A shield master control parameter multi-objective optimization adjustment method and system

Abstract

The present application relates to the technical field of intelligent control of shield tunneling parameters, and discloses a shield master control parameter multi-objective optimization adjustment method and system, which comprises the following steps: collecting and preprocessing shield tunneling multi-source operation data, and constructing a time series prediction model with the input of advance speed, cutterhead rotating speed, slurry parameters and geological parameters, and the output of slurry pressure and tunneling specific energy. Based on the prediction results, a multi-objective optimization function is established with the minimum slurry pressure deviation and the minimum tunneling specific energy as the target, and a hybrid evolution strategy combining the particle swarm algorithm and the genetic algorithm is adopted to perform an optimized search on the master control parameters, so as to obtain an optimal parameter combination considering safety and energy consumption. The parameters are output to the shield control system, so as to realize the rolling multi-objective optimization adjustment of the tunneling parameters.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for tunnel boring machine (TBM) parameters, and more specifically, to a multi-objective optimization adjustment method and system for the main control parameters of a TBM. Background Technology

[0002] In shield tunneling, the setting and adjustment of tunneling parameters directly affect construction safety, tunneling efficiency, and equipment operating status. However, current engineering practice still relies heavily on the manual experience of the shield tunneling operators. Due to the spatial non-uniformity and time-varying nature of geological conditions, adjustments to tunneling parameters often lag under abrupt geological changes or complex geological conditions. Parameter control strategies are relatively broad, making precise adjustments difficult. Furthermore, the experience differences among different operators hinder the formation of a unified and stable optimal parameter system. Simultaneously, shield tunneling performance involves multiple indicators such as energy consumption and cutter wear. The focus on tunneling efficiency, safety, and stability varies under different geological conditions. Therefore, it is necessary to construct a comprehensive tunneling performance evaluation and decision-making mechanism that considers multiple performance indicators to support the optimized control of shield tunneling parameters.

[0003] Currently, extensive research has been conducted both domestically and internationally on intelligent decision-making for tunnel boring machine (TBM) parameters. Related methods mainly include those based on predictive models and those based on optimization algorithms. However, existing technologies largely focus on predicting macro-level indicators such as cost and schedule before construction, primarily serving construction management and proving difficult to apply to real-time parameter adjustment during the tunneling process. Some methods model historical tunneling data using statistical learning or machine learning to learn from the operator's experience, or employ supervised learning models to predict objective functions and manually select parameter combinations by setting weights. However, these methods generally suffer from problems such as reliance on human experience for weights, insufficient adaptability to complex geological conditions, and a lack of dynamic feedback mechanisms, making it difficult to meet the demands for refined, real-time TBM parameter control. For example, CN119513494A uses a supervised learning model to generate TBM main control parameter adjustment commands, but its ability to predict and optimize key tunneling parameters at the multi-objective level is limited. It fails to effectively achieve synergistic coupling between the predictive model and the optimization control model, making it difficult to provide continuous and accurate feedback adjustment support for the intelligent control system of the TBM. Summary of the Invention

[0004] To address the problems existing in the background technology, this invention proposes a multi-objective optimization adjustment method and system for the main control parameters of a tunnel boring machine (TBM). By predicting the passive response parameters during the TBM tunneling process in real time and combining this with a multi-objective optimization algorithm to optimize the active control parameters, the method aims to improve construction efficiency, reduce energy consumption, and ensure safety. The innovation of this method lies in its ability not only to achieve high-precision prediction of passive response parameters but also to realize real-time optimization and adjustment of active control parameters under different geological conditions, providing more scientific and intelligent decision support for TBM tunnel construction.

[0005] This invention proposes a multi-objective optimization adjustment method for the main control parameters of a tunnel boring machine, comprising: Collect multi-source operational data during the tunnel boring machine (TBM) excavation process and preprocess the multi-source operational data. The multi-source operational data includes at least the propulsion speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, TBM attitude parameters, and geological parameters at the corresponding tunneling mileage. Using propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume and geological parameters as input features, and slurry pressure and tunneling specific energy as output response indicators, a time-series prediction model for shield tunneling response parameters is constructed and trained. Based on the prediction results of the time series prediction model, a multi-objective optimization function is constructed with the objectives of minimizing the prediction deviation of slurry pressure and minimizing the tunneling specific energy. Based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, the main control parameters of the tunnel boring machine are optimized through multi-objective search to obtain the optimal combination of main control parameters that meets the requirements of safety and energy consumption. The optimal combination of main control parameters is output to the shield tunneling control system to guide the adjustment of shield tunneling parameters in the next control cycle, thereby realizing rolling multi-objective optimization adjustment of the shield tunneling process.

[0006] Furthermore, when collecting multi-source operational data during the tunnel boring machine (TBM) excavation process and preprocessing the multi-source operational data, the following steps are included: Real-time data collection is performed on the following parameters generated during shield tunneling: propulsion speed, propulsion cylinder pressure, cutterhead rotation speed, cutterhead torque, earth pressure value, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters at the corresponding tunneling mileage. The multi-source operational data from different acquisition devices with different sampling frequencies are timestamped, and the data are unified to a preset sampling period based on the resampling method. A complete synchronous data sequence with time points is obtained based on linear interpolation or spline interpolation methods. The synchronous data sequence is used to identify outliers based on the moving average method, median filtering method, box plot outlier statistics method or density-based anomaly detection algorithm, and outliers exceeding the preset threshold range are removed or replaced. For short-term data gaps that occur during the data collection process, forward padding, backward padding, linear interpolation, or mean estimation based on adjacent time windows are used to fill in the gaps and ensure the continuity of the time series data. The scaling of the operating parameters in each dimension is performed based on the normalization method or the z-score standardization method to make each input feature fall within a uniform numerical range. According to the preset time window length and sliding step size, the preprocessed data is sliced ​​to construct a time series input sample sequence, and the slurry pressure and tunneling specific energy at the corresponding time are used as label data to form a sample set for training the time series prediction model.

[0007] Furthermore, when constructing and training a time-series prediction model for the shield tunneling response parameters, the following steps are included: Based on the temporal variation characteristics of tunnel boring machine (TBM) data, a temporal prediction model for TBM response parameters is constructed. This model comprises an input layer, a long short-term memory (LSTM) network layer, and a fully connected output layer. The input layer is configured to receive the time-series characteristic sequences of propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume and formation parameters within a preset time window, and the output layer is configured to output the predicted values ​​of slurry pressure and tunneling specific energy for the corresponding time steps. The weight matrix and bias terms of the long short-term memory network layer are initialized using a random initialization method, and the hyperparameters required for training the time series prediction model, such as the initial learning rate, batch size, and number of training epochs, are set. The time series sample set is divided into training set, validation set and test set according to a preset ratio, and the parameters of the time series prediction model are trained, hyperparameters are tuned and the generalization performance of the model is evaluated according to the training set, validation set and test set respectively. The training objective function is based on the mean squared error loss function or the weighted loss function, and the long short-term memory network layer is iteratively trained using the backpropagation algorithm combined with the Adam optimizer. The convergence of the loss function during model training is monitored based on the validation set, and model hyperparameters such as learning rate, hidden layer dimension, and time window length are adjusted according to the validation results. After the model training is completed, the optimal long short-term memory network model parameters are stored and a real-time inference interface is built.

[0008] Furthermore, when adjusting model hyperparameters such as learning rate, hidden layer dimension, and time window length based on validation results, the following should be included: Based on the validation set sample data, the time series prediction model in the current training round is used to make predictions, and the corresponding validation set loss function value is obtained. Based on the change of the loss function value on the validation set with each training epoch, the convergence state of the loss function during model training is monitored, where: When the loss function does not meet the preset convergence condition, at least one of the hyperparameters of the model, namely the learning rate, hidden layer dimension, and time window length, is adjusted according to the convergence state.

[0009] Furthermore, based on the prediction results of the time-series prediction model, when constructing a multi-objective optimization function with the objectives of minimizing slurry pressure prediction deviation and minimizing tunneling specific energy, the following functions are included: Using the minimum tunneling specific energy and the minimum difference between the predicted slurry pressure and the corresponding target value as multi-objective optimization objectives, and with the shield attitude parameters meeting the preset allowable range as constraints, a multi-objective optimization control mathematical model for shield tunneling parameters is established by selecting shield tunneling speed, cutterhead torque, cutterhead rotation speed, and total thrust as design variables. Among them, the cutterhead rotation speed and tunneling speed are the main optimization objects, and the shield tunneling parameters are coordinated and optimized. Coordination optimization uses the shield tunneling state parameters in the current control cycle, the predicted values ​​of tunneling specific energy and slurry pressure obtained from the multi-objective optimization control mathematical model, and the initial value ranges of cutterhead rotation speed and tunneling speed, as well as the associated cutterhead torque and total thrust constraints, as input parameters. The cutterhead rotation speed and tunneling speed are determined as the main optimization variables, and the cutterhead torque and total thrust are set as follow-up adjustment variables to establish a coupled and coordinated adjustment relationship between the cutterhead rotation speed and tunneling speed. Based on the main optimization variables, multiple candidate parameter combinations of cutterhead rotation speed and tunneling speed are generated. The corresponding tunneling specific energy evaluation index and slurry pressure prediction deviation are calculated for each candidate parameter combination. Candidate parameter combinations that do not meet the shield attitude parameter constraint range are eliminated. A multi-objective comprehensive evaluation is conducted on the candidate parameter combinations that meet the constraints. A trade-off is made between minimizing the tunneling specific energy and minimizing the slurry pressure prediction deviation, and the optimal combination of cutterhead rotation speed and tunneling speed parameters with the best coordination is selected. The optimal cutterhead rotation speed and tunneling speed are then used as outputs to adjust the corresponding cutterhead torque and total thrust parameters, forming the shield tunneling main control parameter adjustment command for the next control cycle.

[0010] Furthermore, the objective function of the multi-objective optimization model for tunnel boring parameters is set based on Equation 1: Formula 1 Where H represents the objective function of the multi-objective optimization model for the main control parameters of the tunnel boring machine; H1 represents the tunneling specific energy; H2 represents the absolute value of the difference between the actual and predicted values ​​of the slurry pressure; F represents the total thrust; v represents the propulsion speed; and n represents the cutterhead rotation speed. R represents the cutterhead torque, and R represents the cutterhead radius of the tunnel boring machine. This indicates the true value of the mud-water pressure. This represents the predicted value of mud and water pressure.

[0011] Furthermore, the multi-objective optimization is based on Pareto optimality, using the PSO-GA algorithm to generate the Pareto front solution, and combining LSTM to predict the shield tunneling state to iteratively optimize the main control parameters.

[0012] Furthermore, based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, multi-objective search optimization is performed on the main control parameters of the tunnel boring machine to obtain the optimal combination of main control parameters that meets the requirements of safety and energy consumption. This includes: Set evolutionary parameters and initialize the particle swarm. The evolutionary parameters include the swarm size, learning factor, and number of termination iterations. Initialize the particle swarm and store the historical best solutions corresponding to each particle to form a set of memory best solutions. The optimal parameter codes obtained by each particle in the particle swarm optimization algorithm are used as the operation objects of the genetic algorithm. When the parameter codes can represent better master control parameter values, the chromosome codes in the genetic algorithm are updated and overwritten. Based on the crossover and mutation operations of the genetic algorithm and the parameter update operations of the particle swarm optimization algorithm, the search individuals are evolved, and the memory optimal solution set is updated synchronously during the evolution process, wherein: When the preset heuristic factor search conditions are met, a heuristic strategy is used to optimize the search in the subsequent search process, and the set of memory optimal solutions is continuously updated; at the same time, it is determined whether the preset search termination condition has been reached. When the search termination condition is met, the hybrid evolution search ends, and the optimal combination of master control parameters stored in the particle swarm is taken as the global optimal result of the shield tunneling master control parameters.

[0013] Furthermore, the constraints on the tunnel boring machine (TBM) parameters are as follows: x 前 x 后 y 前 y 后 Where, x 前 Indicates the horizontal deviation of the tunnel boring machine's front end, x 后 Indicates the horizontal deviation of the rear end of the tunnel boring machine, y 前 Indicates the vertical deviation of the tunnel boring machine's front end, y 后 This indicates the vertical deviation at the rear end of the tunnel boring machine.

[0014] Compared with existing technologies, the advantages of this invention are as follows: By uniformly collecting and preprocessing multi-source operational data such as advance speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters during shield tunneling, a basic data system capable of comprehensively reflecting the changes in shield tunneling conditions and geological formations is constructed. This provides reliable data support for subsequent model analysis and parameter optimization, effectively avoiding the problem of incomplete information caused by single parameters or empirical judgments. Based on this, by introducing a time-series prediction model with advance speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and geological parameters as input features, dynamic prediction of key response indicators such as slurry pressure and tunneling specific energy is achieved. This allows the shield control system to perceive the trend of state changes during tunneling in advance, reducing the risk of parameter adjustment lag caused by sudden geological changes or fluctuations in operating conditions, thereby improving the stability and controllability of the tunneling process. Furthermore, this invention constructs a multi-objective optimization function based on the prediction results, aiming to minimize the prediction deviation of slurry pressure and the specific energy of tunneling. It employs a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithms to perform global search and collaborative optimization of the shield tunneling master control parameters. Under the premise of meeting construction safety constraints, this effectively reduces tunneling energy consumption and optimizes the selection of master control parameter configurations, avoiding the limitations of traditional methods that rely on manual experience or fixed weights. Finally, by outputting the optimal combination of master control parameters to the shield tunneling control system in real time to guide the adjustment of tunneling parameters in the next control cycle, a rolling, multi-objective optimization adjustment mechanism for shield tunneling parameters is realized. This method helps improve the adaptability of the shield tunneling process to complex geological conditions, enhances construction safety and tunneling efficiency, reduces equipment energy consumption and operational risks, and has significant engineering application value and promotional significance.

[0015] On the other hand, this application also provides a multi-objective optimization and adjustment system for the main control parameters of a tunnel boring machine, including: The acquisition module is configured to acquire multi-source operational data during the tunnel boring machine (TBM) excavation process and preprocess the multi-source operational data. The multi-source operational data includes at least the propulsion speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, TBM attitude parameters, and geological parameters at the corresponding tunneling mileage. The processing module, electrically connected to the acquisition module, is configured to use propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and geological parameters as input features, and slurry pressure and tunneling specific energy as output response indicators to construct and train a time-series prediction model for shield tunneling response parameters. The processing module is also configured to construct a multi-objective optimization function based on the prediction results of the time-series prediction model, with the objectives of minimizing slurry pressure prediction deviation and minimizing tunneling specific energy. The output module, electrically connected to the processing module, is configured to perform multi-objective search optimization of the shield tunneling main control parameters based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, to obtain the optimal combination of main control parameters that meets safety and energy consumption requirements. The output module is also configured to output the optimal combination of main control parameters to the shield tunneling control system to guide the adjustment of shield tunneling parameters in the next control cycle, thereby realizing rolling multi-objective optimization adjustment of the shield tunneling process.

[0016] It is understood that the shield tunneling main control parameter multi-objective optimization adjustment method and system in the above embodiments of the present invention have the same beneficial effects, and will not be described in detail here. Attached Figure Description

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a method for multi-objective optimization adjustment of main control parameters of a tunnel boring machine, provided in an embodiment of the present invention; Figure 2 A flowchart illustrating a method for multi-objective optimization adjustment of main control parameters of a tunnel boring machine, provided in an embodiment of the present invention; Figure 3 The flowchart of the PSO-GA hybrid algorithm for optimizing the main control parameters of the tunnel boring machine provided in this embodiment of the invention is shown below. Figure 4 This is a schematic diagram of the optimal range of the cutter head rotation speed provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the tunneling speed optimization range provided in an embodiment of the present invention; Figure 6 This is a functional block diagram of a shield tunneling main control parameter multi-objective optimization adjustment system provided in an embodiment of the present invention. Detailed Implementation

[0018] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] Shield tunneling refers to the construction process in which a tunnel boring machine (TBM) cuts through the soil with its cutterhead and advances simultaneously to form the tunnel structure during underground tunnel construction, thereby achieving safe and efficient tunnel formation.

[0020] Advance speed refers to the distance a tunnel boring machine (TBM) advances along the tunnel axis per unit time during the tunneling process, and is used to characterize the efficiency of TBM tunneling.

[0021] The cutterhead rotation speed refers to the number of times the cutterhead of a tunnel boring machine rotates per unit time, and is used to reflect the cutting intensity of the cutterhead on the soil.

[0022] Cutter head torque refers to the amount of torque required to drive the cutter head to rotate, and is used to characterize the cutting resistance of the cutter head and the hardness characteristics of the formation.

[0023] Earth pressure refers to the soil pressure acting on the excavation face or chamber of a tunnel boring machine, used to maintain the stability of the excavation face and prevent ground collapse.

[0024] Slag improvement parameters refer to the additive ratio or injection volume parameters used to adjust the fluidity and stability of slag, and are used to improve the performance of tunneling soil.

[0025] Grout injection volume refers to the volume or flow rate of stabilizing or lubricating grout injected into the shield tunneling system to control earth pressure and balance the strata.

[0026] Grout discharge volume refers to the volume or flow rate of grout discharged from the tunnel boring machine system, which reflects the grout circulation and pressure balance.

[0027] Tunnel boring machine (TBM) attitude parameters refer to a set of parameters that describe the spatial position and attitude state of the TBM, and are used to reflect the attitude deviation and control state of the TBM.

[0028] Time series prediction models refer to prediction models built based on time series data, used to characterize the dynamic changes of tunnel boring machine parameters over time.

[0029] Particle swarm optimization (PSO) is an intelligent optimization algorithm that simulates the cooperative behavior of a group and is used to search for the optimal solution in the parameter space.

[0030] Genetic algorithms are global optimization algorithms based on natural selection and genetic mechanisms, used to optimize parameters for complex nonlinear problems.

[0031] Rolling multi-objective optimization adjustment refers to an adjustment method that repeatedly performs prediction and optimization within a continuous control cycle, used to achieve dynamic adaptive optimization of shield tunneling parameters.

[0032] The moving average method refers to the method of taking a local average of continuous data to smooth time series data and reduce the impact of random fluctuations.

[0033] Median filtering is a method that replaces the original value with the median of neighboring data to remove impulse outliers.

[0034] Box plot outlier statistics refers to statistical methods that identify outliers based on quartile ranges and are used to detect extreme outliers.

[0035] Density-based anomaly detection algorithms are algorithms that identify anomalies based on sample density distribution, and are used to discover data anomalies in low-probability or sparse areas.

[0036] Synchronous data sequences refer to multi-source unified time series data formed after time alignment and resampling, used to ensure the temporal consistency between different parameters.

[0037] Normalization refers to the processing method of mapping data to a fixed range, which is used to eliminate the influence of different units on model training.

[0038] z-score standardization is a data standardization method that involves subtracting the mean and dividing by the standard deviation to give the data zero mean and unit variance.

[0039] Scaling transformation refers to the process of converting the numerical range of the original data to improve the stability of model training.

[0040] The initial learning rate refers to the step size of parameter updates at the beginning of model training, and is used to control the convergence speed of the model.

[0041] Batch size refers to the number of samples used in training each time the model parameters are updated, and is used to balance training efficiency and stability.

[0042] The number of training rounds refers to the number of times the model completes iterative training on all training data, and is used to control the adequacy of model training.

[0043] Hyperparameters are parameters that are manually set before model training and are not automatically learned through training. They are used to control the model structure and training process.

[0044] The mean squared error loss function is a function that measures the performance of a model by calculating the squared error between the predicted and the true values, and is used to train regression prediction models.

[0045] A weighted loss function is a loss function that assigns different weights to different prediction targets or sample errors, and is used to balance the needs of multi-objective training.

[0046] The training objective function is an evaluation function used to guide the optimization of model parameters and to measure the model's predictive performance.

[0047] Backpropagation is an algorithm that calculates the gradient of parameters by propagating the error backward, and is used to update the parameters of a neural network model.

[0048] The Adam optimizer is an optimization algorithm that combines momentum and adaptive learning rate to improve model training efficiency and convergence stability.

[0049] Long Short-Term Memory (LSTM) network layers refer to a neural network structure that can capture long-term temporal dependencies and is used to process tunnel boring machine (TBM) time-series data.

[0050] Convergence refers to the stable state of the loss function during model training as it changes over time, and is used to determine whether the model training has reached its optimal state.

[0051] Hidden layer dimension refers to the number of neurons in the hidden layer of a neural network, which is used to control the expressive power of the model.

[0052] The time window length refers to the number of historical time steps included when constructing time series samples, and is used to characterize the range of time dependencies.

[0053] Heuristic factor search refers to a search method that introduces empirical or strategic factors into the optimization process to accelerate algorithm convergence and improve global search capabilities.

[0054] like Figures 1-5 In some embodiments of this application, this embodiment provides a multi-objective optimization adjustment method for the main control parameters of a tunnel boring machine, including: Step S100: Collect multi-source operation data during the shield tunneling process and preprocess the multi-source operation data. The multi-source operation data includes at least the propulsion speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters at the corresponding tunneling mileage.

[0055] Specifically, the process of collecting and preprocessing multi-source operational data during shield tunneling includes: real-time acquisition of data such as propulsion speed, propulsion cylinder pressure, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters at corresponding tunneling mileages; timestamping the multi-source operational data from different acquisition devices with different sampling frequencies; unifying various data types to a preset sampling period based on resampling methods; obtaining a complete synchronous data sequence with time points based on linear interpolation or spline interpolation methods; and processing the data using methods such as moving average, median filtering, box plot outlier statistics, or density-based anomaly detection algorithms. The synchronized data sequence is used to identify outliers, and outliers exceeding a preset threshold range are removed or replaced. For short-term data gaps that occur during the acquisition process, forward padding, backward padding, linear interpolation, or mean estimation based on adjacent time windows are used to fill in the gaps, ensuring the continuity of the time series data. The operating parameters of each dimension are scaled using normalization or z-score standardization methods to ensure that each input feature is within a uniform numerical range. The preprocessed data is sliced ​​according to the preset time window length and sliding step size to construct a time series input sample sequence, and the mud pressure and tunneling specific energy at the corresponding time are used as label data to form a sample set for training the time series prediction model.

[0056] Understandably, by uniformly collecting and systematically preprocessing multi-source, heterogeneous operational data during tunnel boring machine (TBM) excavation, problems such as inconsistent sampling frequencies, missing data, and abnormal noise between different acquisition devices can be eliminated, thereby constructing a high-quality time-series data foundation that is time-aligned, continuous, and scale-consistent. Specifically, firstly, timestamp calibration and resampling are used to synchronize multi-source data on a unified time axis. Then, interpolation, filtering, and anomaly detection methods are used to improve data integrity and reliability. Furthermore, scale transformation is used to eliminate the influence of different physical dimensions on model training. Finally, a time-series sample sequence that reflects the dynamic characteristics of TBM excavation is constructed using a sliding time window approach, providing reliable data support for subsequent time-series prediction models to accurately learn the time-series correlation between excavation parameters and response indicators.

[0057] For example, in actual tunnel boring machine (TBM) construction, parameters such as advance speed, cutterhead rotation speed, and earth pressure are collected by different sensors. These sensors have different sampling periods and are susceptible to outliers or short-term data gaps due to construction disturbances. By timestamping and resampling these data, data from different sources can be aligned to the same time point. Then, moving averages or median filtering are used to remove instantaneous noise, anomaly detection algorithms are used to eliminate extreme outliers, and missing data is appropriately supplemented, resulting in a continuous and stable synchronous data sequence. Subsequently, the processed multidimensional parameters are normalized or standardized, and input samples are constructed according to a set time window. The slurry pressure and tunneling specific energy at the corresponding time are used as output labels, forming a dataset suitable for training time-series prediction models to characterize the dynamic relationship between parameter changes and response indicators during TBM tunneling.

[0058] Step S200: Using propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume and geological parameters as input features, and slurry pressure and tunneling specific energy as output response indicators, construct and train a time series prediction model for shield tunneling response parameters.

[0059] Specifically, the construction and training of a time-series prediction model for shield tunneling response parameters includes: constructing a time-series prediction model for shield tunneling response parameters based on the time-series variation characteristics of shield tunneling data. This model includes an input layer, a long short-term memory (LSTM) network layer, and a fully connected output layer. The input layer is configured to receive time-series feature sequences of propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and ground parameters within a preset time window. The output layer is configured to output predicted values ​​of slurry pressure and tunneling specific energy for the corresponding time step. The weight matrix and bias terms of the LSTM network layer are initialized using a random initialization method, and hyperparameters such as the initial learning rate, batch size, and number of training rounds required for training the time-series prediction model are set. The time-series sample set is divided into training, validation, and test sets according to a preset ratio, and parameter training, hyperparameter adjustment, and model generalization performance evaluation are performed based on the training, validation, and test sets, respectively. The mean squared error loss function or weighted loss function is used as the training objective function, and the model is trained using a backpropagation algorithm combined with Adam. The optimizer iteratively trains the long short-term memory network layers; it monitors the convergence of the loss function during model training based on the validation set, and adjusts model hyperparameters such as learning rate, hidden layer dimension, and time window length according to the validation results; after the model training is completed, the optimal long short-term memory network model parameters are stored and a real-time inference interface is built.

[0060] Specifically, when adjusting model hyperparameters such as learning rate, hidden layer dimension, and time window length based on validation results, the process includes: predicting the time series prediction model for the current training epoch based on validation set sample data to obtain the corresponding validation set loss function value; monitoring the convergence state of the loss function during model training based on the changes in the validation set loss function value with training epochs, wherein: when the loss function does not meet the preset convergence condition, at least one of the model's learning rate, hidden layer dimension, and time window length is adjusted according to the convergence state.

[0061] Understandably, by leveraging the temporal correlations exhibited by multidimensional operational parameters during tunnel boring machine (TBM) excavation over time, a temporal prediction model is constructed, comprising an input layer, a long short-term memory (LSM) network layer, and a fully connected output layer, to dynamically model and predict key response indicators during the tunneling process. This model receives multidimensional temporal features such as propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and geological parameters within a specific time window through the input layer. Utilizing the LSM network's memory and forgetting mechanism for historical information, it captures the nonlinear temporal relationships between tunneling parameters and slurry pressure and tunneling specific energy. Through iterative training using a mean squared error or weighted loss function combined with backpropagation and the Adam optimizer, and by monitoring the convergence of the loss function based on the validation set and adaptively adjusting hyperparameters such as the learning rate, hidden layer dimension, and time window length, the constructed temporal prediction model maintains good prediction accuracy and generalization ability under different working conditions and geological conditions, thus providing a reliable predictive basis for subsequent parameter optimization and control decisions.

[0062] For example, during actual tunnel boring machine (TBM) excavation, the system can continuously collect data such as advance speed, cutterhead rotation speed, slurry injection volume, and geological parameters within a certain time range. It then constructs an input feature sequence according to a preset time window and inputs it into a time-series prediction model containing a long short-term memory (LSTM) network layer. During the training phase, the model continuously learns the mapping relationship between changes in tunneling parameters and slurry pressure and tunneling specific energy using the training set. The system also uses a validation set to monitor the change in the loss function with each training iteration. When the loss on the validation set decreases slowly or fluctuates, the system can adjust the learning rate or the size of the hidden layers accordingly to avoid overfitting or underfitting. After training is complete, the solidified optimal model parameters can be used for real-time inference, predicting the slurry pressure and tunneling specific energy at the next moment during actual tunneling, providing a forward-looking reference for optimizing the main control parameters of the TBM.

[0063] Step S300: Based on the prediction results of the time series prediction model, construct a multi-objective optimization function with the objectives of minimizing the prediction deviation of slurry pressure and minimizing the tunneling specific energy.

[0064] Specifically, based on the prediction results of the time-series prediction model, when constructing a multi-objective optimization function with the objectives of minimizing the slurry pressure prediction deviation and minimizing the tunneling specific energy, the following steps are taken: The minimization of tunneling specific energy and the minimum difference between the predicted slurry pressure value and the corresponding target value are used as multi-objective optimization objectives. The shield attitude parameters satisfying the preset allowable range are used as constraints. The shield tunneling speed, cutterhead torque, cutterhead rotation speed, and total thrust are selected as design variables. A multi-objective optimization control mathematical model for the shield tunneling parameters is established. The cutterhead rotation speed and tunneling speed are used as the main optimization objects to coordinate and optimize the shield tunneling parameters. The coordination optimization uses the shield tunneling state parameters within the current control cycle, the predicted tunneling specific energy and slurry pressure values ​​obtained from the multi-objective optimization control mathematical model, and the initial value ranges corresponding to the cutterhead rotation speed and tunneling speed, along with their associated cutterhead torque and total thrust constraints. The system uses the following parameters as inputs: cutterhead rotation speed and tunneling speed are determined as the main optimization variables, while cutterhead torque and total thrust are set as follow-up adjustment variables, establishing a coupled and coordinated adjustment relationship between cutterhead rotation speed and tunneling speed; multiple candidate parameter combinations of cutterhead rotation speed and tunneling speed are generated based on the main optimization variables; the corresponding tunneling specific energy evaluation index and slurry pressure prediction deviation are calculated for each candidate parameter combination, and candidate parameter combinations that do not meet the shield attitude parameter constraints are eliminated; a multi-objective comprehensive evaluation is performed on the candidate parameter combinations that meet the constraints, balancing the minimum tunneling specific energy with the minimum slurry pressure prediction deviation, and selecting the optimal cutterhead rotation speed and tunneling speed parameter combination; the optimal cutterhead rotation speed and tunneling speed are used as output results, and the corresponding cutterhead torque and total thrust parameters are adjusted accordingly to form the shield main control parameter adjustment command for the next control cycle.

[0065] Specifically, the objective function of the multi-objective optimization model for tunnel boring machine (TBM) parameters is set based on Equation 1: Formula 1 Where H represents the objective function of the multi-objective optimization model for the main control parameters of the tunnel boring machine; H1 represents the tunneling specific energy; H2 represents the absolute value of the difference between the actual and predicted values ​​of the slurry pressure; F represents the total thrust; v represents the propulsion speed; and n represents the cutterhead rotation speed. R represents the cutterhead torque, and R represents the cutterhead radius of the tunnel boring machine. This indicates the true value of the mud-water pressure. This represents the predicted value of mud and water pressure.

[0066] Specifically, the multi-objective optimization is based on Pareto optimality, uses the PSO-GA algorithm to generate Pareto front solutions, and combines LSTM to predict the shield tunneling state to iteratively optimize the main control parameters.

[0067] Understandably, the predicted values ​​of slurry pressure and tunneling specific energy from the time-series prediction model are introduced into the shield tunneling parameter optimization control process. A multi-objective optimization model is constructed with the objectives of minimizing tunneling specific energy and minimizing slurry pressure prediction deviation, achieving coordinated optimization of tunneling parameters while satisfying shield attitude safety constraints. By using propulsion speed, cutterhead rotation speed, cutterhead torque, and total thrust as design variables, with cutterhead rotation speed and tunneling speed as the main optimization variables and cutterhead torque and total thrust as follow-up adjustment variables, a coupling and coordination relationship between the main control parameters is established. The tunneling specific energy and slurry pressure response under different parameter combinations are evaluated based on the prediction model. The multi-objective optimization process adopts an optimization strategy based on Pareto optimality. The PSO-GA hybrid algorithm is used to search and generate a Pareto front solution set that satisfies the constraints. A trade-off is made between optimal energy consumption and pressure stability, and the combination of main control parameters with the best overall performance is selected, thereby achieving dynamic balance control between safety, stability, and energy efficiency in the shield tunneling process.

[0068] For example, within a certain shield tunneling control cycle, the system first predicts the slurry pressure and tunneling specific energy corresponding to different combinations of cutterhead rotation speed and tunneling speed under the current state, based on a pre-trained LSTM time-series prediction model. Then, using a multi-objective function composed of the prediction deviations of tunneling specific energy and slurry pressure as evaluation indicators, multiple sets of candidate parameter combinations are generated, and schemes that do not meet the shield attitude constraints are automatically eliminated. For the remaining candidate schemes, a multi-objective search is performed using the PSO-GA algorithm to form a set of Pareto optimal solutions, from which the optimal parameter combination that balances low energy consumption and pressure stability is selected. Finally, the selected cutterhead rotation speed and tunneling speed are used as the main control outputs, and the corresponding cutterhead torque and total thrust parameters are adjusted simultaneously to generate shield tunneling adjustment commands for the next control cycle, achieving adaptive optimization control of the tunneling process.

[0069] Step S400: Based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, multi-objective search optimization is performed on the main control parameters of the tunnel boring machine to obtain the optimal combination of main control parameters that meets the requirements of safety and energy consumption.

[0070] Specifically, based on a hybrid evolutionary strategy combining particle swarm optimization (PSO) and genetic algorithm, when performing multi-objective search optimization on the main control parameters of a tunnel boring machine (TBM) to obtain the optimal combination of main control parameters that meets safety and energy consumption requirements, the process includes: setting evolutionary parameters and initializing the particle swarm, wherein the evolutionary parameters include swarm size, learning factor, and number of termination iterations; initializing the particle swarm and storing the historical optimal solutions corresponding to each particle to form a set of memory optimal solutions; using the optimal parameter codes obtained by each particle in the PSO algorithm as the operation objects of the genetic algorithm, wherein when the parameter codes can represent better main control parameter values, the genetic algorithm operates on the optimal parameters obtained by each particle in the PSO algorithm. The chromosome encoding in the genetic algorithm is updated and overwritten; based on the crossover and mutation operations of the genetic algorithm and the parameter update operations of the particle swarm algorithm, the search individuals are evolved, and the memory optimal solution set is updated synchronously during the evolution process. Specifically: when the preset heuristic factor search conditions are met, a heuristic strategy is used to optimize the search in the subsequent search process, and the memory optimal solution set is continuously updated; at the same time, it is determined whether the preset search termination condition has been reached; when the search termination condition is reached, the hybrid evolution search ends, and the optimal master control parameter combination stored in the memory of the particle swarm is taken as the global optimal result of the shield tunneling master control parameters.

[0071] Specifically, the constraints on the tunnel boring machine (TBM) parameters are as follows: x 前 x 后 y 前 y 后 Where, x 前 Indicates the horizontal deviation of the tunnel boring machine's front end, x 后 Indicates the horizontal deviation of the rear end of the tunnel boring machine, y 前 Indicates the vertical deviation of the tunnel boring machine's front end, y 后 This indicates the vertical deviation at the rear end of the tunnel boring machine.

[0072] Understandably, by integrating the collaborative search capability of the particle swarm optimization (PSO) algorithm with the global evolutionary capability of the genetic algorithm, a hybrid evolutionary multi-objective optimization strategy is constructed for the efficient search and optimization of the main control parameters during tunnel boring machine (TBM) excavation. In this strategy, the PSO algorithm is responsible for rapid optimization within a continuous parameter space, guiding the search direction through historical optimal solutions and global optimal solutions. The genetic algorithm enhances the diversity of the search process through crossover and mutation operations, preventing the PSO algorithm from getting trapped in local optima. By introducing a memory set of optimal solutions during the evolutionary process, the optimal parameter combinations are continuously preserved and updated, and a heuristic search is switched to improve convergence efficiency when the heuristic factor search conditions are met. Simultaneously, the TBM attitude deviation parameters are incorporated as constraints into the search process, and solutions that do not meet safety operation requirements are constrained and filtered. Finally, within the multi-objective optimization framework, the globally optimal combination of the TBM's main control parameters that simultaneously satisfies safety constraints and energy consumption optimization objectives is obtained.

[0073] For example, during the actual tunneling process of a tunnel boring machine (TBM), the system first initializes the particle swarm based on preset evolutionary parameters such as swarm size, learning factor, and number of iterations. Each particle is assigned a set of master control parameters as an initial solution, and the historical best solution for each particle is recorded. Subsequently, the optimal parameters obtained from the particle swarm are encoded and introduced into a genetic algorithm. New parameter combinations are generated through crossover and mutation operations, and this process is synchronized with the particle swarm parameter update. During the search process, the system continuously evaluates the energy consumption indicators and attitude deviation constraints corresponding to each parameter combination. When the heuristic factor triggering condition is met, a heuristic search strategy is used to accelerate convergence. Finally, after reaching the preset iteration termination condition, the master control parameter combination that satisfies the front and rear horizontal and vertical deviation constraints and has the best overall performance is selected from the set of remembered optimal solutions as the master control parameter adjustment result for the next control cycle of the TBM.

[0074] Step S500: Output the optimal combination of main control parameters to the shield tunneling control system to guide the adjustment of shield tunneling parameters in the next control cycle, thereby realizing rolling multi-objective optimization adjustment of the shield tunneling process.

[0075] Understandably, within each control cycle, the system generates main control parameter adjustment commands based on the current tunneling status and multi-objective optimization results, and sends them to the shield tunneling actuator to adjust the propulsion speed, cutterhead rotation speed, and related follow-up parameters. At the start of the next control cycle, updated operating data is collected again, and prediction and optimization are performed again, thus forming a closed-loop control mechanism of "prediction-optimization-execution-feedback". Through this rolling adjustment method, the shield tunneling parameters can dynamically adapt to changes in geological conditions and fluctuations in construction status, achieving continuous optimization of multiple objective indicators such as energy consumption and construction efficiency while ensuring tunneling safety and attitude stability.

[0076] For example, within a certain tunneling control cycle, the system obtains a set of main control parameters that meet safety constraints and have a low tunneling specific energy based on a multi-objective optimization algorithm. This parameter set is then output to the shield control system for real-time adjustment of the cutterhead rotation speed and propulsion speed. After the shield machine completes the tunneling operation of the current cycle under these parameter conditions, the system synchronously collects new operational data and attitude feedback information. Subsequently, at the start of the next control cycle, the system re-performs time-series prediction and multi-objective optimization based on the updated data and generates a new set of main control parameters, achieving rolling parameter updates. Through iterative adjustments over multiple consecutive control cycles, the shield tunneling process is always kept in a controlled and optimized operating state.

[0077] The method of this invention was verified in an engineering application of an earth pressure shield tunneling project in a subway system. The engineering application section mainly consisted of fine sand, medium sand, silt, and alternating sections of silty clay and sand. Rings 950 to 959, which were alternating sections of silty clay and sand, were selected for verification. The optimization range of the cutterhead rotation speed of the PSO-GA hybrid algorithm varied with the ring number as follows: Figure 4 As shown, the optimization range of the tunneling speed of the PSO-GA hybrid algorithm varies with the ring number as follows: Figure 5 As shown in Table 1, the parameter optimization range after calculation by the hybrid algorithm is as follows.

[0078] Table 1. Parameter optimization range based on the PSO-GA hybrid algorithm By combining the Pareto optimality of the particle swarm optimization algorithm with the good convergence of the genetic algorithm, the resulting parameter control range is more reasonable and objective. This allows the invention to effectively adapt to the geological conditions of the tunneling strata and significantly improve the overall tunneling performance of the shield tunneling machine.

[0079] In the above embodiments, by uniformly collecting and preprocessing multi-source operational data such as advance speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters during shield tunneling, a basic data system capable of comprehensively reflecting the changes in shield tunneling conditions and geological formations is constructed. This provides reliable data support for subsequent model analysis and parameter optimization, effectively avoiding the problem of incomplete information caused by single parameters or empirical judgments. Based on this, by introducing a time-series prediction model with advance speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and geological parameters as input features, dynamic prediction of key response indicators such as slurry pressure and tunneling specific energy is achieved. This allows the shield control system to perceive the trend of state changes during tunneling in advance, reducing the risk of parameter adjustment lag caused by sudden geological changes or fluctuations in operating conditions, thereby improving the stability and controllability of the tunneling process. Furthermore, this invention constructs a multi-objective optimization function based on the prediction results, aiming to minimize the prediction deviation of slurry pressure and the specific energy of tunneling. It employs a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithms to perform global search and collaborative optimization of the shield tunneling master control parameters. Under the premise of meeting construction safety constraints, this effectively reduces tunneling energy consumption and optimizes the selection of master control parameter configurations, avoiding the limitations of traditional methods that rely on manual experience or fixed weights. Finally, by outputting the optimal combination of master control parameters to the shield tunneling control system in real time to guide the adjustment of tunneling parameters in the next control cycle, a rolling, multi-objective optimization adjustment mechanism for shield tunneling parameters is realized. This method helps improve the adaptability of the shield tunneling process to complex geological conditions, enhances construction safety and tunneling efficiency, reduces equipment energy consumption and operational risks, and has significant engineering application value and promotional significance.

[0080] In another preferred embodiment based on the above embodiments, such as Figure 6 As shown in the figure, this embodiment provides a multi-objective optimization and adjustment system for the main control parameters of a tunnel boring machine, including: a data acquisition module, a processing module, and an output module.

[0081] Specifically, the acquisition module is configured to collect multi-source operational data during the shield tunneling process and preprocess the multi-source operational data. The multi-source operational data includes at least the propulsion speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters at the corresponding tunneling mileage. The processing module is electrically connected to the acquisition module and is configured to use the propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and geological parameters as input features, and slurry pressure and tunneling specific energy as output response indicators to construct and train a time-series prediction model for the shield tunneling response parameters. The processing module is also configured to construct a multi-objective optimization function based on the prediction results of the time-series prediction model, with the objectives of minimizing the prediction deviation of slurry pressure and minimizing the tunneling specific energy. The output module is electrically connected to the processing module and is configured to perform multi-objective search optimization on the main control parameters of the tunnel boring machine (TBM) based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, to obtain the optimal combination of main control parameters that meets the requirements of safety and energy consumption. The output module is also configured to output the optimal combination of main control parameters to the TBM control system to guide the adjustment of the TBM tunneling parameters in the next control cycle, thereby realizing the rolling multi-objective optimization adjustment of the TBM tunneling process.

[0082] It is understood that the shield tunneling main control parameter multi-objective optimization adjustment method and system in the above embodiments of the present invention have the same beneficial effects, and will not be described in detail here.

[0083] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0084] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0085] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0086] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for multi-objective optimization and adjustment of main control parameters of a tunnel boring machine, characterized in that, include: Collect multi-source operational data during the tunnel boring machine (TBM) excavation process and preprocess the multi-source operational data. The multi-source operational data includes at least the propulsion speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, TBM attitude parameters, and geological parameters at the corresponding tunneling mileage. Using propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume and geological parameters as input features, and slurry pressure and tunneling specific energy as output response indicators, a time-series prediction model for shield tunneling response parameters is constructed and trained. Based on the prediction results of the time series prediction model, a multi-objective optimization function is constructed with the objectives of minimizing the prediction deviation of slurry pressure and minimizing the tunneling specific energy. Based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, the main control parameters of the tunnel boring machine are optimized through multi-objective search to obtain the optimal combination of main control parameters that meets the requirements of safety and energy consumption. The optimal combination of main control parameters is output to the shield tunneling control system to guide the adjustment of shield tunneling parameters in the next control cycle, thereby realizing rolling multi-objective optimization adjustment of the shield tunneling process.

2. The method for multi-objective optimization and adjustment of shield tunneling main control parameters as described in claim 1, characterized in that, When collecting multi-source operational data during the tunnel boring machine (TBM) excavation process and preprocessing the multi-source operational data, the following steps are included: Real-time data collection is performed on the following parameters generated during shield tunneling: propulsion speed, propulsion cylinder pressure, cutterhead rotation speed, cutterhead torque, earth pressure value, soil improvement parameters, slurry injection volume, slurry discharge volume, shield attitude parameters, and geological parameters at the corresponding tunneling mileage. The multi-source operational data from different acquisition devices with different sampling frequencies are timestamped, and the data are unified to a preset sampling period based on the resampling method. A complete synchronous data sequence with time points is obtained based on linear interpolation or spline interpolation methods. The synchronous data sequence is used to identify outliers based on the moving average method, median filtering method, box plot outlier statistics method or density-based anomaly detection algorithm, and outliers exceeding the preset threshold range are removed or replaced. For short-term data gaps that occur during the data collection process, forward padding, backward padding, linear interpolation, or mean estimation based on adjacent time windows are used to fill in the gaps and ensure the continuity of the time series data. The scaling of the operating parameters in each dimension is performed based on the normalization method or the z-score standardization method to make each input feature fall within a uniform numerical range. According to the preset time window length and sliding step size, the preprocessed data is sliced ​​to construct a time series input sample sequence, and the slurry pressure and tunneling specific energy at the corresponding time are used as label data to form a sample set for training the time series prediction model.

3. The method for multi-objective optimization and adjustment of shield tunneling main control parameters as described in claim 2, characterized in that, When constructing and training a time-series prediction model for the shield tunneling response parameters, the following steps are included: Based on the temporal variation characteristics of tunnel boring machine (TBM) data, a temporal prediction model for TBM response parameters is constructed. This model comprises an input layer, a long short-term memory (LSTM) network layer, and a fully connected output layer. The input layer is configured to receive the time-series characteristic sequences of propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume and formation parameters within a preset time window, and the output layer is configured to output the predicted values ​​of slurry pressure and tunneling specific energy for the corresponding time steps. The weight matrix and bias terms of the long short-term memory network layer are initialized using a random initialization method, and the hyperparameters required for training the time series prediction model, such as the initial learning rate, batch size, and number of training epochs, are set. The time series sample set is divided into training set, validation set and test set according to a preset ratio, and the parameters of the time series prediction model are trained, hyperparameters are tuned and the generalization performance of the model is evaluated according to the training set, validation set and test set respectively. The training objective function is based on the mean squared error loss function or the weighted loss function, and the long short-term memory network layer is iteratively trained using the backpropagation algorithm combined with the Adam optimizer. The convergence of the loss function during model training is monitored based on the validation set, and model hyperparameters such as learning rate, hidden layer dimension, and time window length are adjusted according to the validation results. After the model training is completed, the optimal long short-term memory network model parameters are stored and a real-time inference interface is built.

4. The method for multi-objective optimization and adjustment of shield tunneling main control parameters as described in claim 3, characterized in that, When adjusting model hyperparameters such as learning rate, hidden layer dimension, and time window length based on validation results, the following should be included: Based on the validation set sample data, the time series prediction model in the current training round is used to make predictions, and the corresponding validation set loss function value is obtained. Based on the change of the loss function value on the validation set with each training epoch, the convergence state of the loss function during model training is monitored, where: When the loss function does not meet the preset convergence condition, at least one of the hyperparameters of the model, namely the learning rate, hidden layer dimension, and time window length, is adjusted according to the convergence state.

5. The method for multi-objective optimization and adjustment of shield tunneling main control parameters as described in claim 4, characterized in that, When constructing a multi-objective optimization function based on the prediction results of the time-series prediction model, with the objectives of minimizing slurry pressure prediction deviation and minimizing tunneling specific energy, the following functions are included: Using the minimum tunneling specific energy and the minimum difference between the predicted slurry pressure and the corresponding target value as multi-objective optimization objectives, and with the shield attitude parameters meeting the preset allowable range as constraints, a multi-objective optimization control mathematical model for shield tunneling parameters is established by selecting shield tunneling speed, cutterhead torque, cutterhead rotation speed, and total thrust as design variables. Among them, the cutterhead rotation speed and tunneling speed are the main optimization objects, and the shield tunneling parameters are coordinated and optimized. Coordination optimization uses the shield tunneling state parameters in the current control cycle, the predicted values ​​of tunneling specific energy and slurry pressure obtained from the multi-objective optimization control mathematical model, and the initial value ranges of cutterhead rotation speed and tunneling speed, as well as the associated cutterhead torque and total thrust constraints, as input parameters. The cutterhead rotation speed and tunneling speed are determined as the main optimization variables, and the cutterhead torque and total thrust are set as follow-up adjustment variables to establish a coupled and coordinated adjustment relationship between the cutterhead rotation speed and tunneling speed. Based on the main optimization variables, multiple candidate parameter combinations of cutterhead rotation speed and tunneling speed are generated. The corresponding tunneling specific energy evaluation index and slurry pressure prediction deviation are calculated for each candidate parameter combination. Candidate parameter combinations that do not meet the shield attitude parameter constraint range are eliminated. A multi-objective comprehensive evaluation is conducted on the candidate parameter combinations that meet the constraints. A trade-off is made between minimizing the tunneling specific energy and minimizing the slurry pressure prediction deviation, and the optimal combination of cutterhead rotation speed and tunneling speed parameters with the best coordination is selected. The optimal cutterhead rotation speed and tunneling speed are then used as outputs to adjust the corresponding cutterhead torque and total thrust parameters, forming the shield tunneling main control parameter adjustment command for the next control cycle.

6. The method for multi-objective optimization and adjustment of the main control parameters of a tunnel boring machine as described in claim 5, characterized in that, The objective function of the multi-objective optimization model for shield tunneling parameters is set based on Equation 1: Official 1 Where H represents the objective function of the multi-objective optimization model for the main control parameters of the tunnel boring machine; H1 represents the tunneling specific energy; H2 represents the absolute value of the difference between the actual and predicted values ​​of the slurry pressure; F represents the total thrust; v represents the propulsion speed; and n represents the cutterhead rotation speed. R represents the cutterhead torque, and R represents the cutterhead radius of the tunnel boring machine. This indicates the true value of the mud-water pressure. This represents the predicted value of mud and water pressure.

7. The method for multi-objective optimization and adjustment of main control parameters of a tunnel boring machine as described in claim 5, characterized in that, The multi-objective optimization is based on Pareto optimality. It uses the PSO-GA algorithm to generate Pareto front solutions and combines LSTM to predict the tunnel boring machine's excavation status to iteratively optimize the main control parameters.

8. The method for multi-objective optimization and adjustment of main control parameters of a tunnel boring machine as described in claim 1, characterized in that, A hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm is used to perform multi-objective search optimization on the main control parameters of the tunnel boring machine (TBM). The optimal combination of main control parameters that satisfies both safety and energy consumption requirements includes: Set evolutionary parameters and initialize the particle swarm. The evolutionary parameters include the swarm size, learning factor, and number of termination iterations. Initialize the particle swarm and store the historical best solutions corresponding to each particle to form a set of memory best solutions. The optimal parameter codes obtained by each particle in the particle swarm optimization algorithm are used as the operation objects of the genetic algorithm. When the parameter codes can represent better master control parameter values, the chromosome codes in the genetic algorithm are updated and overwritten. Based on the crossover and mutation operations of the genetic algorithm and the parameter update operations of the particle swarm optimization algorithm, the search individuals are evolved, and the memory optimal solution set is updated synchronously during the evolution process, wherein: When the preset heuristic factor search conditions are met, a heuristic strategy is used to optimize the search in the subsequent search process, and the set of memory optimal solutions is continuously updated; at the same time, it is determined whether the preset search termination condition has been reached. When the search termination condition is met, the hybrid evolution search ends, and the optimal combination of master control parameters stored in the particle swarm is taken as the global optimal result of the shield tunneling master control parameters.

9. The method for multi-objective optimization and adjustment of the main control parameters of a tunnel boring machine as described in claim 8, characterized in that, The constraints on the shield tunneling parameters are as follows: x 前 x 后 y 前 y 后 Where, x 前 Indicates the horizontal deviation of the tunnel boring machine's front end, x 后 Indicates the horizontal deviation of the rear end of the tunnel boring machine, y 前 Indicates the vertical deviation of the tunnel boring machine's front end, y 后 This indicates the vertical deviation at the rear end of the tunnel boring machine.

10. A multi-objective optimization adjustment system for the main control parameters of a tunnel boring machine (TBM), employing the multi-objective optimization adjustment method for the main control parameters of a TBM as described in any one of claims 1-9, characterized in that, include: The acquisition module is configured to acquire multi-source operational data during the tunnel boring machine (TBM) excavation process and preprocess the multi-source operational data. The multi-source operational data includes at least the propulsion speed, cutterhead rotation speed, cutterhead torque, earth pressure, soil improvement parameters, slurry injection volume, slurry discharge volume, TBM attitude parameters, and geological parameters at the corresponding tunneling mileage. The processing module, electrically connected to the acquisition module, is configured to use propulsion speed, cutterhead rotation speed, slurry injection volume, slurry discharge volume, and geological parameters as input features, and slurry pressure and tunneling specific energy as output response indicators to construct and train a time-series prediction model for shield tunneling response parameters. The processing module is also configured to construct a multi-objective optimization function based on the prediction results of the time-series prediction model, with the objectives of minimizing slurry pressure prediction deviation and minimizing tunneling specific energy. The output module is electrically connected to the processing module. The output module is configured to perform multi-objective search optimization on the main control parameters of the tunnel boring machine based on a hybrid evolutionary strategy combining particle swarm optimization and genetic algorithm, so as to obtain the optimal combination of main control parameters that meets the requirements of safety and energy consumption. The output module is also configured to output the optimal combination of main control parameters to the shield control system to guide the adjustment of shield tunneling parameters in the next control cycle, thereby realizing rolling multi-objective optimization adjustment of the shield tunneling process.

Images