Shield tunneling control method and system based on rock-soil particle motion trail

Abstract

The invention relates to the field of shield tunneling, and discloses a shield tunneling control method and system based on a rock-soil particle motion trail, and the method comprises the steps: fusion deployment of various types of high-precision sensors, intelligent anti-noise data collection and preprocessing, precise physical model construction, efficient numerical simulation optimization, and multi-parameter comprehensive detection and correlation analysis. And real-time data-driven adaptive control model construction, control instruction rapid generation and transmission, closed-loop feedback and continuous optimization are realized. According to the method, multiple links including fusion deployment of multiple types of high-precision sensors, intelligent anti-noise processing, physical model construction and the like are covered, all the links have a synergistic effect, data can be accurately obtained, data quality can be guaranteed, a simulation foundation can be provided, comprehensive evaluation, intelligent control and rapid instruction transmission can be achieved, continuous optimization is achieved through closed-loop feedback, and the method is suitable for large-scale popularization and application. And finally, the construction level in all aspects is improved, and complex engineering requirements are met.

Description

Technical Field

[0001] The present invention relates to the field of shield tunneling, and in particular to a shield tunneling control method and system based on rock and soil particle motion trajectories. Background Art

[0002] With the growing demand for urban underground space development, such as subway construction, underground integrated pipe gallery construction and other projects, shield tunneling technology has been widely used as an efficient and safe underground tunnel construction method. It was created to cope with complex underground geological conditions (including soft soil, hard rock, composite strata, etc.), can reduce the disturbance to the surrounding soil, effectively control the construction progress and quality, and adapt to the construction requirements of tunnels of different shapes and sizes. Compared with traditional excavation methods, it has significant advantages in many aspects such as construction safety, efficiency and environmental impact.

[0003] However, in terms of measurement and data acquisition, the existing shield tunneling technology has limited sensor accuracy, insufficient resolution of laser displacement sensors, difficulty in balancing the frame rate and resolution of high-speed cameras, and lens distortion. It is also affected by noise interference such as electromagnetic and particle collisions, and the acquisition range is limited. When establishing and analyzing the model, the physical model has errors in the simplification of rock and soil media and the forces between particles, and the numerical simulation requires large computing resources and is difficult to set boundary conditions. In terms of control strategy, there are problems such as control lag, single parameters, and lack of adaptive capabilities, which make it difficult to adapt to the complex and changeable working conditions of shield tunneling, affecting tunneling efficiency and construction safety.

[0004] Therefore, it is necessary to design a shield tunneling control method and system based on the movement trajectory of rock and soil particles to solve the problems of the existing shield tunneling technology that it is difficult to accurately judge the movement of particles; the acquisition range is limited, the simplified error of the physical model makes the control basis deviate from reality, the simulation results have large deviations, the parameters are single, and the efficiency is low. Summary of the invention

[0005] In view of this, the present invention proposes a shield tunneling control method and system based on the movement trajectory of rock and soil particles, aiming to solve the problems that the existing shield tunneling technology is difficult to accurately judge the movement of particles; the acquisition range is limited, the simplified error of the physical model makes the control basis deviate from the actual situation, the simulation results have large deviations, the parameters are single, and the efficiency is low.

[0006] In one aspect, the present invention provides a shield tunneling control method based on the motion trajectory of rock and soil particles, comprising:

[0007] Design the sensor network topology according to the shield machine structure and sensor distribution;

[0008] Develop intelligent electromagnetic noise monitoring and automatic compensation programs, receive electromagnetic environment detection sensor data, monitor electromagnetic noise intensity and frequency in real time and collect them regularly;

[0009] Collect and analyze scanning images of rock and soil particles, conduct multi-condition combination tests, compare field data with indoor test data, and build and calibrate a physical model of rock and soil particle motion trajectory that takes multiple factors into consideration;

[0010] Introducing parallel frameworks and developing adaptive meshing algorithms, building data acquisition and transmission systems and integrating test optimization to jointly improve the efficiency and accuracy of numerical simulations;

[0011] Design sensor interfaces and connect them, build a data association analysis platform to perform data standardization, association analysis and dimensionality reduction, and build and optimize the shield tunneling status assessment model;

[0012] Determine the network architecture combination and improve data processing, make training adjustments, and build an adaptive control model that can meet the needs of shield tunneling control;

[0013] Optimize the algorithm flow and design multi-threading and caching mechanisms, and ensure that control instructions can be quickly and accurately generated and transmitted to the shield machine actuator through network-related configuration, redundant design, self-healing technology, etc.;

[0014] Develop feedback data collection plans, maintain equipment networks and safely store data, organize and analyze data and quantitatively evaluate and diagnose causes, optimize and improve issues such as sensor layout, data processing algorithms, physical models and control models, and form a closed-loop feedback and continuous optimization system for shield tunneling control.

[0015] Furthermore, the sensor network topology is designed according to the shield machine structure and sensor distribution conditions, including:

[0016] Sensor network construction and debugging: Design the sensor network topology according to the structure of the shield machine and the distribution of sensors, set up switches and routers in the network and plan the data transmission path, determine the network communication protocol and data storage format, debug the sensor network, test the network connection between each sensor node and the data processing center, perform data fusion test, transmit the data collected by different sensors to the data processing center, run the data fusion algorithm, check the algorithm's integration effect on different sensor data, if the data fusion effect is not good, check whether the algorithm parameter settings are reasonable, and adjust the algorithm parameters in time according to the test results until the data fusion effect meets the requirements.

[0017] Furthermore, the development of intelligent electromagnetic noise monitoring and automatic compensation program receives electromagnetic environment detection sensor data, monitors electromagnetic noise intensity and frequency in real time and collects them regularly, including:

[0018] Intelligent electromagnetic noise monitoring and automatic compensation, develop intelligent electromagnetic noise monitoring and automatic compensation program, receive data from electromagnetic environment detection sensors, monitor the intensity and frequency of electromagnetic noise in real time, collect electromagnetic noise data at regular intervals, perform fast Fourier transform analysis on the collected data, obtain the spectrum information of electromagnetic noise, compare the current electromagnetic noise spectrum with the preset threshold spectrum, and if it is found that the electromagnetic noise intensity in a certain frequency band exceeds the preset threshold, calculate the parameter values ​​that need to be adjusted for the electromagnetic filter according to the pre-set adjustment strategy, realize automatic optimization of the electromagnetic filter parameters, and record the historical records of electromagnetic filter parameter adjustments.

[0019] Furthermore, the multi-condition combination test is carried out, the field collected data is compared with the indoor test data, and a physical model of the movement trajectory of rock and soil particles considering multiple factors is constructed and calibrated, including:

[0020] Indoor test: According to the geological conditions and rock and soil particle characteristics that may be encountered in shield construction, a series of indoor test schemes are designed. The test schemes include different combinations of water content, particle surface roughness and different stress states. High-precision mechanical sensors, test containers, loading devices, simulated rock and soil particles and other test equipment and materials are prepared. The simulated rock and soil particles are placed in the test container according to the predetermined particle grading and stacking method. The water content and surface roughness of the particles are adjusted, and then different stress states are applied through the loading device. The interaction force between particles is measured in real time using mechanical sensors, and data such as the magnitude, direction and action time of the force are recorded. Displacement sensors are used to monitor the displacement changes of particles, and the movement and deformation of particles under different conditions are observed. Each test condition is tested repeatedly, and the data of each test is recorded and stored in detail.

[0021] Field data collection and comparative analysis: Install sensors on the shield machine that can collect data on the interaction between rock and soil particles. The sensors are installed on key locations such as the cutter head, shield, and slag conveying pipeline of the shield machine to directly or indirectly obtain information on the interaction between rock and soil particles during the actual excavation process. Continuously collect data on the interaction between rock and soil particles, record the data collection time, collection location, and shield machine operating parameters, establish field data archives, compare and analyze field data with indoor test data, and compare indoor test data and field data under the same geological conditions to analyze the differences and similarities between the two.

[0022] Physical model construction and calibration: Based on indoor tests and field data, a physical model of the motion trajectory of rock and soil particles is established that takes into account factors such as particle shape, surface roughness, and water content. Micromechanics theory and multi-scale analysis methods are introduced to link the microstructural characteristics of rock and soil particles with their macroscopic motion behavior, and to determine key parameters such as the elastic modulus, Poisson's ratio, friction coefficient, and cohesion coefficient of the particles in the model. The model parameters are calibrated and optimized through repeated test data fitting and field data verification. The optimization algorithm is used to compare the model calculation results with the actual test data and field data, and the model parameters are continuously adjusted so that the model can accurately reflect the interaction relationship between rock and soil particles under different geological conditions. The historical records of each parameter adjustment and the corresponding model evaluation indicators are recorded.

[0023] Furthermore, the introduction of parallel framework and development of adaptive meshing algorithm, construction of data acquisition and transmission system and integrated test optimization, jointly improve the efficiency and accuracy of numerical simulation, including:

[0024] The numerical simulation algorithm is optimized by using the discrete element method. Based on the MPI framework, the entire simulation area is divided into multiple sub-areas. The particle motion calculation tasks of each sub-area are assigned to different computing nodes. The large-scale rock and soil particle system is evenly divided into several sub-task areas according to the spatial position. The sub-tasks are dynamically assigned according to the performance characteristics of each computing node, and task scheduling and management are performed. The execution progress of each sub-task is monitored, and the data transmission and synchronization operations between nodes are coordinated. The computing status information of each sub-task is collected in real time, and the task allocation and data transmission strategies are dynamically adjusted according to this information. According to the estimated particle distribution of the shield excavation face and the geometric characteristics of the overall simulation area, a coarser grid is used for division, and a particle motion state monitoring mechanism is established. By real-time analysis of indicators such as particle velocity, acceleration, displacement changes, and particle concentration changes in local areas, the intensity of particle motion and regional feature changes are judged. According to the monitoring results, an adaptive grid refinement algorithm is developed. When an area that needs to be refined is detected, the algorithm automatically inserts finer grid units in the area and sets refinement stop conditions. When the grid is refined to a certain extent, even if the particle motion is still violent, as long as certain accuracy requirements are met, further refinement will be stopped;

[0025] Real-time boundary condition detection, development of data acquisition hardware circuits and software programs, to achieve real-time acquisition and preprocessing of sensor data. The hardware circuit is responsible for conditioning, amplifying, filtering and other processing of the analog or digital signals output by the sensor, and the software program controls the data acquisition card to collect sensor data according to the predetermined sampling frequency and data format, and perform preliminary data verification and error processing; build a high-speed data transmission network to transmit the collected sensor data to the numerical simulation system, using the optical fiber communication network as the main transmission medium. In the network architecture design, a star topology is adopted, and network switches, routers and other equipment are configured to set the network protocol and data transmission rate; integrate sensors, data acquisition systems and numerical simulation systems, conduct system joint debugging tests, simulate the operation of shield machines under different working conditions, check whether the boundary condition monitoring system can accurately collect and transmit data, whether the numerical simulation system can correctly receive and use these data for simulation calculations, and perform performance tests and optimizations on the system in terms of real-time data acquisition, data transmission delay and packet loss rate, and numerical simulation calculation efficiency and accuracy, and continuously optimize system performance by adjusting hardware equipment parameters and software algorithms.

[0026] Furthermore, the design of the sensor interface and connection, the construction of a data association analysis platform for data standardization, association analysis and dimensionality reduction, and the construction and optimization of a shield tunneling state assessment model include:

[0027] Multi-parameter detection, design corresponding interface circuits and communication protocols for different types of sensors. The interface circuit can convert the signal output by the sensor into a format that can be recognized and processed by the shield machine control system; reserve the sensor interface position in the shield machine control system, and connect each sensor to the control system according to the design requirements. After the connection is completed, test each sensor individually to check whether it can work normally and send data to the control system; use timestamp technology to mark the data of each sensor, and align and synchronize the data according to the timestamp during the data processing process, establish a data accuracy verification mechanism, and perform real-time verification of the collected multi-parameter data;

[0028] Data association analysis, development of data standardization processing program, dimensionless processing of each parameter data, normalization processing is used for parameters with clear physical meaning and magnitude range such as thrust pressure and cutter head torque, and their data are mapped to the interval [0,1]. For parameters with large data magnitude differences such as slag discharge, standardization processing is used to make their data obey the standard normal distribution; the correlation analysis algorithm in big data analysis technology is used to calculate the correlation coefficient between each parameter and the motion trajectory of rock and soil particles, and the principal component analysis algorithm is used to reduce the dimension of multi-parameter data, converting multiple related original parameters into a few unrelated principal components, which can retain most of the information of the original data and determine the number of principal components. And the contribution rate of each principal component, select the principal component with larger contribution rate as the input variable for model establishment; according to the results of correlation analysis and principal component analysis, a multi-parameter coupled shield tunneling state evaluation model is established, the screened principal components or key parameters are used as the input variables of the model, and the evaluation index of shield tunneling state is used as the output variable of the model, and the model is trained through historical shield tunneling data; in the process of model training, the cross-validation method is used to evaluate and optimize the performance of the model, and according to the evaluation results, the structural parameters of the model or the parameters of the optimization algorithm are adjusted to continuously improve the accuracy and generalization ability of the model. After multiple training and optimization, the final shield tunneling state evaluation model is obtained, and it is deployed in the actual shield tunneling control system.

[0029] Furthermore, the determination of the network architecture combination and the improvement of data processing, the training adjustment, and the construction of an adaptive control model that can meet the shield tunneling control requirements include:

[0030] Model framework selection and design, analyze the advantages and disadvantages of different architectures in feature extraction, sequence processing and model complexity, combine the real-time and accuracy requirements of shield tunneling control for data processing, determine the most suitable network architecture combination or single architecture, determine the range of the number of layers and nodes when designing hidden layers, consider the impact of different inter-layer connection methods on the model learning ability, compare the performance of the model on the training set and validation set under different connection methods through small-scale experiments, and select the optimal connection strategy; when determining the loss function, explore the construction of a custom loss function based on the characteristics of the control instructions and actual engineering needs;

[0031] Data collection and preprocessing: collect data from the existing shield engineering case database, formulate a supplementary monitoring data plan on site, and clarify the data types and frequencies that need to be collected under different geological conditions, shield tunneling stages, and different construction processes; perform data cleaning, identify and process outliers based on statistical methods, establish a data quality assessment model, comprehensively consider the integrity, consistency, and accuracy of the data, score each data record, and screen out high-quality data for subsequent processing based on the scoring results; use mean, median filling or model-based filling methods to further analyze the impact of different filling methods on model training results, and determine the optimal filling strategy under different data types and missing ratios through comparative experiments; perform data annotation, classify and annotate according to the shield tunneling status, annotate the subdivision features under different states, and use time series segmentation methods to ensure that the training set data precedes the validation set and test set in time to simulate the time-sequential learning process of the model in actual applications;

[0032] Model training and optimization: use the back-propagation algorithm to train the model, set the learning rate, adjust the momentum term parameters, and adopt an adaptive learning rate adjustment strategy to dynamically adjust the learning rate according to the change in loss during model training. Use the early stopping method to monitor the loss value or evaluation index on the validation set and combine it with the prediction stability index of the model on the validation set. When the loss value no longer increases and the prediction stability index also meets certain requirements, it is determined that the model performance no longer improves, and training is stopped. Model integration technology is used to weighted average or vote on excellent models at different stages of multiple training processes. After the model training is completed, the sensitivity analysis method is used to evaluate the importance of the model's input features and determine the key features that have a greater impact on the generation of control instructions.

[0033] Furthermore, the optimization algorithm process is designed with multi-threading and caching mechanisms, and network-related configuration, redundant design, self-healing technology, etc. are used to ensure that control instructions can be quickly and accurately generated and transmitted to the shield machine actuator, including:

[0034] System software development uses multi-threaded processing technology to reasonably allocate thread resources according to different task modules in the data processing process, design thread pool management strategies, set higher priorities for data verification threads to ensure that the accuracy and integrity of the data are guaranteed at the first time; for model input data preparation threads, dynamically adjust their priorities and resource allocation according to the real-time data needs of the model; set a first-level cache at the sensor data receiving end to temporarily store the newly collected data, set a second-level cache at the data processing front end to cache the data that has been preliminarily verified and classified, and set a third-level cache at the model input end to ensure that the model can quickly obtain processed data for instruction generation; use algorithm optimization technology, loop unrolling, code vectorization, etc. to reduce the computational overhead in the data processing process, optimize the process of model generation control instructions, and use model compression technology to reduce the computational workload and memory usage of the model;

[0035] Network construction and system integration, use network performance testing tools to test network delays regularly, establish a network delay monitoring model, adjust network configuration in advance or take network optimization measures based on model prediction results to cope with possible network delay fluctuations, configure redundant equipment at key network nodes, and automatically switch to backup equipment when the main equipment fails to ensure uninterrupted operation of the network, establish a network equipment failure monitoring and early warning system, monitor the operating status of network equipment in real time, discover potential fault hazards in advance, and perform maintenance and repair in a timely manner; use the SDN controller to monitor and analyze network traffic in real time, and automatically adjust the network topology and re-plan the data transmission path when a network failure or congestion is detected to achieve rapid self-healing of the network; during the connection between the control instruction generation system and the shield machine actuator, monitor the connection status, monitor the stability of the connection and the correctness of the data transmission in real time, and use the heartbeat mechanism to send test signals regularly to ensure that both parties can perceive each other's status in a timely manner. Once an abnormal connection is found, immediately initiate emergency measures such as reconnection or switching to the backup connection line.

[0036] Furthermore, the aforementioned formulation of feedback data collection plan, maintenance of equipment network and safe storage of data, collation and analysis of data and quantitative evaluation of diagnosis causes, optimization and improvement of sensor layout, data processing algorithm, physical model and control model, etc., form a closed-loop feedback and continuous optimization system for shield tunneling control, including:

[0037] Data collection and transmission: formulate feedback data collection plans according to different stages and control requirements of shield tunneling, clarify the characteristics of rock and soil particle movement trajectories and key operating status parameters of shield machines that need to be focused on at each stage, determine the time interval setting principles for data collection, formulate multi-sensor collection plans for key data points, ensure data reliability and integrity, plan the maintenance cycle and calibration plan of data collection equipment, ensure that the equipment operates stably throughout the tunneling process, and that data collection accuracy meets requirements; regularly inspect data collection equipment to check the installation firmness of sensors, and check whether they are loose, displaced or damaged; clean and maintain the measuring elements of sensors to prevent dust, mud and other impurities from affecting measurement accuracy. Conduct performance tests on the sensor network and use network monitoring tools to monitor the network status in real time. If the network delay is too high or the packet loss is serious, promptly investigate the cause of the network failure and take appropriate repair measures. Use data encryption technology to encrypt the feedback data and compress the transmitted data to improve data transmission efficiency and reduce network bandwidth usage. Establish a feedback data repository in the data processing center and classify and store it according to data type, collection time, shield machine location and other information. Use a database management system to manage the data and regularly back up the data repository to prevent data loss. The backup data is stored in an off-site server or storage medium to deal with possible hardware failures, natural disasters, etc.

[0038] Data analysis and evaluation: after receiving the feedback data, the data processing center will check the integrity of the data to see if there are any missing data, data format errors, etc. For missing data, interpolation methods are used to supplement the data according to the time series characteristics and correlation of the data. For data format errors, data format conversion and correction are performed; the data is filtered to remove noise interference in the data. Digital filtering technology is used to smooth the data for rock and soil particle motion trajectory data to extract the true trend of particle motion. Corresponding filtering is also performed on the shield machine operation parameter data to accurately reflect the actual operating status of the equipment; the particle motion trajectory deviation rate is calculated to determine the reference trajectory, and mathematical methods are used to calculate the difference between the actual trajectory and the reference trajectory. Deviation: Analyze the distribution of deviation rate in different time stages and different spatial regions, find out the time periods and regions with large deviations, provide a basis for subsequent cause analysis, calculate the variance, standard deviation and other statistical quantities of parameters to measure the degree of fluctuation of the stability index of shield machine operation parameters; establish a comprehensive evaluation model, weightedly integrate the particle motion trajectory deviation rate, shield machine operation parameter stability index and other related indicators, and obtain a comprehensive evaluation value that fully reflects the control effect. Determine the weight coefficient of each indicator through the hierarchical analysis method. According to the comprehensive evaluation results and the analysis of various quantitative indicators, conduct in-depth diagnosis of the reasons for poor control effect. Use the fault tree analysis method to gradually check possible causes and determine the key factors, providing a clear direction for subsequent optimization and adjustment;

[0039] Optimization and adjustment: when the regional rock and soil particle movement trajectory monitoring data is inaccurate, review the geological conditions of the area and re-evaluate the sensor layout plan and simulation analysis; when there are problems with the data processing algorithm, review the optimization algorithm and conduct performance tests; when the physical model or numerical simulation parameters are unreasonable, conduct sensitivity analysis and calibration of the physical model parameters and numerical simulation parameter adjustment experiments; when the control model has poor prediction effect, analyze the control model input and output data and optimize the data acquisition preprocessing link or increase the amount of training data and adjust the model structure parameters and test and verify.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows: through the fusion deployment of multiple types of high-precision sensors, geological information and particle motion data can be accurately acquired; intelligent anti-noise data acquisition and preprocessing ensure data quality; accurate physical model construction provides a basis for simulation; efficient numerical simulation optimization improves computing efficiency and accuracy; multi-parameter comprehensive monitoring and correlation analysis realize comprehensive status assessment; real-time data-driven adaptive control model construction achieves intelligent control; control instructions are quickly generated and transmitted to ensure timeliness; closed-loop feedback and continuous optimization continuously improve performance and reliability, thereby effectively improving the safety, accuracy, adaptability and intelligence of shield tunneling construction, reducing construction risks and costs, improving construction efficiency and quality, and adapting to complex and changeable geological conditions and engineering requirements.

[0041] On the other hand, the present application also provides a shield tunneling control system based on the movement trajectory of rock and soil particles, including:

[0042] The shield machine sensor network construction and debugging module designs the sensor network topology according to the shield machine structure and sensor distribution;

[0043] Intelligent anti-noise data acquisition and preprocessing module, planning electromagnetic shielding and filtering, detecting electromagnetic environment, intelligent monitoring and compensation of electromagnetic noise, and transmitting data to the processing center for format conversion, verification, denoising, integrity check and outlier processing;

[0044] The precise physical model building module conducts multi-condition combination tests, compares the field data with the indoor test data, and builds and calibrates the physical model of the movement trajectory of rock and soil particles that takes into account multiple factors;

[0045] Efficient numerical simulation optimization module, introducing parallel framework and developing adaptive meshing algorithm, building data acquisition and transmission system and integrating test optimization, jointly improving the efficiency and accuracy of numerical simulation;

[0046] Multi-parameter comprehensive monitoring and correlation analysis module, design sensor interfaces and connect them, build a data correlation analysis platform for data standardization, correlation analysis and dimensionality reduction, and build and optimize the shield tunneling status assessment model;

[0047] Real-time data-driven adaptive control model building module determines the network architecture combination and improves data processing, performs training adjustments, and builds an adaptive control model that can meet the needs of shield tunneling control;

[0048] The control instruction rapid generation and transmission module optimizes the algorithm flow and designs multi-threading and caching mechanisms. Through network-related configuration, redundant design, self-healing technology, etc., it ensures that the control instructions can be quickly and accurately generated and transmitted to the shield machine actuator;

[0049] The closed-loop feedback and continuous optimization module formulates feedback data collection plans, maintains equipment networks and safely stores data, organizes and analyzes data and conducts quantitative evaluation and diagnosis of causes. It optimizes and improves issues such as sensor layout, data processing algorithms, physical models and control models, and forms a closed-loop feedback and continuous optimization system for shield tunneling control.

[0050] It can be understood that the above-mentioned shield tunneling control system and method based on the movement trajectory of rock and soil particles have the same beneficial effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0052] Figure 1 A flowchart of a shield tunneling control method based on rock and soil particle motion trajectory provided by an embodiment of the present invention;

[0053] Figure 2 A functional block diagram of a shield tunneling control system based on the motion trajectory of rock and soil particles provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0055] Reference Figure 1 In some embodiments of the present application, a shield tunneling control method based on rock and soil particle motion trajectory includes:

[0056] Step S100: Designing a sensor network topology according to the shield machine structure and sensor distribution conditions.

[0057] Step S200: Develop an intelligent electromagnetic noise monitoring and automatic compensation program, receive electromagnetic environment detection sensor data, monitor the electromagnetic noise intensity and frequency in real time and collect them regularly.

[0058] Step S300: Conduct a multi-condition combination test, compare the field collected data with the indoor test data, and construct and calibrate a physical model of the rock and soil particle motion trajectory that takes multiple factors into consideration.

[0059] Step S400: Introduce a parallel framework and develop an adaptive meshing algorithm, build a data acquisition and transmission system, and integrate test optimization to jointly improve the efficiency and accuracy of numerical simulation.

[0060] Step S500: Design the sensor interface and connect it, build a data association analysis platform to perform data standardization, association analysis and dimensionality reduction processing, and build and optimize the shield tunneling status assessment model.

[0061] Step S600: Determine the network architecture combination and improve data processing, perform training adjustments, and build an adaptive control model that can meet the shield tunneling control requirements.

[0062] Step S700: Optimize the algorithm flow and design multi-threading and caching mechanisms, and ensure that control instructions can be quickly and accurately generated and transmitted to the shield machine actuator through network-related configuration, redundant design, self-healing technology, etc.

[0063] Step S800: Formulate a feedback data collection plan, maintain the equipment network and safely store data, organize and analyze the data and quantify and evaluate the causes of diagnosis, optimize and improve issues such as sensor layout, data processing algorithms, physical models and control models, and form a closed-loop feedback and continuous optimization system for shield tunneling control.

[0064] Specifically, according to the structure of the shield machine and the distribution of sensors, the sensor network topology is designed, switches and routers are set up in the network, and the data transmission path is planned. The network communication protocol is determined, such as the industrial Ethernet protocol or other special communication protocols suitable for the shield construction environment, to ensure the stability and real-time performance of data transmission. The data transmission rate is set, and the bandwidth is reasonably allocated according to the amount of sensor data and the requirements for real-time performance. For example, for high-speed camera data with a large amount of data, a higher bandwidth is allocated to ensure that the image data can be transmitted in time, while for sensor data with a small amount of data, bandwidth is appropriately allocated to avoid waste of network resources. The data storage format is determined, and a standardized data format, such as CSV format or binary format, is used to facilitate data processing and analysis. The sensor network is debugged, the network connection between each sensor node and the data processing center is tested, and a data fusion test is performed to transmit the data collected by different sensors to the data processing center. The data fusion algorithm is run to check the integration effect of the algorithm on different sensor data. If the data fusion effect is not good, check whether the algorithm parameter settings are reasonable, and adjust the algorithm parameters in time according to the test results until the data fusion effect meets the requirements.

[0065] It is understandable that installing sensors and setting up a debugging network ensures stable and efficient data collection and transmission.

[0066] Specifically, develop an intelligent electromagnetic noise monitoring and automatic compensation program, receive data from electromagnetic environment detection sensors, monitor the intensity and frequency of electromagnetic noise in real time, collect electromagnetic noise data at regular intervals, perform fast Fourier transform analysis on the collected data, obtain the spectrum information of the electromagnetic noise, compare the current electromagnetic noise spectrum with the preset threshold spectrum, and if it is found that the electromagnetic noise intensity in a certain frequency band exceeds the preset threshold, calculate the parameter values ​​that need to be adjusted for the electromagnetic filter according to the pre-set adjustment strategy, realize automatic optimization of the electromagnetic filter parameters, and record the historical records of the electromagnetic filter parameter adjustments.

[0067] Specifically, the test design team designed a series of indoor test plans based on the geological conditions and rock and soil particle characteristics that may be encountered during shield construction. The test plan should include combinations of different moisture contents, particle surface roughness, and different stress states, such as setting different moisture levels (from dry to saturated moisture content), using particle samples with different roughness treatments (changing the particle surface roughness by grinding, coating, etc.), and applying different pressure and shear conditions. Prepare test equipment and materials, including high-precision mechanical sensors, test containers, loading devices, simulated rock and soil particles, etc. Mechanical sensors should have high sensitivity and high precision, and be able to accurately measure the interaction forces such as friction and cohesion between particles. If a micro-force sensor is used, its measurement accuracy can reach 0.01N. The test container should be able to simulate the actual stacking environment of rock and soil particles, such as using a cylindrical or square container, and setting adjustable boundary conditions inside the container to simulate different constraints. Perform the test according to the test plan. Place the simulated rock and soil particles in the test container according to the predetermined particle grading and stacking method, adjust the moisture content and surface roughness of the particles, and then apply different stress states through the loading device. During the test, mechanical sensors are used to measure the interaction force between particles in real time, and data such as the magnitude, direction and duration of the force are recorded. At the same time, displacement sensors are used to monitor the displacement changes of particles and observe the movement and deformation of particles under different conditions. Each test condition is tested repeatedly to ensure the reliability and accuracy of the test data. The data of each test should be recorded and stored in detail, including information such as test parameters, test process data and test results.

[0068] Specifically, at the shield construction site, the sensor installation and data acquisition team installs sensors on the shield machine that can collect data on the interaction between rock and soil particles, such as pressure sensors and strain sensors. These sensors should be installed in key locations such as the cutterhead, shield, and slag conveying pipeline of the shield machine, and can directly or indirectly obtain information on the interaction between rock and soil particles during the actual excavation process. Establish a field data acquisition system to ensure that sensor data can be transmitted to the data processing center in real time and stably. The data acquisition system should have functions such as data caching, data compression, and data encryption to improve data transmission efficiency and ensure data security. During the shield excavation process, continuously collect data on the interaction between rock and soil particles, record the data collection time, collection location, and shield machine operating parameters (such as propulsion speed, cutterhead speed, etc.), and establish a field data archive. Compare and analyze the field data with the indoor test data. Data analysts use data analysis software to compare the indoor test data and field data under the same geological conditions, and analyze the differences and similarities between the two. For example, compare the measured values ​​of inter-particle friction and cohesion to observe the deviation between the inter-particle interaction force and the indoor test results in the actual engineering environment. Through comparative analysis, the differences between the indoor test model and the actual engineering situation are found out, providing a basis for the establishment of the physical model and parameter calibration.

[0069] Specifically, the theoretical modeling team established a physical model of the motion trajectory of rock and soil particles based on indoor tests and field data, taking into account factors such as particle shape, surface roughness, and water content. In the process of model construction, micromechanics theory and multi-scale analysis methods were introduced to link the microstructural characteristics of rock and soil particles with macroscopic motion behavior. For example, the discrete element method (DEM) was used to establish a particle model, and each rock and soil particle was regarded as an independent unit. The interaction between particles, such as contact force, friction force, and cohesion force, was considered, and the motion state of the particles was calculated by Newton's second law. The key parameters in the model, such as the elastic modulus, Poisson's ratio, friction coefficient, and cohesion coefficient of the particles, were determined. The initial values ​​of these parameters can be estimated based on indoor test data and related theoretical research. Then, the model parameters are calibrated and optimized through repeated test data fitting and field data verification. Optimization algorithms, such as the least squares method and genetic algorithm, are used to compare the model calculation results with the actual test data and field data, and the model parameters are continuously adjusted so that the model can accurately reflect the interaction relationship between rock and soil particles under different geological conditions and improve the accuracy of the prediction of particle motion trajectory. During the model calibration process, the historical records of each parameter adjustment and the corresponding model evaluation indicators (such as root mean square error, correlation coefficient, etc.) are recorded for subsequent analysis and model improvement.

[0070] It is understandable that by conducting multi-condition combination tests and comparing field and indoor test data to construct and calibrate a physical model of the movement trajectory of rock and soil particles that takes into account multiple factors, we can gain deep insights into the micro and macro characteristics of rock and soil particles and accurately determine the interaction relationship between particles, thereby providing an extremely accurate theoretical basis for shield tunneling control. It can predict in advance the movement and change trends of rock and soil particles under different geological conditions, effectively optimize tunneling strategies, and significantly enhance the ability to respond to complex geological conditions during shield tunneling. It greatly improves the safety and stability of construction, and effectively ensures the smooth progress of the project and accurately achieves the expected results.

[0071] Specifically, according to the selected numerical simulation algorithm, a parallel computing framework is selected for integration. For example, if the discrete element method is used, the entire simulation area can be divided into multiple sub-areas based on the MPI framework, and the particle motion calculation tasks of each sub-area are assigned to different computing nodes. Design a reasonable task decomposition strategy to ensure data interaction and load balancing between subtasks. For example, a large-scale rock and soil particle system is evenly divided into several sub-task areas according to spatial position, and subtasks are dynamically allocated according to the performance characteristics of each computing node (such as the number of CPU cores, memory size, etc.), so that the computing load of each node is relatively balanced, avoiding the situation where the computing pressure of individual nodes is too large and the overall efficiency is reduced. Perform task scheduling and management, monitor the execution progress of each subtask, coordinate data transmission and synchronization operations between nodes, collect the computing status information of each subtask in real time, such as the completed time step, the remaining computing amount, etc., and dynamically adjust the task allocation and data transmission strategy based on this information to cope with possible node failures or changes in computing load.

[0072] Specifically, in the initialization stage of the simulation, a coarser grid is used for division according to the estimated particle distribution of the shield tunneling face and the geometric characteristics of the overall simulation area. For example, for areas with relatively uniform particle distribution and slow movement, larger grid units are used to reduce the number of grids and computing resource requirements. As the simulation proceeds, a particle motion state monitoring mechanism is established. The intensity of particle motion and regional feature changes are judged by real-time analysis of indicators such as particle velocity, acceleration, displacement change, and particle concentration change in local areas. For example, when the average velocity of particles in a certain area exceeds a preset threshold or the particle concentration changes at a large rate in a short period of time, it is determined that the particle motion in the area is more intense or obvious local changes have occurred. Based on the monitoring results, an adaptive grid refinement algorithm is developed. When an area that needs to be refined is detected, the algorithm automatically inserts finer grid units in the area to improve the simulation accuracy. At the same time, in order to avoid the waste of computing resources caused by excessive refinement, a refinement stop condition is set. For example, when the grid is refined to a certain extent, even if the particle motion is still more intense, as long as a certain accuracy requirement is met, further refinement is stopped. During the mesh refinement process, the transition and data transfer between adjacent mesh elements must also be considered to ensure the continuity and accuracy of the simulation.

[0073] Specifically, data acquisition hardware circuits and software programs are developed to realize real-time acquisition and preprocessing of sensor data. The hardware circuit is responsible for conditioning, amplifying, filtering and other processing of the analog or digital signals output by the sensor so that it can meet the input requirements of the data acquisition card; the software program controls the data acquisition card to collect sensor data according to the predetermined sampling frequency and data format, and performs preliminary data verification and error processing. For example, check whether the data is within a reasonable range, whether there is data loss or abnormal fluctuation, etc., mark the abnormal data or take corresponding corrective measures. Build a high-speed data transmission network to transmit the collected sensor data to the numerical simulation system. Use the optical fiber communication network as the main transmission medium to ensure the high speed and stability of data transmission. In the network architecture design, a star topology is used to reduce data transmission delays and failure points. At the same time, configure network switches, routers and other equipment, set network protocols (such as TCP / IP protocol) and data transmission rates to ensure that data can be transmitted to the numerical simulation system in a timely and accurate manner. In the process of data transmission, data encryption and compression technology should also be considered to improve the efficiency and security of data transmission and prevent data from being stolen or tampered with during transmission.

[0074] Specifically, the sensors, data acquisition system and numerical simulation system are integrated to ensure the compatibility and collaborative working ability of each part. The system joint debugging test is carried out to simulate the operation of the shield machine under different working conditions, and to check whether the boundary condition monitoring system can accurately collect and transmit data, and whether the numerical simulation system can correctly receive and use these data for simulation calculation. For example, during the advancement of the shield machine, observe the real-time display and application of data such as advancement speed, cutterhead speed and earth pressure in the numerical simulation system, and check whether the simulation results are reasonably adjusted with the changes in boundary conditions. The system is tested and optimized for performance, including the real-time performance of data acquisition, data transmission delay and packet loss rate, and the computational efficiency and accuracy of numerical simulation. By adjusting the hardware equipment parameters (such as sensor sampling frequency, data acquisition card cache size, network bandwidth, etc.) and software algorithms (such as data processing algorithms, simulation calculation algorithms, etc.), the system performance is continuously optimized to ensure that the system can meet the requirements of actual engineering applications. For example, when it is found that the data transmission delay is high, you can try to increase the network bandwidth or optimize the data transmission protocol; when the numerical simulation calculation efficiency is low, further optimize the parallel computing algorithm or adjust the computing resource allocation strategy.

[0075] It is understandable that the introduction of a parallel framework and the development of an adaptive meshing algorithm can make full use of computing resources, flexibly adjust mesh accuracy according to the simulation process, reduce unnecessary computing overhead and improve simulation accuracy in key areas; building a data acquisition and transmission system and integrating test optimization can accurately and in real time obtain shield tunneling boundary condition data and efficiently transmit it to the numerical simulation system, providing more accurate and efficient numerical simulation support for shield tunneling control methods based on the motion trajectory of rock and soil particles, thereby accurately predicting the shield tunneling status under different working conditions in advance, effectively optimizing tunneling parameters, and significantly improving the safety, reliability and overall efficiency of shield construction, effectively ensuring the smooth progress of the project and accurately achieving the expected goals.

[0076] Specifically, for different types of sensors (such as thrust pressure sensor, cutter head torque sensor, grouting pressure sensor, shield tail seal pressure sensor, slag discharge monitoring equipment, etc.), design corresponding interface circuits and communication protocols. The interface circuit should be able to convert the signal output by the sensor into a format that can be recognized and processed by the shield machine control system, such as converting analog voltage signals into digital signals, and perform necessary signal amplification, filtering and isolation processing to improve signal quality and anti-interference ability. Reserve sensor interface positions in the shield machine control system, and connect each sensor to the control system according to design requirements. During the connection process, ensure that the wiring is correct and firm to avoid problems such as virtual connection and short circuit. For sensors that use wireless communication, set wireless communication modules and frequency bands to ensure stable and reliable wireless communication between the sensor and the control system. After the connection is completed, test each sensor individually to check whether it can work normally and send data to the control system.

[0077] Specifically, a data synchronization control module is developed to ensure the synchronization of data collection of each sensor. Since the sampling frequency and data transmission delay of different sensors may be different, in order to ensure the consistency of multi-parameter data in time, the module uses timestamp technology to mark the data of each sensor, and aligns and synchronizes the data according to the timestamp during data processing. For example, based on the data acquisition time of the propulsion pressure sensor, the data collected by other sensors at the same time or similar time are matched and integrated, so that the multi-parameter data can accurately reflect the operating status of the shield machine at the same moment. A data accuracy verification mechanism is established to perform real-time verification of the collected multi-parameter data. Redundant measurement or comparative verification methods are used, such as installing multiple sensors of the same type at key locations, and judging the accuracy of the data by comparing their measured values; or comparing the sensor measurement value with the theoretical calculated value or empirical value. When the data deviation is found to exceed the allowable range, the data is corrected or an alarm is given in time. At the same time, the sensor is calibrated and maintained regularly to ensure that its measurement accuracy always meets the requirements. For example, the sensor is calibrated using standard calibration equipment at regular intervals (such as one week or one month), the calibration data is recorded, and the sensor parameters are adjusted to ensure the accuracy and reliability of the data.

[0078] Specifically, a data standardization program is developed to perform dimensionless processing on the data of each parameter. Standardization methods such as normalization or standardization are used for parameter data of different types and magnitudes. For parameters with clear physical meanings and magnitude ranges, such as propulsion pressure and cutter head torque, normalization processing can be used to map their data to the interval [0,1]. For parameters with large data magnitude differences, such as slag discharge, standardization processing is used to make their data obey the standard normal distribution. Through data standardization processing, the dimension differences between the data of each parameter are eliminated, which is convenient for subsequent data analysis and model establishment.

[0079] Specifically, the correlation analysis algorithm in big data analysis technology is used to calculate the correlation coefficient between each parameter and the trajectory of rock and soil particles. For example, the Pearson correlation coefficient or Spearman's rank correlation coefficient is used to analyze the linear or nonlinear correlation between parameters such as propulsion pressure, cutter head torque, grouting pressure and particle movement speed, direction, distribution and other characteristics. Through correlation analysis, the parameters closely related to the particle movement trajectory are preliminarily screened out to provide a basis for further in-depth analysis and model construction. The principal component analysis algorithm (PCA) is used to reduce the dimension of multi-parameter data. Since the collected multi-parameter data may have certain redundant information, the principal component analysis is used to convert multiple related original parameters into a few unrelated principal components, which can retain most of the information of the original data. In the process of principal component analysis, the number of principal components and the contribution rate of each principal component are determined, and the principal component with a larger contribution rate is selected as the input variable for the subsequent model establishment, thereby reducing the data dimension, reducing the complexity of the model, and improving the calculation efficiency and generalization ability of the model.

[0080] Specifically, according to the results of correlation analysis and principal component analysis, a model construction method is selected, such as a multivariate linear regression model, a neural network model or a support vector machine model, to establish a multi-parameter coupled shield tunneling state evaluation model. The selected principal components or key parameters are used as the input variables of the model, and the evaluation indicators of the shield tunneling state (such as tunneling stability index, particle movement abnormality, etc.) are used as the output variables of the model. The model is trained through a large amount of historical shield tunneling data. During the model training process, a cross-validation method (such as k-fold cross-validation) is used to evaluate and optimize the performance of the model. The historical data is divided into k non-overlapping subsets, and k-1 subsets are used as training sets each time, and the remaining subset is used as a validation set. The training and validation are repeated k times, and the mean square error (MSE), mean absolute error (MAE) and other evaluation indicators of the model on the validation set are calculated. According to the evaluation results, the structural parameters of the model (such as the number of layers and nodes of the neural network, the coefficients of the regression model, etc.) or the parameters of the optimization algorithm (such as learning rate, regularization parameter, etc.) are adjusted to continuously improve the accuracy and generalization ability of the model. After multiple training and optimizations, the final shield tunneling status assessment model is obtained and deployed in the actual shield tunneling control system.

[0081] It can be understood that the design of sensor interface and connection ensures stable and accurate data interaction between various sensors and control systems of the shield machine, providing a rich and reliable data source for subsequent analysis; building a data correlation analysis platform for data standardization, correlation analysis and dimensionality reduction processing can eliminate dimensional differences, explore the intrinsic connections between parameters and reduce data complexity, thereby improving data availability and analysis efficiency; and then building and optimizing the shield tunneling status assessment model, which can comprehensively and accurately assess the shield tunneling status based on multi-parameter data, predict potential risks and anomalies in advance, and provide a scientific decision-making basis for the shield tunneling control method based on the motion trajectory of rock and soil particles, effectively optimize the tunneling strategy, and significantly improve the intelligence level, safety and quality control effect of shield construction, to ensure the smooth and efficient progress of the project.

[0082] Specifically, in addition to considering recurrent neural networks (RNNs) and their variants (such as long short-term memory networks (LSTMs), it is also necessary to study the feasibility of other deep learning architectures such as convolutional neural networks (CNNs) in processing rock and soil particle motion image data (if there is relevant image monitoring data) and fusing other parameter data. Analyze the advantages and disadvantages of different architectures in feature extraction, sequence processing, and model complexity, and combine the real-time and accuracy requirements of shield tunneling control for data processing to determine the most suitable network architecture combination or single architecture. For the LSTM architecture to be adopted, when determining the number of input layer nodes according to the characteristic dimension of the input data, deeply analyze the importance and correlation of each data type, and use feature engineering methods to reduce the dimension of the data or optimize the feature combination. For example, for rock and soil particle motion trajectory data, if its three-dimensional coordinate information has a strong correlation, it can be considered to convert it into a smaller feature representation through methods such as principal component analysis (PCA) to determine the number of input nodes. When designing the hidden layer, in addition to determining the approximate range of the number of layers and nodes, it is also necessary to consider the impact of different inter-layer connection methods, such as full connection and skip connection, on the model learning ability. Through small-scale experiments, the performance of the model on the training set and the validation set under different connection methods is compared, and the optimal connection strategy is selected. At the same time, for the activation function, in addition to the commonly used ReLU function, other variants such as LeakyReLU, ELU, etc. can also be tried to observe their impact on the convergence speed and performance of the model when processing shield tunneling data. When determining the loss function, in addition to the mean square error (MSE) and cross entropy loss functions, explore the construction of custom loss functions based on the characteristics of the control instructions and actual engineering needs. For example, if there are high requirements for the stability of the control instructions, a penalty term for the rate of change of the control instructions can be added to the loss function to guide the model to generate a smoother sequence of control instructions.

[0083] Specifically, in the data collection process, it is necessary not only to obtain data from the existing shield engineering case database, but also to formulate a detailed plan for supplementary monitoring data on site. It is necessary to clarify the data types and frequencies that need to be collected under different geological conditions, shield tunneling stages (such as starting, normal tunneling, receiving, etc.) and different construction processes (such as different grouting methods, slag improvement methods, etc.). For example, when crossing complex strata, the frequency of collecting rock and soil particle trajectory data and stress data of key parts of the shield machine should be increased to obtain richer samples for model training. In the data cleaning stage, in addition to identifying and processing outliers based on statistical methods, a data quality assessment model should be established. This model can comprehensively consider multiple factors such as data integrity, consistency, and accuracy, score each data record, and select high-quality data for subsequent processing based on the scoring results. For data missing value processing, when using mean, median filling or model-based filling methods, the impact of different filling methods on model training results is further analyzed, and the optimal filling strategy under different data types and missing ratios is determined through comparative experiments. In the data labeling work, in addition to simply labeling according to the shield tunneling status, it is also necessary to label the subdivided features under different states. For example, for unstable tunneling states, further labeling is due to stratum changes, equipment failure, or improper operation, so that the model can more accurately learn the control strategy response mechanism under different situations. When dividing the training set, validation set, and test set, the time series segmentation method is used to ensure that the training set data precedes the validation set and test set in time, so as to simulate the time sequence learning process of the model in actual application and improve the generalization ability of the model.

[0084] Specifically, when using the back-propagation algorithm to train a model, in addition to setting the learning rate, the momentum term parameters also need to be adjusted. The momentum term can accelerate the convergence of the model in the gradient direction and reduce oscillations. The momentum value range is determined through experiments, and the model training process is optimized in coordination with the learning rate. At the same time, an adaptive learning rate adjustment strategy is adopted, such as the learning rate decay combined with the learning rate annealing method, to dynamically adjust the learning rate according to the loss changes during the model training process, so that the model can learn quickly in the early stage of training and fine-tune the parameters in the later stage. When using the early stopping method, in addition to monitoring the loss value or evaluation index on the validation set, the prediction stability index of the model on the validation set, such as the variance of the prediction value, can also be combined. When the loss value no longer increases and the prediction stability index also meets certain requirements, it is determined that the model performance is no longer improved and the training is stopped. In addition, model integration technology can also be used, such as weighted averaging or voting combination of excellent models at different stages of multiple training processes to improve the overall performance and robustness of the model. After the model training is completed, the sensitivity analysis method is used to evaluate the importance of the model's input features and determine the key features that have a greater impact on the generation of control instructions, so that the data collection and processing links related to these features can be focused on in the subsequent optimization process.

[0085] It is understandable that by determining the appropriate network architecture combination and carefully improving the data processing process, it is possible to efficiently integrate and deeply mine various types of data in the shield excavation process, and accurately extract key information; on this basis, scientific training and adjustment can enable the constructed adaptive control model to fully learn the regular characteristics under different geological conditions and excavation conditions, so as to quickly and accurately generate optimized control instructions based on real-time data, effectively respond to complex and changeable geological conditions and construction environments, greatly improve the automation and intelligence level of shield excavation, significantly enhance the safety and stability of construction, effectively ensure the timely and quality completion of the project, and provide highly targeted and efficient solutions for shield excavation control.

[0086] Specifically, when adopting multi-threaded processing technology, thread resources are reasonably allocated according to different task modules in the data processing process, such as data verification, classification processing, model input data preparation, etc. Design a thread pool management strategy to ensure that thread creation, destruction and task allocation are carried out efficiently, and avoid thread resource waste and competition conflicts. For example, for the data verification thread, a higher priority is set to ensure that the accuracy and integrity of the data are guaranteed at the first time; for the model input data preparation thread, its priority and resource allocation are dynamically adjusted according to the real-time demand of the model for data. In terms of data caching mechanism, a multi-level cache structure is designed. A first-level cache is set at the sensor data receiving end to temporarily store the newly collected data to avoid data loss; a second-level cache is set in front of the data processing module to cache the data that has been preliminarily verified and classified to meet the data processing module's demand for batch data processing; a third-level cache is set at the model input end to ensure that the model can quickly obtain the processed data for instruction generation. At the same time, a cache elimination strategy, such as the least recently used (LRU) algorithm, is adopted to ensure the effective use of cache space and timely eliminate expired or infrequently used data. When optimizing the algorithm process, a detailed time complexity analysis is performed on the entire process of data processing and instruction generation. Adopt algorithm optimization techniques, such as loop unrolling and code vectorization, to reduce the computational overhead in the data processing process. For example, in the data verification link, if there are a large number of repetitive verification calculations, multiple loop calculations can be combined into one calculation through loop unrolling to improve the computational efficiency. At the same time, optimize the process of model generation control instructions, and adopt model compression techniques, such as pruning and quantization, to reduce the model's computational workload and memory usage, thereby speeding up instruction generation.

[0087] Specifically, in the process of network construction, in addition to ensuring that the network bandwidth meets the real-time requirements of control instruction transmission, it is also necessary to accurately measure and analyze network delay. Use network performance testing tools such as Ping and Traceroute to test network delay regularly and establish a network delay monitoring model. According to the model prediction results, adjust the network configuration in advance or take network optimization measures, such as optimizing routing strategies and adding network cache devices, to cope with possible network delay fluctuations. When adopting redundant design, in addition to network link redundancy, network device redundancy must also be considered. Configure redundant devices at key network nodes (such as switches, routers, etc.). When the main device fails, it can automatically switch to the backup device to ensure uninterrupted operation of the network. At the same time, establish a network device fault monitoring and early warning system to monitor the operating status of network devices in real time, discover potential fault hazards in advance, and perform maintenance and repair in time. In terms of network self-healing technology, a self-healing mechanism based on software-defined networking (SDN) is adopted. The SDN controller monitors and analyzes network traffic in real time. When a network failure or congestion is detected, the network topology can be automatically adjusted and the data transmission path can be replanned to achieve rapid self-healing of the network. In the process of connecting the control command generation system with the shield machine actuator, a connection status monitoring module is designed to monitor the stability of the connection and the correctness of data transmission in real time. A heartbeat mechanism is used to send test signals regularly to ensure that both parties can perceive the status of each other in time. Once an abnormal connection is found, emergency measures such as reconnection or switching to a backup connection line are immediately initiated.

[0088] Specifically, according to the different stages of shield tunneling (such as the starting section, normal tunneling section, crossing special strata, etc.) and control requirements, a detailed feedback data collection plan is formulated. The characteristics of the movement trajectory of rock and soil particles that need to be focused on at each stage (such as the velocity variation range of particles in a specific area, the concentration of the movement direction, etc.) and the key operating state parameters of the shield machine (such as the propulsion force fluctuation range, the cutter head torque change rate, etc.) are clarified. The time interval setting principle for collecting data is determined. For example, when the geological conditions are stable and the tunneling speed is uniform, the collection interval can be appropriately extended; when crossing complex strata or approaching important buildings, the collection interval can be shortened to obtain more detailed feedback information. Considering the redundancy of data collection, a multi-sensor collection plan is formulated for key data points (such as the movement of rock and soil particles around the head of the shield machine, the pressure data of the main propulsion cylinder, etc.) to ensure the reliability and integrity of the data. At the same time, the maintenance cycle and calibration plan of the data acquisition equipment are planned to ensure that the equipment operates stably throughout the tunneling process and the data collection accuracy meets the requirements.

[0089] Specifically, data acquisition equipment (including various sensors, data acquisition cards, etc.) should be inspected regularly. Check the installation firmness of the sensor to see if it is loose, displaced or damaged; clean and maintain the measuring elements of the sensor to prevent dust, mud and other impurities from affecting the measurement accuracy. For example, the optical lens of the laser displacement sensor should be wiped and calibrated regularly to ensure that it can accurately capture the position changes of rock and soil particles. Perform performance tests on the sensor network, including the detection of indicators such as network delay and packet loss rate. Use network monitoring tools, such as the Ping command and network traffic analysis software, to monitor the network status in real time. If the network delay is too high or the packet loss is serious, promptly check the cause of the network failure, which may be caused by network cable damage, network switch failure or electromagnetic interference. Take appropriate repair measures, such as replacing cables, restarting switches or adding shielding measures, to ensure that data can be transmitted to the data processing center in a timely and accurate manner.

[0090] Specifically, during the data transmission process, data encryption technology is used to encrypt the feedback data to prevent the data from being stolen or tampered with during the transmission process. An encryption algorithm, such as the AES (Advanced Encryption Standard) encryption algorithm, is selected to ensure data security. At the same time, the transmitted data is compressed to improve data transmission efficiency and reduce network bandwidth usage. For example, a lossless compression algorithm (such as LZ77, LZMA, etc.) is used to compress the data, and then the data is decompressed and restored at the receiving end. A special feedback data repository is established in the data processing center, and classified and stored according to data type (such as rock and soil particle motion trajectory data, shield machine operation parameter data, etc.), acquisition time, shield machine location and other information. A database management system (such as MySQL, Oracle, etc.) is used to manage the data to facilitate data query, retrieval and subsequent analysis. The data repository is backed up regularly to prevent data loss, and the backup data is stored in an off-site server or storage medium to cope with possible hardware failures, natural disasters, etc.

[0091] Specifically, after receiving the feedback data, the data processing center first performs a data integrity check. Check whether there are problems such as missing data and data format errors. For missing data, interpolation methods such as linear interpolation and spline interpolation are used to supplement according to the time series characteristics and correlation of the data. For data format errors, data format conversion and correction are performed to ensure that the data can be correctly identified and processed by subsequent analysis software. The data is filtered to remove noise interference in the data. For the rock and soil particle motion trajectory data, since it may be affected by factors such as shield machine vibration and electromagnetic interference during the collection process, digital filtering technology (such as low-pass filtering, Kalman filtering, etc.) is used to smooth the data and extract the true trend of particle movement. For the shield machine operation parameter data, such as high-frequency fluctuations in pressure and torque data, corresponding filtering processing is also performed to accurately reflect the actual operating status of the equipment.

[0092] Specifically, when calculating the deviation rate of particle motion trajectory, the reference trajectory is first determined. The reference trajectory can be an ideal trajectory calculated based on a theoretical model, or it can be a relatively stable particle motion trajectory data collected during the previous excavation process under the same geological conditions. Mathematical methods are used to calculate the deviation between the actual trajectory and the reference trajectory, such as calculating the average distance between trajectory points, root mean square error (RMSE) and other indicators. At the same time, the distribution of the deviation rate in different time stages and different spatial regions is analyzed to find out the time periods and regions with large deviations, providing a basis for subsequent cause analysis. For the stability index of shield machine operation parameters, the variance, standard deviation and other statistics of the parameters are calculated to measure their degree of fluctuation. For example, for the propulsion pressure data, its variance over a period of time is calculated. If the variance is too large, it indicates that the propulsion pressure is unstable, which may affect the stability and safety of shield excavation. In addition, time series analysis methods, such as autocorrelation function analysis, can be used to determine whether there are periodic fluctuations or abnormal change trends in the operating parameters, and further understand the operating status of the equipment.

[0093] Specifically, a comprehensive evaluation model is established to weight and fuse the particle motion trajectory deviation rate, shield machine operation parameter stability index and other related indicators (such as the stability of slag discharge, the rationality of grouting pressure, etc.) to obtain a comprehensive evaluation value that fully reflects the control effect. The weight coefficients of each indicator are determined by methods such as the analytic hierarchy process (AHP) to make the comprehensive evaluation results more scientific and reasonable. According to the comprehensive evaluation results and the analysis of various quantitative indicators, the reasons for the poor control effect are deeply diagnosed. Consider various factors that may affect the control effect, such as changes in geological conditions (sudden increase in formation hardness, changes in water content, etc.), shield machine equipment failures (tool wear, hydraulic system leakage, etc.), and unreasonable control strategies (improper control parameter settings, inaccurate model predictions, etc.). Fault tree analysis (FTA) and other methods are used to gradually investigate possible causes and determine key factors, providing a clear direction for subsequent optimization and adjustment.

[0094] Specifically, if it is determined that the inaccurate monitoring data of the trajectory of rock and soil particles in a certain area leads to poor control effect, the optimization team will first conduct a detailed review of the geological conditions in the area. Use geological radar, core drilling and other means to further understand the stratigraphic structure, rock and soil characteristics (such as particle size distribution, particle shape, hardness, etc.) and groundwater conditions. Re-evaluate the sensor layout plan based on the review results and the measurement principle of the sensor. Consider increasing the number of sensors or adjusting the installation position and angle of the sensors to better cover the monitoring area and improve the accuracy of data collection. For example, if it is found that the particle movement in a certain area changes greatly in the vertical direction, and the original sensors are arranged more in the horizontal direction, the number of sensors installed in the vertical direction can be appropriately increased or their angles can be adjusted to enable them to capture the particle movement trajectory more accurately. At the same time, the new sensor layout plan is simulated and analyzed, and numerical simulation software or physical model experiments are used to predict the data collection effect under the new plan, verify its feasibility and effectiveness, and ensure that it can meet the control requirements before the actual sensor adjustment and installation.

[0095] Specifically, when problems are found in the data processing algorithm when processing feedback data (such as poor data filtering effect, inaccurate feature extraction, etc.), the data processing algorithm optimization team conducts a comprehensive review and optimization of the algorithm. First, analyze the shortcomings of the existing algorithm when processing the current data. For example, if the filtering algorithm used cannot effectively remove the noise of a specific frequency, resulting in data distortion, study other more suitable filtering algorithms or adjust the parameters of the existing algorithm. In view of the problem of inaccurate data feature extraction, feature engineering technology is used for optimization. Re-analyze the characteristics of the feedback data and explore new feature combinations or feature transformation methods to improve the correlation between data features and shield tunneling control objectives. For example, for the trajectory data of rock and soil particles, in addition to the conventional position and velocity characteristics, try to extract the acceleration characteristics and trajectory curvature characteristics of the particle movement, and use methods such as principal component analysis (PCA) to screen out the key features that have a greater impact on the control effect, and use them as the input of the data processing algorithm to improve the algorithm's understanding and processing capabilities of the data. At the same time, the optimized algorithm is performance tested, and the algorithm is verified using a large amount of historical feedback data and simulation data to ensure that the algorithm has been significantly improved in terms of accuracy, efficiency and stability.

[0096] Specifically, if the poor control effect is caused by the inaccurate description of the actual geological conditions by the physical model or the unreasonable setting of the numerical simulation parameters, the physical model and numerical simulation optimization team will carry out the following work. First, a sensitivity analysis is performed on the parameters in the physical model to determine which parameters have a greater impact on the model output (such as the prediction results of the motion trajectory of rock and soil particles). For example, in the model considering the interaction force between particles, the sensitivity of parameters such as friction coefficient and cohesion is analyzed. According to the results of the sensitivity analysis, combined with the monitoring data and geological survey data in the actual project, the key parameters are recalibrated. More accurate parameter values ​​are obtained through indoor tests, on-site in-situ tests, etc., and they are substituted into the physical model for verification and optimization. At the same time, the numerical simulation parameters (such as grid division accuracy, time step, etc.) are adjusted. If it is found during the simulation process that the particle motion simulation results in some areas deviate greatly from the actual monitoring data, consider locally refining the grid in that area to improve the simulation accuracy; or adjust the time step to ensure that the simulation process can accurately capture the dynamic changes of particle motion. During the parameter adjustment process, multiple groups of parameter combinations are used to carry out numerical simulation experiments, and the simulation results under different combinations are compared with the actual feedback data. The optimal parameter combination is selected to make the physical model and numerical simulation results closer to the actual engineering situation, providing a more reliable basis for the formulation of control strategies.

[0097] Specifically, when the control model does not predict well under certain geological conditions, the control model optimization team first conducts a detailed analysis of the input data and output results of the model under these geological conditions. Check whether the quality and feature distribution of the input data are significantly different from those during model training, for example, whether there are new geological features or changes in the operating status of the shield machine that are not fully considered by the model. If it is an input data problem, optimize and adjust the data acquisition and preprocessing links to ensure that the data input to the model can accurately reflect the actual situation. If it is a problem with the model itself, consider increasing the amount of training data, collecting more shield tunneling data under the same geological conditions, including rock and soil particle motion trajectory data, shield machine operating parameter data, and corresponding control instruction data, and retrain the model. At the same time, adjust the structure and parameters of the model, such as increasing the number of hidden layers, changing the way neurons are connected, and optimizing the parameters of the activation function and loss function. Use cross-validation, model fusion and other technologies to improve the generalization ability and robustness of the model. For example, integrate multiple models with different structures or parameter settings, and obtain the final control instructions through weighted average or voting to reduce the prediction error of a single model and improve the control effect. After the model is optimized, rigorous testing and verification are carried out, and the model performance is evaluated using independent test data sets and actual engineering scenario simulation data to ensure that the model can accurately and stably generate control instructions under various geological conditions and achieve efficient and safe control of shield tunneling.

[0098] In the above-mentioned embodiment, a shield tunneling control method based on the movement trajectory of rock and soil particles can accurately obtain geological information and particle movement data through the fusion deployment of multiple types of high-precision sensors; intelligent anti-noise data acquisition and preprocessing ensure data quality; accurate physical model construction provides a basis for simulation; efficient numerical simulation optimization improves calculation efficiency and accuracy; multi-parameter comprehensive monitoring and correlation analysis realize comprehensive status assessment; real-time data-driven adaptive control model construction achieves intelligent control; control instructions are quickly generated and transmitted to ensure timeliness; closed-loop feedback and continuous optimization continuously improve performance and reliability, thereby effectively improving the safety, accuracy, adaptability and intelligence of shield tunneling construction, reducing construction risks and costs, improving construction efficiency and quality, and adapting to complex and changeable geological conditions and engineering requirements.

[0099] Another preferred method based on the above embodiment is as follows: Figure 2 As shown, this embodiment provides a shield tunneling control system based on the movement trajectory of rock and soil particles, including:

[0100] The shield machine sensor network construction and debugging module designs the sensor network topology according to the shield machine structure and sensor distribution;

[0101] Intelligent electromagnetic noise monitoring and automatic compensation module, develop intelligent electromagnetic noise monitoring and automatic compensation program, receive electromagnetic environment detection sensor data, monitor electromagnetic noise intensity and frequency in real time and collect them regularly;

[0102] The precise physical model building module conducts multi-condition combination tests, compares the field data with the indoor test data, and builds and calibrates the physical model of the movement trajectory of rock and soil particles that takes into account multiple factors;

[0103] Efficient numerical simulation optimization module, introducing parallel framework and developing adaptive meshing algorithm, building data acquisition and transmission system and integrating test optimization, jointly improving the efficiency and accuracy of numerical simulation;

[0104] Multi-parameter comprehensive monitoring and correlation analysis module, design sensor interfaces and connect them, build a data correlation analysis platform for data standardization, correlation analysis and dimensionality reduction, and build and optimize the shield tunneling status assessment model;

[0105] Real-time data-driven adaptive control model building module determines the network architecture combination and improves data processing, performs training adjustments, and builds an adaptive control model that can meet the shield tunneling control needs;

[0106] The control instruction rapid generation and transmission module optimizes the algorithm flow and designs multi-threading and caching mechanisms. Through network-related configuration, redundant design, self-healing technology, etc., it ensures that the control instructions can be quickly and accurately generated and transmitted to the shield machine actuator;

[0107] The closed-loop feedback and continuous optimization module formulates feedback data collection plans, maintains equipment networks and safely stores data, organizes and analyzes data and conducts quantitative evaluation and diagnosis of causes. It optimizes and improves issues such as sensor layout, data processing algorithms, physical models and control models, and forms a closed-loop feedback and continuous optimization system for shield tunneling control.

[0108] It can be understood that the above-mentioned shield tunneling control system and method based on the movement trajectory of rock and soil particles have the same beneficial effects and will not be repeated here.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does 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 shield tunneling control method based on rock and soil particle motion trajectory, characterized in that: include: Design the sensor network topology according to the shield machine structure and sensor distribution; Develop intelligent electromagnetic noise monitoring and automatic compensation programs, receive electromagnetic environment detection sensor data, monitor electromagnetic noise intensity and frequency in real time and collect them regularly; Conduct multi-condition combination tests, compare field data with indoor test data, and construct and calibrate a physical model of rock and soil particle motion trajectory that takes multiple factors into account; Introducing parallel frameworks and developing adaptive meshing algorithms, building data acquisition and transmission systems and integrating test optimization to jointly improve the efficiency and accuracy of numerical simulations; Design sensor interfaces and connect them, build a data association analysis platform to perform data standardization, association analysis and dimensionality reduction, and build and optimize the shield tunneling status assessment model; Determine the network architecture combination and improve data processing, make training adjustments, and build an adaptive control model that can meet the needs of shield tunneling control; Optimize the algorithm flow and design multi-threading and caching mechanisms, and ensure that control instructions can be quickly and accurately generated and transmitted to the shield machine actuator through network-related configuration, redundant design, self-healing technology, etc.; Develop feedback data collection plans, maintain equipment networks and safely store data, organize and analyze data and quantitatively evaluate and diagnose causes, optimize and improve issues such as sensor layout, data processing algorithms, physical models and control models, and form a closed-loop feedback and continuous optimization system for shield tunneling control.

2. A shield tunneling control method based on rock and soil particle motion trajectory according to claim 1, characterized in that: The sensor network topology is designed according to the shield machine structure and sensor distribution conditions, including: Sensor network construction and debugging: Design the sensor network topology according to the structure of the shield machine and the distribution of sensors, set up switches and routers in the network and plan the data transmission path, determine the network communication protocol and data storage format, debug the sensor network, test the network connection between each sensor node and the data processing center, perform data fusion test, transmit the data collected by different sensors to the data processing center, run the data fusion algorithm, check the algorithm's integration effect on different sensor data, if the data fusion effect is not good, check whether the algorithm parameter settings are reasonable, and adjust the algorithm parameters in time according to the test results until the data fusion effect meets the requirements.

3. The shield tunneling control method based on the movement trajectory of rock and soil particles according to claim 1 is characterized in that: The development of intelligent electromagnetic noise monitoring and automatic compensation program receives electromagnetic environment detection sensor data, monitors electromagnetic noise intensity and frequency in real time and collects them regularly, including: Intelligent electromagnetic noise monitoring and automatic compensation, develop intelligent electromagnetic noise monitoring and automatic compensation program, receive data from electromagnetic environment detection sensors, monitor the intensity and frequency of electromagnetic noise in real time, collect electromagnetic noise data at regular intervals, perform fast Fourier transform analysis on the collected data, obtain the spectrum information of electromagnetic noise, compare the current electromagnetic noise spectrum with the preset threshold spectrum, and if it is found that the electromagnetic noise intensity in a certain frequency band exceeds the preset threshold, calculate the parameter values ​​that need to be adjusted for the electromagnetic filter according to the pre-set adjustment strategy, realize automatic optimization of the electromagnetic filter parameters, and record the historical records of electromagnetic filter parameter adjustments.

4. The shield tunneling control method based on rock and soil particle motion trajectory according to claim 1 is characterized in that: The multi-condition combination test is carried out to compare the field collected data with the indoor test data, and to construct and calibrate a physical model of the movement trajectory of rock and soil particles that takes into account multiple factors, including: Indoor test: According to the geological conditions and rock and soil particle characteristics that may be encountered in shield construction, a series of indoor test schemes are designed. The test schemes include different combinations of water content, particle surface roughness and different stress states. High-precision mechanical sensors, test containers, loading devices, simulated rock and soil particles and other test equipment and materials are prepared. The simulated rock and soil particles are placed in the test container according to the predetermined particle grading and stacking method. The water content and surface roughness of the particles are adjusted, and then different stress states are applied through the loading device. The interaction force between particles is measured in real time using mechanical sensors, and data such as the magnitude, direction and action time of the force are recorded. Displacement sensors are used to monitor the displacement changes of particles, and the movement and deformation of particles under different conditions are observed. Each test condition is tested repeatedly, and the data of each test is recorded and stored in detail. Field data collection and comparative analysis: Install sensors on the shield machine that can collect data on the interaction between rock and soil particles. The sensors are installed on key locations such as the cutter head, shield, and slag conveying pipeline of the shield machine to directly or indirectly obtain information on the interaction between rock and soil particles during the actual excavation process. Continuously collect data on the interaction between rock and soil particles, record the data collection time, collection location, and shield machine operating parameters, establish field data archives, compare and analyze field data with indoor test data, and compare indoor test data and field data under the same geological conditions to analyze the differences and similarities between the two. Physical model construction and calibration: Based on indoor tests and field data, a physical model of the motion trajectory of rock and soil particles is established that takes into account factors such as particle shape, surface roughness, and water content. Micromechanics theory and multi-scale analysis methods are introduced to link the microstructural characteristics of rock and soil particles with their macroscopic motion behavior, and to determine key parameters such as the elastic modulus, Poisson's ratio, friction coefficient, and cohesion coefficient of the particles in the model. The model parameters are calibrated and optimized through repeated test data fitting and field data verification. The optimization algorithm is used to compare the model calculation results with the actual test data and field data, and the model parameters are continuously adjusted so that the model can accurately reflect the interaction relationship between rock and soil particles under different geological conditions. The historical records of each parameter adjustment and the corresponding model evaluation indicators are recorded.

5. The shield tunneling control method based on rock and soil particle motion trajectory according to claim 1 is characterized in that: The introduction of parallel framework and development of adaptive meshing algorithm, construction of data acquisition and transmission system and integrated test optimization, jointly improve the efficiency and accuracy of numerical simulation, including: The numerical simulation algorithm is optimized by using the discrete element method. Based on the MPI framework, the entire simulation area is divided into multiple sub-areas. The particle motion calculation tasks of each sub-area are assigned to different computing nodes. The large-scale rock and soil particle system is evenly divided into several sub-task areas according to the spatial position. The sub-tasks are dynamically assigned according to the performance characteristics of each computing node, and task scheduling and management are performed. The execution progress of each sub-task is monitored, and the data transmission and synchronization operations between nodes are coordinated. The computing status information of each sub-task is collected in real time, and the task allocation and data transmission strategies are dynamically adjusted according to this information. According to the estimated particle distribution of the shield excavation face and the geometric characteristics of the overall simulation area, a coarser grid is used for division, and a particle motion state monitoring mechanism is established. By real-time analysis of indicators such as particle velocity, acceleration, displacement changes, and particle concentration changes in local areas, the intensity of particle motion and regional feature changes are judged. According to the monitoring results, an adaptive grid refinement algorithm is developed. When an area that needs to be refined is detected, the algorithm automatically inserts finer grid units in the area and sets refinement stop conditions. When the grid is refined to a certain extent, even if the particle motion is still violent, as long as certain accuracy requirements are met, further refinement will be stopped; Real-time boundary condition detection, development of data acquisition hardware circuits and software programs, to achieve real-time acquisition and preprocessing of sensor data. The hardware circuit is responsible for conditioning, amplifying, filtering and other processing of the analog or digital signals output by the sensor, and the software program controls the data acquisition card to collect sensor data according to the predetermined sampling frequency and data format, and perform preliminary data verification and error processing; build a high-speed data transmission network to transmit the collected sensor data to the numerical simulation system, using the optical fiber communication network as the main transmission medium. In the network architecture design, a star topology is adopted, and network switches, routers and other equipment are configured to set the network protocol and data transmission rate; integrate sensors, data acquisition systems and numerical simulation systems, conduct system joint debugging tests, simulate the operation of shield machines under different working conditions, check whether the boundary condition monitoring system can accurately collect and transmit data, whether the numerical simulation system can correctly receive and use these data for simulation calculations, and perform performance tests and optimizations on the system in terms of real-time data acquisition, data transmission delay and packet loss rate, and numerical simulation calculation efficiency and accuracy, and continuously optimize system performance by adjusting hardware equipment parameters and software algorithms.

6. The shield tunneling control method based on rock and soil particle motion trajectory according to claim 1 is characterized in that: The design of sensor interfaces and connection, construction of a data association analysis platform for data standardization, association analysis and dimensionality reduction, and construction and optimization of a shield tunneling status assessment model include: Multi-parameter detection, design corresponding interface circuits and communication protocols for different types of sensors. The interface circuit can convert the signal output by the sensor into a format that can be recognized and processed by the shield machine control system; reserve the sensor interface position in the shield machine control system, and connect each sensor to the control system according to the design requirements. After the connection is completed, test each sensor individually to check whether it can work normally and send data to the control system; use timestamp technology to mark the data of each sensor, and align and synchronize the data according to the timestamp during the data processing process, establish a data accuracy verification mechanism, and perform real-time verification of the collected multi-parameter data; Data association analysis, development of data standardization processing program, dimensionless processing of each parameter data, normalization processing is used for parameters with clear physical meaning and magnitude range such as thrust pressure and cutter head torque, and their data are mapped to the interval [0,1]. For parameters with large data magnitude differences such as slag discharge, standardization processing is used to make their data obey the standard normal distribution; the correlation analysis algorithm in big data analysis technology is used to calculate the correlation coefficient between each parameter and the motion trajectory of rock and soil particles, and the principal component analysis algorithm is used to reduce the dimension of multi-parameter data, converting multiple related original parameters into a few unrelated principal components, which can retain most of the information of the original data and determine the number of principal components. And the contribution rate of each principal component, select the principal component with larger contribution rate as the input variable for model establishment; according to the results of correlation analysis and principal component analysis, a multi-parameter coupled shield tunneling state evaluation model is established, the screened principal components or key parameters are used as the input variables of the model, and the evaluation index of shield tunneling state is used as the output variable of the model, and the model is trained through historical shield tunneling data; in the process of model training, the cross-validation method is used to evaluate and optimize the performance of the model, and according to the evaluation results, the structural parameters of the model or the parameters of the optimization algorithm are adjusted to continuously improve the accuracy and generalization ability of the model. After multiple training and optimization, the final shield tunneling state evaluation model is obtained, and it is deployed in the actual shield tunneling control system.

7. The shield tunneling control method based on rock and soil particle motion trajectory according to claim 1 is characterized in that: The above-mentioned determination of network architecture combination and improvement of data processing, training and adjustment, and construction of an adaptive control model that can meet the shield tunneling control requirements include: Model framework selection and design, analyze the advantages and disadvantages of different architectures in feature extraction, sequence processing and model complexity, combine the real-time and accuracy requirements of shield tunneling control for data processing, determine the most suitable network architecture combination or single architecture, determine the range of the number of layers and nodes when designing hidden layers, consider the impact of different inter-layer connection methods on the model learning ability, compare the performance of the model on the training set and validation set under different connection methods through small-scale experiments, and select the optimal connection strategy; when determining the loss function, explore the construction of a custom loss function based on the characteristics of the control instructions and actual engineering needs; Data collection and preprocessing: collect data from the existing shield engineering case database, formulate a supplementary monitoring data plan on site, and clarify the data types and frequencies that need to be collected under different geological conditions, shield tunneling stages, and different construction processes; perform data cleaning, identify and process outliers based on statistical methods, establish a data quality assessment model, comprehensively consider the integrity, consistency, and accuracy of the data, score each data record, and screen out high-quality data for subsequent processing based on the scoring results; use mean, median filling or model-based filling methods to further analyze the impact of different filling methods on model training results, and determine the optimal filling strategy under different data types and missing ratios through comparative experiments; perform data annotation, classify and annotate according to the shield tunneling status, annotate the subdivision features under different states, and use time series segmentation methods to ensure that the training set data precedes the validation set and test set in time to simulate the time-sequential learning process of the model in actual applications; Model training and optimization: use the back-propagation algorithm to train the model, set the learning rate, adjust the momentum term parameters, and adopt an adaptive learning rate adjustment strategy to dynamically adjust the learning rate according to the change in loss during model training. Use the early stopping method to monitor the loss value or evaluation index on the validation set and combine it with the prediction stability index of the model on the validation set. When the loss value no longer increases and the prediction stability index also meets certain requirements, it is determined that the model performance no longer improves, and training is stopped. Model integration technology is used to weighted average or vote on excellent models at different stages of multiple training processes. After the model training is completed, the sensitivity analysis method is used to evaluate the importance of the model's input features and determine the key features that have a greater impact on the generation of control instructions.

8. The shield tunneling control method based on rock and soil particle motion trajectory according to claim 1 is characterized in that: The optimization algorithm process described above designs a multi-threading and cache mechanism, and ensures that control instructions can be quickly and accurately generated and transmitted to the shield machine actuator through network-related configuration, redundant design, self-healing technology, etc., including: System software development uses multi-threaded processing technology to reasonably allocate thread resources according to different task modules in the data processing process, design thread pool management strategies, set higher priorities for data verification threads to ensure that the accuracy and integrity of the data are guaranteed at the first time; for model input data preparation threads, dynamically adjust their priorities and resource allocation according to the real-time data needs of the model; set a first-level cache at the sensor data receiving end to temporarily store the newly collected data, set a second-level cache at the data processing front end to cache the data that has been preliminarily verified and classified, and set a third-level cache at the model input end to ensure that the model can quickly obtain processed data for instruction generation; use algorithm optimization technology, loop unrolling, code vectorization, etc. to reduce the computational overhead in the data processing process, optimize the process of model generation control instructions, and use model compression technology to reduce the computational workload and memory usage of the model; Network construction and system integration, use network performance testing tools to test network delays regularly, establish a network delay monitoring model, adjust network configuration in advance or take network optimization measures based on model prediction results to cope with possible network delay fluctuations, configure redundant equipment at key network nodes, and automatically switch to backup equipment when the main equipment fails to ensure uninterrupted operation of the network, establish a network equipment failure monitoring and early warning system, monitor the operating status of network equipment in real time, discover potential fault hazards in advance, and perform maintenance and repair in a timely manner; use the SDN controller to monitor and analyze network traffic in real time, and automatically adjust the network topology and re-plan the data transmission path when a network failure or congestion is detected to achieve rapid self-healing of the network; during the connection between the control instruction generation system and the shield machine actuator, monitor the connection status, monitor the stability of the connection and the correctness of the data transmission in real time, and use the heartbeat mechanism to send test signals regularly to ensure that both parties can perceive each other's status in a timely manner. Once an abnormal connection is found, immediately initiate emergency measures such as reconnection or switching to the backup connection line.

9. The shield tunneling control method based on rock and soil particle motion trajectory according to claim 1, characterized in that: The aforementioned formulation of feedback data collection plan, maintenance of equipment network and safe storage of data, collation and analysis of data and quantitative evaluation and diagnosis of causes, optimization and improvement of sensor layout, data processing algorithm, physical model and control model, etc., form a closed-loop feedback and continuous optimization system for shield tunneling control, including: Data collection and transmission: formulate feedback data collection plans according to different stages and control requirements of shield tunneling, clarify the characteristics of rock and soil particle movement trajectories and key operating status parameters of shield machines that need to be focused on at each stage, determine the time interval setting principles for data collection, formulate multi-sensor collection plans for key data points, ensure data reliability and integrity, plan the maintenance cycle and calibration plan of data collection equipment, ensure that the equipment operates stably throughout the tunneling process, and that data collection accuracy meets requirements; regularly inspect data collection equipment to check the installation firmness of sensors, and check whether they are loose, displaced or damaged; clean and maintain the measuring elements of sensors to prevent dust, mud and other impurities from affecting measurement accuracy. Conduct performance tests on the sensor network and use network monitoring tools to monitor the network status in real time. If the network delay is too high or the packet loss is serious, promptly investigate the cause of the network failure and take appropriate repair measures. Use data encryption technology to encrypt the feedback data and compress the transmitted data to improve data transmission efficiency and reduce network bandwidth usage. Establish a feedback data repository in the data processing center and classify and store it according to data type, collection time, shield machine location and other information. Use a database management system to manage the data and regularly back up the data repository to prevent data loss. The backup data is stored in an off-site server or storage medium to deal with possible hardware failures, natural disasters, etc. Data analysis and evaluation: after receiving the feedback data, the data processing center will check the integrity of the data to see if there are any missing data, data format errors, etc. For missing data, interpolation methods are used to supplement the data according to the time series characteristics and correlation of the data. For data format errors, data format conversion and correction are performed; the data is filtered to remove noise interference in the data. Digital filtering technology is used to smooth the data for rock and soil particle motion trajectory data to extract the true trend of particle motion. Corresponding filtering is also performed on the shield machine operation parameter data to accurately reflect the actual operating status of the equipment; the particle motion trajectory deviation rate is calculated to determine the reference trajectory, and mathematical methods are used to calculate the difference between the actual trajectory and the reference trajectory. Deviation: Analyze the distribution of deviation rate in different time stages and different spatial regions, find out the time periods and regions with large deviations, provide a basis for subsequent cause analysis, calculate the variance, standard deviation and other statistical quantities of parameters to measure the degree of fluctuation of the stability index of shield machine operation parameters; establish a comprehensive evaluation model, weightedly integrate the particle motion trajectory deviation rate, shield machine operation parameter stability index and other related indicators, and obtain a comprehensive evaluation value that fully reflects the control effect. Determine the weight coefficient of each indicator through the hierarchical analysis method. According to the comprehensive evaluation results and the analysis of various quantitative indicators, conduct in-depth diagnosis of the reasons for poor control effect. Use the fault tree analysis method to gradually check possible causes and determine the key factors, providing a clear direction for subsequent optimization and adjustment; Optimization and adjustment: when the regional rock and soil particle movement trajectory monitoring data is inaccurate, review the geological conditions of the area and re-evaluate the sensor layout plan and simulation analysis; when there are problems with the data processing algorithm, review the optimization algorithm and conduct performance tests; when the physical model or numerical simulation parameters are unreasonable, conduct sensitivity analysis and calibration of the physical model parameters and numerical simulation parameter adjustment experiments; when the control model has poor prediction effect, analyze the control model input and output data and optimize the data acquisition preprocessing link or increase the amount of training data and adjust the model structure parameters and test and verify.

10. A shield tunneling control system based on the movement trajectory of rock and soil particles, characterized in that: include: The shield machine sensor network construction and debugging module designs the sensor network topology according to the shield machine structure and sensor distribution; Intelligent electromagnetic noise monitoring and automatic compensation module, develop intelligent electromagnetic noise monitoring and automatic compensation program, receive electromagnetic environment detection sensor data, monitor electromagnetic noise intensity and frequency in real time and collect them regularly; The precise physical model building module conducts multi-condition combination tests, compares the field data with the indoor test data, and builds and calibrates the physical model of the movement trajectory of rock and soil particles that takes into account multiple factors; Efficient numerical simulation optimization module, introducing parallel framework and developing adaptive meshing algorithm, building data acquisition and transmission system and integrating test optimization, jointly improving the efficiency and accuracy of numerical simulation; Multi-parameter comprehensive monitoring and correlation analysis module, design sensor interfaces and connect them, build a data correlation analysis platform for data standardization, correlation analysis and dimensionality reduction, and build and optimize the shield tunneling status assessment model; Real-time data-driven adaptive control model building module determines the network architecture combination and improves data processing, performs training adjustments, and builds an adaptive control model that can meet the needs of shield tunneling control; The control instruction rapid generation and transmission module optimizes the algorithm flow and designs multi-threading and caching mechanisms. Through network-related configuration, redundant design, self-healing technology, etc., it ensures that the control instructions can be quickly and accurately generated and transmitted to the shield machine actuator; The closed-loop feedback and continuous optimization module formulates feedback data collection plans, maintains equipment networks and safely stores data, organizes and analyzes data and conducts quantitative evaluation and diagnosis of causes. It optimizes and improves issues such as sensor layout, data processing algorithms, physical models and control models, and forms a closed-loop feedback and continuous optimization system for shield tunneling control.

Images