Meta-Learning-Based Shield Tunneling Optimization Method, Device, Equipment, and Storage Medium

Abstract

The present invention relates to the technical field of shield tunneling. The present invention discloses an optimization method, device, equipment and storage medium for shield tunneling based on meta-learning. By obtaining the tunneling parameters of an atmospheric pressure cutterhead shield machine, the tunneling parameters of a pressurized cutterhead shield machine and geological parameters, a comprehensive task set is generated based on these data, enabling the meta-learning model to learn the relationship between different geological conditions and tunneling parameters; when facing complex and changeable geological conditions, the optimized meta-learning model can map and generate the tunneling parameters required by the atmospheric pressure cutterhead shield machine in the main tunnel based on the tunneling parameters provided by the pressurized cutterhead in the service tunnel, thereby establishing an effective coordination mechanism between the main tunnel and the service tunnel, greatly improving the overall construction efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of shield tunneling, and in particular to a shield tunneling optimization method, device, equipment and storage medium based on meta-learning. Background Art

[0002] In the field of shield tunnel construction, especially in complex environments such as undersea tunnels, tunnel construction faces many severe challenges. The excavation of an undersea tunnel mainly includes the excavation of the main tunnel and the service tunnel. Among them, the main tunnel has a huge amount of work, and its construction usually uses an atmospheric pressure cutter head shield machine, and its construction speed is relatively slow. In contrast, the service tunnel is relatively small in scale, and a pressure cutter head shield machine is used for construction, and its excavation speed is relatively fast.

[0003] However, in the existing construction mode, the excavation work of the main tunnel and the service tunnel is basically carried out independently, which directly leads to the difficulty in coordinating and unifying the excavation progress between the main tunnel and the service tunnel; in the actual construction process, various key information will be continuously obtained during the excavation of the service tunnel, including the detailed situation of geological changes and the valuable experience of adjusting excavation parameters, etc., and these important information have not been fully utilized and learned in the excavation work of the main tunnel; due to the main tunnel being unable to obtain the information of the service tunnel in time, when facing geological condition changes, the main tunnel cannot quickly respond and adjust the excavation parameters, resulting in a serious mismatch between the excavation parameters and the actual geological conditions, which makes the main tunnel frequently encounter problems such as increased cutter head wear, fluctuating excavation speed or even complete stagnation during the excavation process, not only seriously reducing the overall construction efficiency, greatly increasing the equipment maintenance cost, but also greatly extending the project construction period. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a shield tunneling optimization method, device, equipment and storage medium based on meta-learning, which obtains the excavation parameters of the atmospheric pressure cutter head shield machine, the excavation parameters of the pressure cutter head shield machine and geological parameters, generates a comprehensive task set based on these data, so that the meta-learning model can learn the relationship between different geological conditions and excavation parameters; when facing complex and changeable geological conditions, the optimized meta-learning model can map and generate the excavation parameters required by the atmospheric pressure cutter head shield machine in the main tunnel based on the excavation parameters provided by the pressure cutter head in the service tunnel, thereby establishing an effective coordination mechanism between the main tunnel and the service tunnel, and greatly improving the overall construction efficiency.

[0005] The first aspect of the present invention provides an optimization method for shield tunneling based on meta-learning, including: obtaining shield tunneling parameters and geological parameters; generating a training task set and an evaluation task set according to the shield tunneling parameters and geological parameters; constructing a meta-learning model; training the meta-learning model using the training task set to obtain a basic meta-learning model; evaluating the performance of the basic meta-learning model using the evaluation task set to obtain an evaluation result, and optimizing the performance of the basic meta-learning model according to the evaluation result to obtain an optimized meta-learning model; analyzing the mapping relationship between the main tunnel and the service tunnel using the optimized meta-learning model to obtain optimized shield tunneling parameters for an atmospheric pressure cutterhead shield machine.

[0006] Optionally, in the first implementation manner of the first aspect of the present invention, the obtaining of the shield tunneling parameters and geological parameters includes: obtaining the shield tunneling parameters of the atmospheric pressure cutterhead shield machine from the atmospheric pressure cutterhead shield machine; obtaining the shield tunneling parameters of the pressurized cutterhead shield machine from the pressurized cutterhead shield machine; integrating the shield tunneling parameters of the atmospheric pressure cutterhead shield machine and the shield tunneling parameters of the pressurized cutterhead shield machine to obtain the shield tunneling parameters; obtaining the geological parameters of the service tunnel from the service tunnel; obtaining the geological parameters of the main tunnel from the main tunnel; integrating the geological parameters of the service tunnel and the geological parameters of the main tunnel to obtain the geological parameters.

[0007] Optionally, in the second implementation manner of the first aspect of the present invention, the generating of the training task set and the evaluation task set according to the shield tunneling parameters and geological parameters includes: generating a plurality of task frameworks according to the geological parameters; allocating the shield tunneling parameters to different task frameworks to obtain a plurality of training tasks; dividing all the training tasks according to a preset division criterion to obtain a training task set and an evaluation task set.

[0008] Optionally, in the third implementation manner of the first aspect of the present invention, the training of the meta-learning model using the training task set to obtain a basic meta-learning model includes: performing inner-loop training on the meta-learning model using the training task set to obtain a plurality of temporary meta-learning models; performing outer-loop training on all the temporary meta-learning models to obtain a basic meta-learning model.

[0009] Optionally, in the fourth implementation manner of the first aspect of the present invention, the performing of inner-loop training on the meta-learning model using the training task set to obtain a plurality of temporary meta-learning models includes: establishing a plurality of temporary training models according to the meta-learning model; extracting the corresponding training tasks from the temporary training set for each temporary training model; inputting different training tasks into their corresponding temporary training models and performing gradient calculation to obtain a plurality of updated parameters; optimizing their corresponding temporary training models using different updated parameters to obtain a plurality of temporary meta-learning models.

[0010] Optionally, in the fifth implementation manner of the first aspect of the present invention, the outer loop training of all temporary meta-learning models to obtain a basic meta-learning model includes: performing loss analysis on all temporary meta-learning models to obtain a task loss set; calculating the meta-gradient of the initial parameters of the meta-learning model according to the task loss set to obtain gradient data; and performing optimization analysis on the meta-learning model according to a preset optimization algorithm and the gradient data to obtain a basic meta-learning model.

[0011] Optionally, in the sixth implementation manner of the first aspect of the present invention, analyzing the mapping relationship between the main tunnel and the service tunnel by using the optimized meta-learning model to obtain optimized tunneling parameters of the atmospheric pressure cutterhead shield machine includes: obtaining the real-time tunneling parameters of the pressurized cutterhead shield machine; obtaining the real-time geological parameters of the main tunnel and the real-time geological parameters of the service tunnel; inputting the real-time tunneling parameters of the pressurized cutterhead shield machine, the real-time geological parameters of the main tunnel, and the real-time geological parameters of the service tunnel into the optimized meta-learning model, and using the optimized meta-learning model to perform mapping analysis between the main tunnel and the service tunnel to obtain optimized tunneling parameters of the atmospheric pressure cutterhead shield machine.

[0012] The second aspect of the present invention provides a shield tunneling optimization device based on meta-learning, including: a data acquisition module for acquiring the tunneling parameters of the atmospheric pressure cutterhead shield machine, the tunneling parameters of the pressurized cutterhead shield machine, and geological parameters; a task division module for generating a training task set and an evaluation task set according to the tunneling parameters of the atmospheric pressure cutterhead shield machine, the tunneling parameters of the pressurized cutterhead shield machine, and geological parameters; a model construction module for constructing a meta-learning model; a model training module for training the meta-learning model framework by using the training task set to obtain a basic meta-learning model; a model optimization module for performing performance evaluation on the basic meta-learning model by using the evaluation task set to obtain an evaluation result, and performing performance optimization on the basic meta-learning model according to the evaluation result to obtain an optimized meta-learning model; and a parameter mapping module for analyzing the mapping relationship between the main tunnel and the service tunnel by using the optimized meta-learning model to obtain optimized tunneling parameters of the atmospheric pressure cutterhead shield machine.

[0013] The third aspect of the present invention provides a shield tunneling optimization device based on meta-learning. The shield tunneling optimization device based on meta-learning includes: a memory and at least one processor, wherein instructions are stored in the memory; and at least one of the processors calls the instructions in the memory so that the shield tunneling optimization device based on meta-learning executes each step of the shield tunneling optimization method based on meta-learning described in any one of the above.

[0014] The fourth aspect of the present invention provides a computer-readable storage medium, on which instructions are stored, and when the instructions are executed by a processor, each step of the shield tunneling optimization method based on meta-learning described in any one of the above is implemented.

[0015] In the technical solution of the present invention, first, the corresponding shield machine tunneling parameters and geological parameters are obtained from the main tunnel and the service tunnel respectively, and then a training data set and an evaluation data set for training the meta-learning model are generated according to these data; then a meta-learning model framework is constructed, and the training data set and the evaluation data set are used to train and evaluate the meta-learning model framework to generate an optimized meta-learning model; based on the relationship between different geological conditions and tunneling parameters learned by the optimized meta-learning model, a mapping relationship between the main tunnel and the service tunnel can be established, and the tunneling parameters provided by the pressure-balanced cutterhead in the service tunnel can be converted into the tunneling parameters required by the atmospheric-pressure cutterhead shield machine in the tunnel, thereby establishing an effective cooperation mechanism between the main tunnel and the service tunnel, and greatly improving the overall construction efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present invention will become obvious and easy to understand from the description of the embodiments in conjunction with the following drawings, wherein:

[0017] Figure 1 is the first flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0018] Figure 2 is the second flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0019] Figure 3 is the third flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0020] Figure 4 is the fourth flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0021] Figure 5 is the fifth flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0022] Figure 6 is the sixth flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0023] Figure 7 is the seventh flow chart of the shield tunneling optimization method based on meta-learning provided by the embodiment of the present invention;

[0024] Figure 8 is a schematic structural diagram of a shield tunneling optimization device based on meta-learning provided by the embodiment of the present invention;

[0025] Figure 9 is a schematic structural diagram of a shield tunneling optimization device based on meta-learning provided by the embodiment of the present invention. Detailed implementation mode

[0026] The present invention provides a shield tunneling optimization method, system, device and storage medium based on meta-learning. Firstly, the tunneling parameters and geological parameters of the shield machine are respectively obtained from the main tunnel and the service tunnel, and then training data sets and evaluation data sets for training the meta-learning model are generated according to these data; then a meta-learning model framework is constructed, and the meta-learning model framework is trained and evaluated by using the training data sets and evaluation data sets to generate an optimized meta-learning model; based on the relationship between different geological conditions and tunneling parameters learned by the optimized meta-learning model, a mapping relationship between the main tunnel and the service tunnel can be established, and the tunneling parameters provided by the pressurized cutterhead in the service tunnel can be converted into the tunneling parameters required by the atmospheric pressure cutterhead shield machine in the tunnel, thereby establishing an effective coordination mechanism between the main tunnel and the service tunnel and greatly improving the overall construction efficiency.

[0027] The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and drawings of the present invention are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments described herein can be implemented in an order different from that shown or described herein. In addition, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] For the convenience of understanding, the specific process of the embodiment of the present invention will be described below. Please refer to Figure 1 , an embodiment of the shield tunneling optimization method based on meta-learning in the embodiment of the present invention includes:

[0029] 101. Obtain the tunneling parameters and geological parameters of the shield machine;

[0030] In this embodiment, the tunneling parameters and geological parameters of the shield machine are respectively obtained from the main tunnel and the service tunnel; the tunneling parameters of the shield machine include the working data of the shield machine such as the thrust, torque, tunneling efficiency and tool wear of the shield machine, and the tunneling parameters of the shield machine directly reflect the operating state of the shield machine; the geological parameters include geological condition data such as rock hardness, water content, and surrounding rock stress, and the geological parameters reflect the working conditions faced by the shield machine. By obtaining the tunneling parameters and geological parameters, the working conditions of the shield machine under different geological conditions can be intuitively reflected, providing a rich information basis for subsequent model training and optimization.

[0031] 102. Generate a training task set and an evaluation task set based on the tunneling parameters and geological parameters of the shield machine;

[0032] In this embodiment, the tunneling process of the shield machine under different geological conditions can be regarded as different tasks. By reasonably grouping the collected data, a training task set and an evaluation task set are constructed. The training task set is used to provide materials for model training, enabling the model to learn the complex relationships and laws between the tunneling parameters of the shield machine under different geological conditions. The evaluation task set is used to evaluate the performance of the model, promptly detect problems such as overfitting or underfitting of the model, and improve the reliability and accuracy of the model in practical applications. In actual operation, usually 80% of all the data is used to construct the training task set, and the remaining 20% is used to construct the evaluation task set.

[0033] 103. Construct a meta-learning model;

[0034] In this embodiment, a meta-learning model is used as the framework for analyzing the mapping relationship between the main tunnel and the service tunnel. Meta-learning aims to enable the model to quickly adapt to new tasks by training the model on multiple related tasks. Specifically, in this embodiment, the model-agnostic meta-learning (MAML) framework is adopted. The MAML framework optimizes the initial parameters of the model, enabling it to quickly adapt with only a small number of gradient updates when facing new tasks, which matches the working condition of scarce data in undersea tunnel excavation. At the same time, the application of the MAML framework enables the model to learn common features and laws from multiple related tasks, improving the generalization ability of the model and enabling it to be applicable to different types of shield machines and geological conditions.

[0035] 104. Use the training task set to train the meta-learning model to obtain a basic meta-learning model;

[0036] In this embodiment, multiple training tasks are extracted from the training task set to train the meta-learning model, enabling the model to learn the commonalities and differences between different tasks and obtain a basic meta-learning model. The trained basic meta-learning model can quickly adapt between tasks under different geological conditions and has strong parameter prediction ability. It can relatively accurately predict the tunneling parameters of the atmospheric pressure cutterhead shield machine based on the parameters of the pressurized cutterhead shield machine and geological parameters, providing a preliminary basis for parameter adjustment during the tunneling process of the shield machine. At the same time, by learning the experience of multiple tasks, the model can capture some general tunneling laws, which also helps to improve the versatility of the model.

[0037] 105. Use the evaluation task set to evaluate the performance of the basic meta-learning model to obtain an evaluation result, and optimize the performance of the basic meta-learning model according to the evaluation result to obtain an optimized meta-learning model;

[0038] In this embodiment, during the shield tunneling process, the tunneling parameters of the pressure - maintained cutterhead shield machine for verifying the basic meta - learning model are collected, and then these parameters are passed as input data to the basic meta - learning model. The model processes the input data according to the parameter mapping relationship constructed inside it and combines the tunneling laws learned previously under different geological conditions to generate simulated data. By comparing the simulated data with the actual data in the evaluation task set, the possibility of the simulated data in actual application is evaluated. Specifically, the matching degree of the same tunneling parameters (such as tunneling speed, thrust, etc.) between the simulated data and the actual data can be calculated, or the mean square error (MSE) between the simulated data and the actual data can be calculated to measure the prediction accuracy of the model. Through the calculation of these evaluation indicators, a comprehensive evaluation result is obtained, which can quantify the performance of the model on the evaluation task set. A qualified standard for the evaluation result is set in advance, for example, the prediction accuracy reaches a certain threshold (above 90%), the mean square error and the mean absolute error are lower than specific values, etc.

[0039] When the evaluation result meets these qualified standards, it indicates that the model can accurately predict the tunneling parameters of the atmospheric - pressure cutterhead shield machine in actual application and has good performance. At this time, the basic meta - learning model is officially marked as the optimized meta - learning model; the optimized meta - learning model can be officially applied to the actual tunneling construction.

[0040] When the evaluation result shows that the performance of the basic meta - learning model is unqualified, it means that the prediction ability of the model in the actual application scenario does not meet the expected standard and needs to be further optimized. Specifically, a detailed optimization analysis can be carried out on the part with errors. For example, when it is found that the error of the cutterhead torque of the atmospheric - pressure cutterhead shield machine in predicting high - hardness rock geological conditions is large, the input data related to this parameter is focused on. Using methods such as feature analysis and feature importance evaluation, determine which factors have a greater impact on the prediction error, and then calculate the corresponding optimization parameters. The optimization parameters include weight adjustment, learning rate adjustment, hyperparameter adjustment, etc. The optimization parameters provide a clear direction and basis for model improvement, can specifically solve the problems exposed by the model during the evaluation process, and guide the model to develop in a direction that better adapts to the actual shield tunneling engineering requirements. Use the optimization parameters to perform corresponding adjustment operations on the basic meta - learning model to generate a relatively optimized meta - learning model, and evaluate the performance of this meta - learning model again with the evaluation task set. Through multiple iterative adjustments, continuously optimize the performance of the meta - learning model until the performance of the model on the evaluation task set reaches the qualified standard, and finally obtain the optimized meta - learning model.

[0041] 106. Analyze the mapping relationship between the main tunnel and the service tunnel using the optimized meta - learning model to obtain the optimized tunneling parameters of the atmospheric - pressure cutterhead shield machine.

[0042] In this embodiment, an optimized meta - learning model is used to analyze the mapping relationship between the main tunnel and the service tunnel, so as to establish an effective cooperation mechanism between the main tunnel and the service tunnel. During the tunneling process of the main tunnel and the service tunnel, the tunneling parameters and geological parameters of the pressure - balanced shield machine in the service tunnel are obtained in real time, and then the actual working condition data such as the tunneling data and geological information of the service tunnel are converted into the tunneling parameters required for the atmospheric - pressure shield machine in the main tunnel, so as to guide the tunneling construction of the atmospheric - pressure shield machine in the main tunnel. This data - driven mapping analysis enhances the scientificity and rationality of parameter selection during the tunneling process of the main tunnel, effectively improving the tunneling efficiency and ensuring the safety and quality of construction.

[0043] In the embodiment of the present invention, first, the corresponding shield machine tunneling parameters and geological parameters are respectively obtained from the main tunnel and the service tunnel, and then a training data set and an evaluation data set for training the meta - learning model are generated based on these data. Then, a meta - learning model is constructed, and the training data set and the evaluation data set are used to train and evaluate the meta - learning model framework to generate an optimized meta - learning model. Based on the relationship between different geological conditions and tunneling parameters learned by the optimized meta - learning model, a mapping relationship between the main tunnel and the service tunnel can be established, and the tunneling parameters provided by the pressure - balanced shield in the service tunnel are converted into the tunneling parameters required for the atmospheric - pressure shield machine in the tunnel, thus establishing an effective cooperation mechanism between the main tunnel and the service tunnel and greatly improving the overall construction efficiency.

[0044] Please refer to Figure 2 , two embodiments of the shield tunneling optimization method based on meta - learning in the embodiment of the present invention include:

[0045] 201. Obtain the tunneling parameters of the atmospheric - pressure shield machine from the atmospheric - pressure shield machine;

[0046] In this embodiment, the tunneling parameters of the atmospheric - pressure shield machine are key information reflecting its working state and performance. Specifically, the tunneling parameters are collected in real time through various sensors installed on the atmospheric - pressure shield machine. The tunneling parameters include cutterhead torque, cutterhead speed, shield machine thrust, wear condition of the cutterhead, etc. Accurately obtaining the tunneling parameters of the atmospheric - pressure shield machine can grasp the tunneling situation of the main tunnel in real time, providing necessary conditions for subsequent integration and comparative analysis with the data of the service tunnel.

[0047] 202. Obtain the tunneling parameters of the pressure - balanced shield machine from the pressure - balanced shield machine;

[0048] In this embodiment, the tunneling parameters of the pressure - maintained cutterhead shield machine are key information reflecting its working state and performance; similar to the acquisition of tunneling parameters of the atmospheric - pressure cutterhead shield machine, the acquisition of tunneling parameters of the pressure - maintained cutterhead shield machine is also based on various sensors installed on the shield machine; accurately obtaining the tunneling parameters of the pressure - maintained cutterhead shield machine can real - time master the tunneling situation of the service tunnel and provide necessary conditions for subsequent data analysis.

[0049] 203. Integrate the tunneling parameters of the atmospheric - pressure cutterhead shield machine and the pressure - maintained cutterhead shield machine to obtain the tunneling parameters of the shield machine.

[0050] In this embodiment, integrating the tunneling parameters of the two types of shield machines is to comprehensively and systematically understand the working state of the entire shield tunneling project and realize the collaborative utilization of data; by summarizing the tunneling parameters of different types of shield machines under different geological conditions, more potential laws and relationships can be discovered, providing a richer data basis for establishing an accurate meta - learning model.

[0051] 204. Obtain the geological parameters of the service tunnel from the service tunnel.

[0052] In this embodiment, during the tunneling process of the service tunnel, geological parameters are real - time collected through geological exploration equipment and sensors, and these equipment and sensors include instruments for measuring parameters such as rock hardness, water content, and surrounding rock stress; the geological parameters of the service tunnel can provide a reference for the geological environment of the entire shield tunneling project, help analyze the influence of different geological conditions on the tunneling of the shield machine, and provide a basis for predicting the geological conditions of the main tunnel and optimizing the tunneling parameters of the main tunnel.

[0053] 205. Obtain the geological parameters of the main tunnel from the main tunnel.

[0054] In this embodiment, similar to the acquisition of geological parameters of the service tunnel, during the tunneling process of the main tunnel, geological parameters are real - time collected through geological exploration equipment and sensors; the geological parameters of the main tunnel are key factors directly affecting the tunneling of the atmospheric - pressure cutterhead shield machine; in specific construction, because the excavation progress of the service tunnel is faster, more geological data and tunneling parameters can be collected, while the main tunnel has a larger project volume, faces more challenges, and more complex geological conditions; obtaining the geological parameters of the main tunnel can enable the model to more accurately analyze and predict according to the actual geological conditions of the main tunnel.

[0055] 206. Integrate the geological parameters of the service tunnel and the geological parameters of the main tunnel to obtain geological parameters.

[0056] In this embodiment, the geological parameters of the main tunnel and the service tunnel are integrated to construct a complete geological information model, so as to more comprehensively and accurately understand the geological conditions of the entire shield tunneling area. Since the main tunnel and the service tunnel are spatially related, the strata they pass through are to a certain extent similar and continuous. By integrating the geological parameters, this relevance can be fully utilized to improve the accuracy of geological prediction and provide a more reliable geological basis for the tunneling of the shield machine at different positions. Specifically, the geological parameters of the main tunnel and the service tunnel near the same mileage stake or within the same formation depth range can be compared and merged to form a data set containing the complete geological information of this area.

[0057] Please refer to Figure 3 , the three embodiments of the shield tunneling optimization method based on meta-learning in the embodiments of the present invention include:

[0058] 301. Generate multiple task frameworks according to geological parameters;

[0059] In this embodiment, geological parameters are the key factors determining the tunneling environment of the shield machine. Under different geological conditions, the tunneling strategies and parameter requirements of the shield machine vary greatly. By using geological parameters as the basis for task division, task frameworks closely related to the actual geological conditions can be constructed. Specifically, the obtained geological parameters (such as rock hardness, water content, surrounding rock stress, etc.) are analyzed and classified. According to different ranges of rock hardness (such as high hardness, medium hardness, low hardness), combined with different combinations of water content and surrounding rock stress, multiple task frameworks are generated. Each task framework represents a specific combination of geological conditions, enabling the model to fully learn the tunneling characteristics under different geological environments during the training process.

[0060] 302. Allocate the shield tunneling parameters to different task frameworks to obtain multiple training tasks;

[0061] In this embodiment, after constructing the task frameworks, according to the parameters collected during the tunneling of the shield machine (such as the cutter head rotation speed, thrust, propulsion efficiency, torque, etc.) of the shield machine, these parameters are allocated to the corresponding task frameworks according to their corresponding geological conditions to form a complete training task, enabling the model to learn the relationship between the shield tunneling parameters under different geological conditions during the training process, thereby establishing an accurate parameter mapping model.

[0062] 303. Divide all training tasks according to a preset division criterion to obtain a training task set and an evaluation task set;

[0063] In this embodiment, all training tasks are randomly divided according to a certain ratio (e.g., 80% for training and 20% for evaluation), and the constructed training tasks are divided into a training task set and an evaluation task set; during the division process, it is necessary to ensure that the training task set and the evaluation task set are representative and independent in terms of geological condition distribution, tunneling parameter range, etc., to avoid data deviation affecting the model evaluation result, and it is necessary to ensure that both sets can cover various situations and there are no obvious differences. The training task set provides sufficient learning data for the model, and the evaluation task set serves as a validator for the model performance.

[0064] Please refer to Figure 4 , the four embodiments of the shield tunneling optimization method based on meta-learning in the embodiments of the present invention include:

[0065] 401. Use the training task set to perform in-loop training on the meta-learning model to obtain multiple temporary meta-learning models;

[0066] In this embodiment, first, multiple training tasks are extracted from the training task set, and in-loop training of the meta-learning model is carried out for each training task; each training task includes the tunneling parameters of the shield machine under specific geological conditions (such as the parameters of the current ring and the next ring of the pressure-balanced cutterhead shield machine, the parameters of the current ring of the atmospheric-pressure cutterhead shield machine) and geological parameters (such as rock hardness, water content, surrounding rock stress, etc.); the in-loop training aims to improve the analysis ability of the meta-learning model for specific geological conditions and tunneling parameters, so that the model can adjust its own parameters according to different geological conditions and tunneling situations, thereby improving the performance on each task.

[0067] 402. Perform out-of-loop training on all temporary meta-learning models to obtain a basic meta-learning model;

[0068] In this embodiment, the learning experiences of multiple temporary meta-learning models on their respective tasks are integrated, and the initial parameters of the meta-learning model are optimized from a global perspective, so that the meta-learning model can learn the common laws under different geological conditions; the out-of-loop training is a key link for the meta-learning model to be able to quickly transfer learning between multiple tasks, which can avoid the overfitting of the model to a single task, enhance the generalization ability of the model, so that the meta-learning model can quickly adapt to the shield tunneling tasks under different geological conditions. When it encounters new geological conditions, it does not need to be retrained with a large amount of new data, and only needs a small amount of gradient updates to quickly adapt and give reasonable predictions of tunneling parameters.

[0069] Please refer to Figure 5 , the five embodiments of the shield tunneling optimization method based on meta-learning in the embodiments of the present invention include:

[0070] 501. Establish multiple temporary training models according to the meta-learning model;

[0071] In this embodiment, first, based on the parameters of the current meta-learning model, a plurality of independent temporary training models are established. These temporary training models have the same structure and parameters in the initial state. However, during subsequent training processes, they will independently update their parameters according to different training tasks, thus gradually forming their own unique parameter configurations to meet the requirements of different tasks.

[0072] 502. For each temporary training model, extract the corresponding training task from the temporary training set.

[0073] In this embodiment, in order to enable each temporary training model to learn and optimize specifically, corresponding training tasks need to be assigned to it. Each training task includes the tunneling parameters of the shield machine under specific geological conditions (such as the parameters of the current ring and the next ring of the pressure-balanced cutterhead shield machine, and the parameters of the current ring of the atmospheric-pressure cutterhead shield machine) and geological parameters (such as rock hardness, water content, surrounding rock stress, etc.). For example, for a temporary training model, a tunneling task under the geological condition of "high-hardness rock - low water content" may be assigned. The data of this task will be used to train this temporary training model so that it can learn the relationship of tunneling parameters under such specific geological conditions. By reasonably assigning training tasks, it is ensured that each temporary training model can be trained in a specific task scenario, improving the model's targeted processing ability for different geological conditions and tunneling situations.

[0074] 503. Input different training tasks into their corresponding temporary training models and perform gradient calculations to obtain multiple updated parameters.

[0075] In this embodiment, for each temporary training model, use its corresponding training task as the input. The model predicts the tunneling parameters of the atmospheric-pressure cutterhead shield machine based on these inputs. During the training process, the model needs to calculate the loss function between the prediction result and the actual tunneling parameters of the atmospheric-pressure cutterhead shield machine in the training task. The mean squared error (MSE) can be used for the calculation of the loss function. Then, according to the gradient of the loss function with respect to the model parameters, use the gradient descent algorithm to calculate the parameter update amount (i.e., the updated parameter) of the model. The updated parameter is used to optimize the temporary training model.

[0076] 504. Use different updated parameters to optimize their corresponding temporary training models to obtain multiple temporary meta-learning models.

[0077] In this embodiment, apply the updated parameters calculated by each temporary training model to its own model parameters to replace the original parameter values, thereby optimizing the performance of the temporary training model and obtaining a temporary meta-learning model. Each temporary meta-learning model has high performance under specific geological conditions or tunneling scenarios, providing rich local optimal solutions for subsequent outer-loop training.

[0078] Please refer to Figure 6 , in the embodiments of the present invention, the six embodiments of the shield tunneling optimization method based on meta-learning include:

[0079] 601. Perform loss analysis on all temporary meta-learning models to obtain a task loss set;

[0080] In this embodiment, by calculating the loss value of each temporary meta-learning model on its corresponding training task, a task loss set reflecting the performance of each model on different tasks can be obtained. The task loss set will be used as the basis for subsequent calculation of meta-gradients, to grasp the performance distribution of the model on various tasks from a global perspective, so as to improve the overall performance of the model on various tasks. The loss analysis includes mean square error (MSE) or mean absolute error (MAE).

[0081] 602. Calculate the meta-gradients of the initial parameters of the meta-learning model according to the task loss set to obtain gradient data;

[0082] In this embodiment, using the principle of the backpropagation algorithm, the loss value of each temporary meta-learning model is backpropagated to the initial parameters of the meta-learning model, and the meta-gradients (gradient data) corresponding to each initial parameter are calculated; according to the gradient data, it can be known which initial parameters need to be increased or decreased, and the magnitude of the adjustment; for example, if the meta-gradient of an initial parameter related to the cutterhead torque in the gradient data is positive, this means that increasing the value of this parameter may reduce the overall task loss, thereby improving the accuracy of the model's prediction of the cutterhead torque under different geological conditions, and further enhancing the tunneling efficiency and stability of the shield machine in various situations.

[0083] 603. Perform optimization analysis on the meta-learning model according to the preset optimization algorithm and gradient data to obtain a basic meta-learning model;

[0084] In this embodiment, the preset optimization algorithm is the Adam algorithm. The Adam algorithm can dynamically adjust the learning rate according to the meta-gradient values in the gradient data, so as to determine the step size and direction of each parameter update; through multiple iterative updates, the initial parameters of the meta-learning model gradually converge to an optimal value that can perform well on different tasks, thereby obtaining a basic meta-learning model. The basic meta-learning model obtained through optimization analysis has strong generalization ability and can accurately predict the tunneling parameters of an atmospheric pressure cutterhead shield machine in shield tunneling tasks under different geological conditions. In actual engineering, when encountering new geological conditions or tunneling situations, the basic meta-learning model can quickly give reasonable suggestions for optimizing the tunneling parameters of the atmospheric pressure cutterhead shield machine based on the real-time obtained tunneling parameters and geological parameters of the pressure-balanced cutterhead shield machine, to help the operator optimize the tunneling strategy of the shield machine.

[0085] Please refer to Figure 7, the seven embodiments of the shield tunneling optimization method based on meta - learning in the embodiments of the present invention include:

[0086] 701. Obtain the tunneling parameters of the real - time pressure - maintained cutterhead shield machine;

[0087] In this embodiment, various sensors installed on the pressure - maintained cutterhead shield machine are used to obtain the tunneling parameters of the pressure - maintained cutterhead shield machine in real time. In actual construction, since the tunneling efficiency of the pressure - maintained cutterhead shield machine in the service tunnel is faster than that of the normal - pressure cutterhead shield machine in the main tunnel, its tunneling parameters can provide reference for the tunneling work of the main tunnel and provide real - time data support for the main tunnel. By collecting the tunneling parameters of the real - time pressure - maintained cutterhead shield machine, it is prepared for the mapping analysis of the subsequent meta - learning model.

[0088] 702. Obtain the real - time geological parameters of the main tunnel and the real - time geological parameters of the service tunnel;

[0089] In this embodiment, geological parameter monitoring devices are respectively arranged in the main tunnel and the service tunnel to collect the latest real - time geological parameters of the main tunnel and the real - time geological parameters of the service tunnel in real time. Geological parameters are the decisive factors for the tunneling environment of the shield machine. These two sets of data play different roles in the subsequent model analysis. Specifically, in actual construction, since the tunneling speed of the service tunnel is faster, more geological data can be collected. These real - time geological parameters of the service tunnel can be used to assist in judging the geological data of the main tunnel, while the real - time geological parameters of the main tunnel reflect the real geological conditions of the main tunnel and are the most important reference for the operation of the shield machine. By collecting real - time geological data, it is prepared for the mapping analysis of the subsequent meta - learning model.

[0090] 703. Input the tunneling parameters of the real - time pressure - maintained cutterhead shield machine, the real - time geological parameters of the main tunnel, and the real - time geological parameters of the service tunnel into the optimized meta - learning model, and use the optimized meta - learning model to perform mapping analysis between the main tunnel and the service tunnel to obtain the optimized tunneling parameters of the normal - pressure cutterhead shield machine;

[0091] In this embodiment, the trained optimized meta-learning model can establish the mapping relationship between the main tunnel and the service tunnel, so as to convert the working experience during the construction of the service tunnel into relevant data for guiding the construction of the main tunnel. Specifically, after inputting various parameters collected in real time into the optimized meta-learning model, the optimized meta-learning model can quickly analyze and predict the optimized tunneling parameters of the atmospheric pressure cutter head shield machine under the current geological conditions according to the mastered rules. This mapping analysis based on real-time data realizes data sharing and collaborative optimization between the main tunnel and the service tunnel, avoids excessive progress differences from affecting the overall project progress, and improves the overall efficiency of shield tunneling. At the same time, because the optimized meta-learning model has learned the tunneling parameters under different geological conditions, for some safety problems that have occurred in the service tunnel, the model can identify and take targeted preventive measures in advance. The output tunneling parameters of the atmospheric pressure cutter head shield machine have better dynamic adaptability, can better maintain the stability of the surrounding rock, reduce the risk of geological disasters such as cave-ins and water leakage, and improve the ability of the entire shield tunneling project to cope with complex and changeable environments.

[0092] The above describes the shield tunneling optimization method based on meta-learning in the embodiments of the present invention. Next, the shield tunneling optimization system based on meta-learning in the embodiments of the present invention will be described. Please refer to Figure 8 , an embodiment of the shield tunneling optimization system based on meta-learning in the embodiments of the present invention includes:

[0093] A data acquisition module 801, configured to acquire the tunneling parameters of the atmospheric pressure cutter head shield machine, the tunneling parameters of the pressure-balanced cutter head shield machine, and geological parameters;

[0094] A task division module 802, configured to generate a training task set and an evaluation task set according to the tunneling parameters of the atmospheric pressure cutter head shield machine, the tunneling parameters of the pressure-balanced cutter head shield machine, and geological parameters;

[0095] A model construction module 803, configured to construct a meta-learning model;

[0096] A model training module 804, configured to perform model training on the meta-learning model framework by using the training task set to obtain a basic meta-learning model;

[0097] A model optimization module 805, configured to perform performance evaluation on the basic meta-learning model by using the evaluation task set to obtain an evaluation result, and perform performance optimization on the basic meta-learning model according to the evaluation result to obtain an optimized meta-learning model;

[0098] A parameter mapping module 806, configured to analyze the mapping relationship between the main tunnel and the service tunnel by using the optimized meta-learning model to obtain optimized tunneling parameters of the atmospheric pressure cutter head shield machine.

[0099] In this embodiment, first, the data acquisition module 801 obtains the corresponding shield tunneling parameters and geological parameters from the main tunnel and the service tunnel. Then, the task division module 802 generates a training data set and an evaluation data set for training the meta-learning model based on these data. After the model construction module 803 constructs the meta-learning model, the model training module 804 uses the training data set to train the meta-learning model to generate a basic meta-learning model. Then, the model optimization module 805 optimizes the performance of the basic meta-learning model using the evaluation task set to obtain an optimized meta-learning model. Finally, the parameter mapping module 806 analyzes the mapping relationship between the main tunnel and the service tunnel using the optimized meta-learning model, and converts the tunneling parameters provided by the pressurized cutterhead in the service tunnel into the tunneling parameters required by the atmospheric-pressure cutterhead shield machine in the tunnel, thereby establishing an effective coordination mechanism between the main tunnel and the service tunnel and greatly improving the overall construction efficiency.

[0100] Figure 9 FIG. is a schematic structural diagram of a shield tunneling optimization device based on meta-learning provided by an embodiment of the present invention. The shield tunneling optimization device 900 based on meta-learning may vary greatly due to configuration or performance differences, and may include one or more processors (central processing units, CPUs) 910 (for example, one or more processors) and a memory 920, and one or more storage media 930 for storing application programs 933 or data 932 (for example, one or more mass storage devices). Among them, the memory 920 and the storage media 930 may be transient storage or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the shield tunneling optimization device 900 based on meta-learning. Further, the processor 910 may be configured to communicate with the storage media 930 and execute a series of instruction operations in the storage media 930 on the shield tunneling optimization device 900 to implement the steps of the shield tunneling optimization method based on meta-learning provided by the above method embodiments.

[0101] The shield tunneling optimization device 900 based on meta-learning may further include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that Figure 9 The shown structural diagram of the shield tunneling optimization device based on meta-learning does not limit the shield tunneling optimization device based on meta-learning, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0102] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is caused to execute the steps of the shield tunneling optimization method based on meta-learning.

[0103] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, device, or unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0104] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0105] Finally, it should be noted that the above are only preferred examples of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A shield tunneling optimization method based on meta-learning, characterized in that: include: Obtain shield machine excavation parameters and geological parameters; Generate training task set and evaluation task set according to shield machine excavation parameters and geological parameters; Build a meta-learning model; The meta-learning model is trained using the training task set to obtain a basic meta-learning model; the basic meta-learning model is evaluated for performance using the evaluation task set to obtain evaluation results, and the basic meta-learning model is optimized based on the evaluation results to obtain an optimized meta-learning model; The mapping relationship between the main tunnel and the service tunnel is analyzed using the optimized meta-learning model to obtain the optimized tunneling parameters of the atmospheric pressure cutterhead shield machine. The method uses an optimized meta-learning model to analyze the mapping relationship between the main tunnel and the service tunnel to obtain optimized normal pressure cutterhead shield machine excavation parameters, including: obtaining real-time pressure cutterhead shield machine excavation parameters; obtaining real-time main tunnel geological parameters and real-time service tunnel geological parameters; inputting the real-time pressure cutterhead shield machine excavation parameters, real-time main tunnel geological parameters and real-time service tunnel geological parameters into the optimized meta-learning model, and using the optimized meta-learning model to perform mapping analysis between the main tunnel and the service tunnel to obtain optimized normal pressure cutterhead shield machine excavation parameters.

2. The shield tunneling optimization method based on meta-learning according to claim 1 is characterized in that: The obtaining of shield machine excavation parameters and geological parameters includes: Obtain the excavation parameters of the atmospheric pressure cutterhead shield machine from the atmospheric pressure cutterhead shield machine; Obtain the excavation parameters of the shield machine with pressure cutterhead from the shield machine with pressure cutterhead; The tunneling parameters of the shield machine with normal pressure cutterhead and the shield machine with pressure cutterhead are integrated to obtain the tunneling parameters of the shield machine; Obtaining service tunnel geological parameters from the service tunnel; Obtaining the geological parameters of the main tunnel from the main tunnel; The geological parameters of the service tunnel and the main tunnel are integrated to obtain the geological parameters.

3. The shield tunneling optimization method based on meta-learning according to claim 1 is characterized in that: The generating of the training task set and the evaluation task set according to the shield machine excavation parameters and the geological parameters includes: Generate multiple task frameworks based on geological parameters; The tunneling parameters of the shield machine are allocated to different task frameworks to obtain multiple training tasks; all training tasks are divided according to preset division criteria to obtain training task sets and evaluation task sets.

4. The shield tunneling optimization method based on meta-learning according to claim 1 is characterized in that: The meta-learning model is trained using the training task set to obtain a basic meta-learning model, including: Using the training task set to perform inner loop training on the meta-learning model to obtain multiple temporary meta-learning models; All temporary meta-learning models are trained in an outer loop to obtain the base meta-learning model.

5. The shield tunneling optimization method based on meta-learning according to claim 4 is characterized in that: The inner loop training of the meta-learning model using the training task set to obtain multiple temporary meta-learning models includes: Establish multiple temporary training models based on the meta-learning model; For each temporary training model, extract the corresponding training task from the temporary training set; input different training tasks into the corresponding temporary training model, and perform gradient calculation to obtain multiple update parameters; Different update parameters are used to optimize the corresponding temporary training models to obtain multiple temporary meta-learning models.

6. The shield tunneling optimization method based on meta-learning according to claim 4 is characterized in that: The outer loop training is performed on all temporary meta-learning models to obtain a basic meta-learning model, including: Perform loss analysis on all temporary meta-learning models to obtain the task loss set; The meta-gradient of the initial parameters of the meta-learning model is calculated according to the task loss set to obtain the gradient data; the meta-learning model is optimized and analyzed according to the preset optimization algorithm and gradient data to obtain the basic meta-learning model.

7. A shield tunneling optimization device based on meta-learning, characterized in that: include: A data acquisition module is used to obtain the excavation parameters of the normal pressure cutterhead shield machine, the excavation parameters of the pressure cutterhead shield machine and the geological parameters; A task division module is used to generate a training task set and an evaluation task set according to the tunneling parameters of the normal pressure cutterhead shield machine, the tunneling parameters of the pressure cutterhead shield machine and the geological parameters; Model building module, used to build meta-learning models; The model training module is used to train the meta-learning model framework using the training task set to obtain a basic meta-learning model; A model optimization module is used to evaluate the performance of the basic meta-learning model using the evaluation task set to obtain evaluation results, and optimize the performance of the basic meta-learning model according to the evaluation results to obtain an optimized meta-learning model; The parameter mapping module is used to analyze the mapping relationship between the main tunnel and the service tunnel using the optimized meta-learning model to obtain the optimized excavation parameters of the atmospheric pressure cutterhead shield machine; The method uses an optimized meta-learning model to analyze the mapping relationship between the main tunnel and the service tunnel to obtain optimized normal pressure cutterhead shield machine excavation parameters, including: obtaining real-time pressure cutterhead shield machine excavation parameters; obtaining real-time main tunnel geological parameters and real-time service tunnel geological parameters; inputting the real-time pressure cutterhead shield machine excavation parameters, real-time main tunnel geological parameters and real-time service tunnel geological parameters into the optimized meta-learning model, and using the optimized meta-learning model to perform mapping analysis between the main tunnel and the service tunnel to obtain optimized normal pressure cutterhead shield machine excavation parameters.

8. A shield tunneling optimization device based on meta-learning, characterized in that: The shield tunneling optimization device based on meta-learning includes: a memory and at least one processor, wherein instructions are stored in the memory; At least one of the processors calls the instructions in the memory so that the meta-learning-based shield tunneling optimization equipment performs the various steps of the meta-learning-based shield tunneling optimization method as described in any one of claims 1-6.

9. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, the various steps of the shield tunneling optimization method based on meta-learning as described in any one of claims 1-6 are implemented.

Images