation distance, set to 1m; t is the tunneling time; f v (x i,n ,x i,F ) is the mapping formula for the shield tunneling speed of the i-th ring; Wherein, the cutter head rotation speed x i,n ∈[n min ,n max Total thrust x i,F ∈[F min ,F max When t takes a smaller value, a better fitness value is obtained. When the global optimal solution is obtained, that is, when the condition of the shortest construction period is met, the optimal parameter combination is obtained. When using the PSO algorithm to optimize tunnel boring machine (TBM) operation, various parameters are set, including an inertia weight w = 1.

5. The velocity constants are c1 = c2 = 0.5, and the maximum particle velocity is V. max =1, number of particles N=200, number of iterations G max =300, After optimization calculations, the optimal cutterhead rotation speed n and total thrust F are obtained.

Images