Shield tunnel risk early warning method and system based on digital twinning and federated learning

Abstract

The application discloses a shield tunnel risk early warning method and system based on digital twinning and federated learning, a shield tunnel construction party client, which is an edge computing terminal and acquires basic parameters; a shield machine production party client, the shield tunnel construction party client and a third party client, all of which construct first, second and third shield digital twinning systems according to the basic parameters; a shield construction risk early warning model is arranged in each shield digital twinning system; the shield machine production party client, the shield tunnel construction party client and the third party client are all participants, and each participant sends model parameters of the respective shield construction risk early warning model to a central server; the central server performs an average value operation on the parameters of the three models, and sends the average values of the parameters to each participant; each participant updates the model of the participant according to the average values of the parameters, and realizes shield construction risk early warning based on the updated model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of shield tunnel construction, in particular to a shield tunnel risk early warning method and system based on digital twinning and federated learning. BACKGROUND

[0002] In the process of tunnel construction, shield method has been more and more widely used due to its high degree of automation, high construction efficiency, small influence on the surrounding environment and other advantages. Shield tunnel is developing towards long distance, large section, large burial depth and more complex geological conditions, which puts forward higher requirements and greater challenges to shield tunnel construction.

[0003] The traditional shield construction risk early warning faces the phenomena of network disconnection, power failure and network speed fluctuation in data transmission, and the data transmission is frequently interrupted. The real-time data of shield monitoring pours into the center server, causing network congestion, data packet loss and insufficient timely early warning.

[0004] The traditional shield construction only analyzes the single data of the construction site, does not consider transmitting the multi-source data of the construction site to a unified digital model, and lacks fusion analysis and utilization of geological data and construction data.

[0005] The traditional shield construction relies on manual experience, and the key shield tunneling parameters are determined by the operator, so that the shield construction decision has great subjectivity and randomness. If the parameter adjustment is improper, it will affect the construction progress, or even cause a huge construction accident. SUMMARY

[0006] In order to solve the problems of the prior art, the present application provides a shield tunnel risk early warning method and system based on digital twinning and federated learning;

[0007] On the one hand, a shield tunnel risk early warning method based on digital twinning and federated learning is provided, comprising:

[0008] The shield tunnel construction party client, as an edge computing terminal, acquires geological parameters and construction parameters; the shield tunnel construction party client transmits the acquired data to the shield machine production party client and the third party client after homomorphic encryption processing; the shield machine production party client transmits the shield equipment parameters to the shield tunnel construction party client and the third party client after homomorphic encryption processing; the geological parameters, the construction parameters and the shield equipment parameters are collectively referred to as basic parameters;

[0009] The shield machine production party client, the shield tunnel construction party client and the third party client all construct first, second and third shield digital twinning systems respectively according to the basic parameters; each shield digital twinning system is provided with a shield construction risk early warning model;

[0010] The shield tunnel construction party client, the shield machine manufacturer client, the third party client and the central server jointly constitute a federated learning system; the shield tunnel construction party client, the shield machine manufacturer client and the third party client are participants, and each participant sends model parameters of a respective shield construction risk early warning model to the central server;

[0011] The central server performs an average value operation on the parameters of the three models, and sends the average values of the parameters to each participant; each participant updates the model thereof according to the average values of the parameters, and realizes shield construction risk early warning based on the updated model.

[0012] On the other hand, a shield tunnel risk early warning system based on digital twinning and federated learning is provided, comprising: a shield tunnel construction party client, a shield machine manufacturer client, a third party client and a central server;

[0013] The shield tunnel construction party client, as an edge computing terminal, acquires geological parameters and construction parameters; the shield tunnel construction party client transmits the acquired data to the shield machine manufacturer client and the third party client after homomorphic encryption processing; the shield machine manufacturer client transmits shield equipment parameters to the shield tunnel construction party client and the third party client after homomorphic encryption processing; the geological parameters, the construction parameters and the shield equipment parameters are collectively referred to as basic parameters;

[0014] The shield machine manufacturer client, the shield tunnel construction party client and the third party client each construct a first, a second and a third shield digital twinning system respectively according to the basic parameters; each shield digital twinning system is provided with a shield construction risk early warning model;

[0015] The shield machine manufacturer client, the shield tunnel construction party client, the third party client and the central server jointly constitute a federated learning system; the shield machine manufacturer client, the shield tunnel construction party client and the third party client are participants, and each participant sends model parameters of a respective shield construction risk early warning model to the central server;

[0016] The central server performs an average value operation on the parameters of the three models, and sends the average values of the parameters to each participant; each participant updates the model thereof according to the average values of the parameters, and realizes shield construction risk early warning based on the updated model.

[0017] The above technical solution has the following advantages or beneficial effects:

[0018] (1) The edge computing system in the application is deployed on the network edge close to the data source, directly processes data near the data source, including shield tunnel non-excavation state data deletion, abnormal data correction, noise processing, transmits necessary data to the data cloud platform, reduces bandwidth demand and transmission cost, reduces transmission time, and ensures faster response speed.

[0019] (2) The shield digital twin system of the application can accurately reflect the state and behavior of the shield machine entity by constructing the mapping relationship between the physical world and the digital space, adopts various machine learning and deep learning algorithms for data analysis and mining, realizes functions such as stratum parameter prediction and excavation parameter optimization, provides strong data-driven capability for simulation, prediction and optimization of the complex system of the shield machine, and accelerates the process of intelligent construction and construction automation.

[0020] (3) The federal learning system in the application provides a platform for mutual communication and learning for different shield machine manufacturers, tunnel construction parties, research institutions and colleges and universities, learns model parameters and experience of all parties under the premise of protecting the data privacy of all parties, and improves the prediction and simulation capabilities of the local digital twin system.

[0021] The shield machine manufacturer, the tunnel construction party and the research institution and the college and university sign a compliant smart contract, use encrypted engineering data from the construction party and encrypted equipment data from the shield machine manufacturer.

[0022] For the shield machine manufacturer, by analyzing the equipment state of the entire construction interval and integrating important learning experience into the subsequent product design stage, the shield machine considers the running state of the whole life cycle in the design stage.

[0023] For different construction parties, by learning model parameters of different projects and integrating more rich construction experience, the shield machine's ability to cope with complex geological environment is improved, and the intelligent development of shield construction is promoted.

[0024] For research institutions and colleges and universities, by strengthening learning and communication with other parties, specific solutions to on-site problems can be proposed, more intelligent and scientific prediction models can be developed, and more efficient decision guidance can be provided for on-site shield construction decision-making. BRIEF DESCRIPTION OF DRAWINGS

[0025] The drawings accompanying the specification of the application form part of the application and serve to provide a further understanding of the application, the illustrative embodiments of the application and their description serve to explain the application without constituting an improper limitation thereof.

[0026] Figure 1 Conceptual diagram of the federal learning system in the embodiment of the application.

[0027] Figure 2 The edge computing system composition framework diagram in the embodiment of the application.

[0028] Figure 3 The edge computing system data processing and transmission flow chart in the embodiment of the application.

[0029] Figure 4 The digital twin system module function framework diagram in the embodiment of the application.

[0030] Figure 5 The overall flow chart in the embodiment of the application. DETAILED DESCRIPTION

[0031] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the application pertains.

[0032] The wide application of digital twin technology, which can accurately reflect the state and behavior of the shield machine entity by constructing the mapping relationship between the physical world and the digital space, provides strong data-driven capabilities for the simulation, prediction and optimization of the complex system of the shield machine, and accelerates the process of intelligent construction and construction automation. The effective implementation of digital twin technology depends on a large amount of real-time data, which comes from sensors installed on the shield machine and the surrounding construction environment. Since shield construction adopts all-weather construction operation mode, sensors usually collect data in seconds, and a large amount of raw data is generated during the construction process, which inevitably contains abnormal data and noise. The two processes of tunneling construction and segment assembly are alternately performed, and the time used is close. The downtime data in the segment assembly process will also be collected by the sensor, which has little significance for model establishment. If all the data collected on site is directly transmitted to the cloud platform, the cloud computing platform will face high network latency, mass device access, mass data processing difficulty, insufficient bandwidth, and high power consumption and other high-difficulty challenges; the existence of a large amount of downtime data, abnormal data and noise not only increases the data transmission pressure, but also causes certain interference to subsequent analysis and modeling.

[0033] In order to solve the disadvantages of high latency and lack of real-time data analysis capability under the traditional data processing mode, edge computing technology emerges as the times require. Edge computing technology is to provide edge intelligent services in the network edge close to the data source through the fusion of network, computing, storage and application core capabilities of the distributed open platform. Edge computing directly processes data at the data source, only transmits necessary data to the cloud platform, reduces bandwidth demand and transmission cost, reduces transmission time, and ensures faster response speed.

[0034] Shield construction is facing more and more complex geological environment, complete historical data and valuable construction experience are extremely important to ensure construction safety and improve construction efficiency, but due to industry competition, privacy security, complex administrative procedures and other problems, the integration of data between the parties involved in tunnel construction (such as shield machine design and production parties, construction parties, research institutions, etc.) faces many obstacles, and the data between parties exists in the form of data island. Federated learning enables multiple participants to continue machine learning while protecting data privacy and meeting legal and regulatory requirements, solving the problem of data island. In the process of machine learning, each participant can use the data of other participants to jointly model, establish a shared machine learning model, and adjust their own model parameters based on the shared model to improve the accuracy of the model.

[0035] Embodiment one

[0036] The embodiment provides a shield tunnel risk early warning method based on digital twinning and federated learning;

[0037] As shown in Figure 5 The shield tunnel risk early warning method based on digital twinning and federated learning comprises:

[0038] S101: The shield tunnel construction party client, as an edge computing terminal, acquires geological parameters and construction parameters; the shield tunnel construction party client transmits the acquired data to the shield machine production party client and the third party client after homomorphic encryption processing; the shield machine production party client transmits shield equipment parameters to the shield tunnel construction party client and the third party client after homomorphic encryption processing; the geological parameters, the construction parameters and the shield equipment parameters are collectively referred to as basic parameters;

[0039] S102: The shield machine production party client, the shield tunnel construction party client and the third party client each construct a first, a second and a third shield digital twinning system according to the basic parameters; each shield digital twinning system is provided with a shield construction risk early warning model;

[0040] S103: The shield machine production party client, the shield tunnel construction party client, the third party client and the central server jointly constitute a federated learning system; the shield machine production party client, the shield tunnel construction party client and the third party client are participants, and each participant sends the model parameters of the shield construction risk early warning model to the central server;

[0041] S104: The central server performs an average value operation on the parameters of the three models, and sends the average value of the parameters to each participant; each participant updates its own model according to the average value of the parameters, and realizes shield construction risk early warning based on the updated model.

[0042] Further, the geological parameter refers to geological information of a tunnel design interval obtained through geological exploration and advanced geological prediction technology; the construction parameter refers to real-time tunneling parameters such as a pushing speed, a pushing force, a cutter head torque, and a penetration degree of the shield machine; and the shield machine equipment parameter refers to detailed equipment parameters such as a cutter head diameter, a cutter type and an installation position, and a pushing cylinder quantity.

[0043] It should be understood that the shield machine construction parameter is collected through a sensor. The geological parameter is collected and summarized through pre-construction geotechnical exploration and advanced geological prediction during construction. The shield machine equipment parameter is provided by a shield machine manufacturer. The sensor and the edge computing client are connected through a communication network.

[0044] As shown in Figure 2 The edge computing terminal pushes computing resources and services from a cloud data center to the network edge, i.e., closer to the data source. This architecture can improve data processing speed, reduce latency, enhance privacy protection, and reduce bandwidth usage. The edge computing terminal includes a physical layer, an information layer, and a communication layer.

[0045] The physical layer is an actual hardware device deployed at the network edge, including various sensors, edge servers, storage devices, and the like arranged in the shield machine and the surrounding stratum environment. The physical layer is responsible for the collection of multi-source heterogeneous data and is also a running platform of the information layer.

[0046] The information layer is a data processing center in the edge computing system and contains various software applications and services such as data processing algorithms. The information layer depends on the physical layer and is responsible for data processing, analysis, and storage.

[0047] The communication layer is responsible for data transmission and communication in the edge computing system and includes network protocols, interfaces, routers, switches, and other network devices. The communication layer is responsible for ensuring data flow between the physical layer and the information layer and between the information layer and the cloud data center, realizing fast transmission and synchronization of data.

[0048] The information layer processes and analyzes the data collected by the physical layer, including shield machine non-tunneling state data deletion, abnormal data correction, and noise processing, to realize the simplification of massive raw data.

[0049] Further, as shown in Figure 3 The S101 further includes:

[0050] The shield tunnel construction party client pre-processes the shield machine working parameters. The pre-processing includes:

[0051] Step (11) retains parameters of the shield machine in a tunneling state and deletes parameters of the shield machine in a non-tunneling state.

[0052] Step (12) corrects the abnormal data;

[0053] Step (13) processes the noise data.

[0054] Further, the step (11) retains the parameters of the shield machine in the tunneling state and deletes the parameters of the shield machine in the non-tunneling state, specifically including:

[0055] The non-tunneling state data deletion is realized by a binary discriminant function, and the on-site construction condition is analyzed. When the push speed, cutterhead torque and cutterhead speed appear 0 value, the shield machine is in a shutdown state, so the tunneling parameters in the shutdown state are deleted to reduce the data transmission pressure. Because the tunneling parameters at this time have little influence on subsequent analysis and learning, and even will interfere with the model accuracy;

[0056] ; ; ;

[0057] wherein, is the cutterhead torque, is the cutterhead speed, is the tunneling speed, is the product of the cutterhead torque, the cutterhead speed and the tunneling speed, is the original data set, is the data set of normal tunneling.

[0058] It should be understood that the non-tunneling state data deletion is realized by a binary discriminant function, because in the shield construction process, forward tunneling and segment assembly are operated in a series manner. When the tunneling reaches the width of a ring, the shield machine stops and starts segment assembly. At this time, the sensor still works according to the set frequency, and many shield machine tunneling related parameters are zero value, which is meaningless for analysis. The zero value tunneling parameters can be deleted to reduce the data transmission pressure.

[0059] Further, the step (12) corrects the abnormal data, specifically by using the isolated forest algorithm to find the abnormal value, and using the average value of the adjacent data of the abnormal value as the replacement data of the abnormal value.

[0060] It should be understood that the abnormal data correction is realized by the isolation forest algorithm, which is a tree-based machine learning algorithm that "isolates" observations by randomly selecting features and randomly selecting the split value of the feature to realize anomaly detection. The expert experience is combined with statistical analysis of the abnormal values selected by the isolation forest, and the abnormal threshold of the isolation forest algorithm is appropriately adjusted to ensure the accuracy and reasonableness of the abnormal data processing, and to avoid misjudging normal data as abnormal. In order to ensure the integrity of the data stream, the average value of adjacent data is used to replace the abnormal data.

[0061] Further, the step (13) processes the noise data by a weighted average filtering algorithm, which gives higher weights to the nearest points and lower weights to the farther points, retains as much information as possible of the current state, and improves the accuracy and reliability of noise processing.

[0062] It should be understood that the noise processing is realized by the weighted moving average filtering technology, which is a signal processing technology used to smooth time series data, reduce the influence of noise, and retain the trend information of the data.

[0063] The edge computing terminal provides data support for the shield digital twin system. The digital twin system updates the model and learns data using the transmitted data, and provides online decision guidance for the site construction.

[0064] It should be understood that the first, second and third shield digital twin systems are established according to the geological parameters, construction parameters and shield machine equipment parameters. The first, second and third shield digital twin systems can analyze the data from the edge computing system in real time, and update the first, second and third shield digital twin systems in time. Deeply mine the key parameters such as tunneling speed, shield posture and settlement deformation, predict equipment state and stratum condition through various machine learning and deep learning algorithms, and timely repair and replace cutters and reasonably adjust construction strategy. At the same time, the first, second and third digital twin systems have parameter prediction and optimization modules, which make the shield machine work according to the optimized parameters through the automatic control module, and provide more scientific and efficient tunneling construction scheme.

[0065] According to the geological survey results, a three-dimensional geological modeling analysis software EVS (Earth Volumetric Studio) software is established in the entire tunnel design interval construction geological three-dimensional model, the stratum interface visualization is realized through the analysis and interpolation of the drilling information, the model is imported into Unity, and the stratum parameter information in the three-dimensional geological model is displayed, such as the compressive strength, bearing capacity and natural density of the rock. The above parameters are derived from field tests and indoor tests. According to the shield machine design drawing, a shield machine model is established in 3DMAX, the shield machine model is imported into the Unity system where the three-dimensional geological model is located, the tunneling simulation of the shield machine in the three-dimensional geological model is realized, the data interaction between the edge computing system and the above model is established, the shield digital twin system is established, and the real-time synchronization of the construction dynamics and the digital twin model is realized.

[0066] Further, the shield machine manufacturer client, the shield tunnel construction client and the third party client each construct a first, second and third shield digital twin system according to the basic parameters, specifically including:

[0067] The shield machine manufacturer client constructs a first shield digital twin system;

[0068] The shield tunnel construction client constructs a second shield digital twin system;

[0069] The third party client constructs a third shield digital twin system.

[0070] Further, as shown in Figure 4 The first shield digital twin system includes a first data storage module, a first data encryption module, a first stratum parameter prediction module, a first ground settlement prediction module, a first cutter wear prediction module, a first shield attitude deviation prediction module, a first tunneling parameter prediction module, a first tunneling parameter optimization module, a first shield construction risk early warning module, a first automatic control module and a first visualization module.

[0071] The second shield digital twin system includes a second data storage module, a second data encryption module, a second stratum parameter prediction module, a second ground settlement prediction module, a second cutter wear prediction module, a second shield attitude deviation prediction module, a second tunneling parameter prediction module, a second tunneling parameter optimization module, a second shield construction risk early warning module, a second automatic control module and a second visualization module.

[0072] The third shield digital twin system comprises a third data storage module, a third data encryption module, a third stratum parameter prediction module, a third ground settlement prediction module, a third cutter wear prediction module, a third shield posture deviation prediction module, a third tunneling parameter prediction module, a third tunneling parameter optimization module, a third shield construction risk early warning module, a third automatic control module and a third visualization module.

[0073] Further, the first, second and third data storage modules are configured to store real-time data streams from the shield tunnel construction party client and update the shield digital twin model.

[0074] Further, the first, second and third data encryption modules are configured to encrypt data required by the federal learning contract party using a homomorphic encryption algorithm. The homomorphic encryption supports calculation and learning on encrypted data, protects data privacy while ensuring learning effectiveness, and provides security for cross-organizational data cooperation and sharing. The shield machine manufacturer client provides basic parameters of the shield equipment, and the shield tunnel construction party client provides geological parameters and construction parameters. The encrypted data is shared with other participants. The data encryption module has a decryption function, and each participant uses decryption technology to process encrypted data, including basic parameters from other participants and model parameters from the central server.

[0075] Further, the first, second and third stratum parameter prediction modules are configured to establish a database based on geological survey data and shield machine historical tunneling data. The geological survey data is used to obtain stratum category labels and surrounding rock grade labels. Tunneling parameters such as cutterhead speed and penetration are selected for model training. The trained model is used to realize stratum category prediction and surrounding rock grade prediction.

[0076] Further, the stratum category prediction is realized based on the trained support vector machine. In the training process of the support vector machine, the input values of the support vector machine are cutterhead torque, cutterhead speed, thrust, penetration, tunneling speed, torque penetration index (TPI) and field penetration index (FPI) values, and rock uniaxial compressive strength. The output value of the support vector set is the stratum category. The stratum category specifically includes: super-soft stratum, relatively homogeneous soft rock, fully weathered soft and hard uneven stratum, strongly weathered and moderately weathered soft and hard uneven stratum, medium-soft stratum, soft and hard uneven soft stratum, soft and hard uneven hard stratum, medium-hard stratum, relatively homogeneous hard rock, and super-hard stratum. When the loss function value of the support vector machine reaches the minimum value or the number of iterations reaches the set number, the training is stopped, and the trained support vector machine is obtained.

[0077]

[0078] In the formula: T is the cutterhead torque; P is the penetration;

[0079]

[0080] F is the thrust force.

[0081] It should be understood that, since the support vector machine (SVM) is a commonly used classification algorithm, it can be used to process multi-class classification problems, and the SVM algorithm is selected for stratum category prediction. According to the early geological drilling data and the historical tunneling parameters of the shield machine, a database for establishing a stratum category prediction model is established, the cutter torque, the cutter rotating speed, the thrust force, the penetration, the tunneling speed, the TPI, the FPI, and the uniaxial compressive strength of the rock are selected as the input variables of the SVM model, and the super-soft and weak stratum, the relatively homogeneous soft rock, the fully weathered soft and hard uneven stratum, the strong weathered and medium weathered soft and hard uneven stratum, the medium-soft stratum, the soft and hard uneven soft stratum, the soft and hard uneven hard stratum, the medium-hard stratum, the relatively homogeneous hard rock, and the super-hard stratum are selected as the output variables of the SVM model. The SVM model is trained by using the established database to obtain a stratum category prediction model based on the SVM algorithm, and in the shield digital twin system, real-time tunneling parameters are input to realize accurate prediction of 10 kinds of stratum categories.

[0082] Further, the surrounding rock grade prediction is realized based on the trained multi-layer perceptron, and in the training process, the input values of the multi-layer perceptron are the disc rotating speed, the penetration, the support shoe pressure, the support shoe pump pressure, the support shoe pitch angle, the control pump pressure, the advancing speed, and the support shoe rolling angle; the output values of the multi-layer perceptron are the surrounding rock grades I, II, III, IV, and V; when the loss function value of the multi-layer perceptron reaches the minimum value or the iteration number reaches the set number, the training is stopped to obtain the trained multi-layer perceptron.

[0083] It should be understood that, since the multi-layer perceptron (MLP) is a classification algorithm based on neural networks, it can learn complex nonlinear relationships and process multi-class classification tasks, and therefore the MLP algorithm is selected for surrounding rock grade prediction. According to the early geological survey data and the historical tunneling parameters of the shield machine, a database for establishing a stratum category prediction model is established, the disc rotating speed, the penetration, the support shoe pressure, the support shoe pump pressure, the support shoe pitch angle, the control pump pressure, the advancing speed, and the support shoe rolling angle are selected as the input variables of the MLP model, and the surrounding rock grades I, II, III, IV, and V are selected as the output variables of the model. The MLP model is trained by using the established database to obtain a surrounding rock grade prediction based on the MLP algorithm, and in the shield digital twin system, real-time tunneling parameters are input to realize accurate prediction of 5 kinds of surrounding rock grades.

[0084] Further, the first, second, and third ground surface settlement prediction modules realize the prediction of the ground surface settlement by using the trained CNN-BiLSTM model.

[0085] The CNN-BiLSTM model comprises a CNN network and a BiLSTM network connected in sequence.

[0086] In the training process, the input values of the CNN-BiLSTM model are the bearing capacity, the natural unit weight, the tunnel depth, the distance between the monitoring point and the cutter head, the penetration, the thrust force, the incision pressure, the working cabin pressure, the cutter head speed and the cutter head torque; the output value of the CNN-BiLSTM model is the ground settlement prediction result; when the loss function value of the CNN-BiLSTM model reaches the minimum value or the iteration number reaches the set number, the training is stopped, and the trained CNN-BiLSTM model is obtained.

[0087] It should be understood that the convolutional neural network is introduced to improve the bidirectional long short-term memory neural network, better capture the local features of the ground settlement data, and establish a ground settlement prediction model based on the CNN-BiLSTM algorithm. According to the geological survey report, the historical tunneling data of the shield and the ground settlement monitoring value, a database is established, and the bearing capacity, the natural unit weight, the tunnel depth, the distance between the monitoring point and the cutter head, the penetration, the thrust force, the incision pressure, the working cabin pressure, the cutter head speed and the cutter head torque are selected as the input parameters of the ground settlement prediction model, and the ground settlement value is the output parameter of the prediction. For a specific monitoring point, the distance between the monitoring point and the cutter head is constantly changing, specifically: before the shield passes through, the distance between the monitoring point and the cutter head constantly decreases, and the distance at this time is set as a positive value; when the cutter head of the shield reaches the section where the monitoring point is located, the distance is 0; after the cutter head of the shield passes through the section, the distance constantly increases, but in order to facilitate the distinction between the two different states before and after the passing, the distance is set as a negative value, and the negative sign is only used to distinguish the two position states. The established database is used for model training to obtain the ground settlement value prediction based on the CNN-BiLSTM algorithm, and in the shield digital twin system, the real-time tunneling parameters and the bearing capacity, the natural unit weight and the tunnel depth value of the stratum where the shield is located in the digital twin system are input to realize the ground settlement prediction of different positions before, during and after the shield passing.

[0088] Further, the first, second and third tool wear prediction modules predict tool wear based on the trained BP neural network.

[0089] In the training process, the BP neural network establishes a database depending on the historical tunneling parameters of the shield, the stratum parameter information and the tool wear information, selects the cutter head torque, the cutter head speed, the thrust force, the tunneling speed, the penetration, the cutter diameter, the installation radius, the cutter spacing, the rock uniaxial compressive strength, the stratum category and the surrounding rock grade as the input values; the output value of the BP neural network is the tool wear amount; when the loss function value of the BP neural network reaches the minimum value or the iteration number reaches the set number, the training is stopped, and the trained BP neural network is obtained.

[0090] It should be understood that the tool wear prediction model is established based on the BP neural network algorithm, a database is established according to shield historical tunneling parameters, stratum parameter information and tool wear information, the cutter torque, the cutter rotating speed, the advancing force, the tunneling speed, the penetration, the cutter diameter, the installation radius, the cutter spacing, the rock uniaxial compressive strength, the stratum category and the surrounding rock grade are selected as the input features, and the tool wear amount is the output variable. The tool wear prediction model is learned and trained based on the historical database, the tool wear prediction model based on the BP neural network is established, the tool wear condition is predicted in real time, the residual life of the cutter is predicted according to the tool wear trend referring to the cutter replacement standard, the cutter is repaired and replaced in time, and the construction progress is ensured.

[0091] Further, the first, second and third shield posture deviation prediction modules are based on the trained long short-term memory neural network model LSTM to realize the prediction of the shield posture deviation.

[0092] In the training process of the model, the input values of the long short-term memory neural network model LSTM are the cutter torque, the cutter rotating speed, the tunneling speed, the advancing force, the upper and lower cylinder thrust difference, the cohesion, the internal friction angle, the lateral pressure coefficient and the tunnel depth; the output values are the incision horizontal deviation, the incision vertical deviation, the shield tail horizontal deviation and the shield tail vertical deviation; when the loss function value of the model decreases to the minimum value or the iteration number reaches the set number of times, the training is stopped, and the trained long short-term memory neural network model LSTM is obtained.

[0093] It should be understood that first, the shield posture deviation prediction is realized based on the long short-term memory neural network model (LSTM for short), the cutter torque, the cutter rotating speed, the tunneling speed, the advancing force, the upper and lower cylinder thrust difference, the cohesion, the internal friction angle, the lateral pressure coefficient and the tunnel depth are selected as the input variables of the posture deviation prediction model, and the output variables include the incision horizontal deviation, the incision vertical deviation, the shield tail horizontal deviation and the shield tail vertical deviation.

[0094] Further, the first, second and third tunneling parameter prediction modules are based on the trained XGBoost model to realize the tunneling parameter prediction.

[0095] In the training process, the input variables of the XGBoost model are the cutter extrusion force, the penetration, the incision pressure, the tunneling speed, the grouting pressure, the grouting flow, the grout discharge flow, the stratum category, the surrounding rock grade, the torque cutting depth index (TPI) and the field cutting depth index (FPI) values, and the output variables of the XGBoost model are the cutter torque, the cutter rotating speed and the total thrust; when the loss function value of the XGBoost model decreases to the minimum value or the iteration number reaches the set number of times, the training is stopped, and the trained XGBoost model is obtained.

[0096] wherein the stratum category and the surrounding rock grade are from output values of a corresponding stratum parameter prediction module.

[0097] Further, the first, second, and third tunneling parameter optimization modules select the cutterhead rotating speed, cutterhead torque, total thrust, and penetration as optimization variables based on multi-objective optimization theory, and realize multi-objective optimization of maximum tunneling speed and minimum tunneling specific energy by using the NSGA-II algorithm. The multi-objective optimization framework is as follows:

[0098] ;

[0099] ; ;

[0100] ; ;

[0101] ; ;

[0102] wherein, is the tunneling specific energy (kJ / m3), n is the cutterhead rotating speed (r / min), T is the cutterhead torque (MN·m), F is the total thrust (unit: MN), R is the cutterhead radius of the shield machine (m), and P is the shield cutterhead penetration (mm / rot).

[0103] is the objective function of the model, and the model has two objectives, including minimizing the tunneling specific energy and the tunneling speed The tunneling specific energy can be directly calculated through real-time cutterhead penetration and other parameters, and the tunneling speed does not have a direct calculation formula, and thus a tunneling speed prediction model based on XGBoost needs to be established, with the cutterhead rotating speed, total thrust, cutterhead torque, and cutterhead penetration as input parameters and the tunneling speed as output parameter. Due to large differences in tunneling parameters under different geological conditions, the cutterhead penetration is taken as an example. The cutterhead penetration in hard rock stratum can reach 30 mm / rot, while the cutterhead penetration in soft soil stratum generally fluctuates around 2 mm / rot, and thus the maximum and minimum values of the parameters in the constraint condition need to be determined according to specific engineering characteristics.

[0104] The preliminary settings of the NSGA-II algorithm parameters are as follows: population size 250, maximum iteration number 100, crossover probability 0.8, and mutation probability 0.1, which can be adjusted according to specific conditions. A multi-objective optimization model of shield tunneling parameters based on the NSGA-II algorithm is established to obtain the optimal solution. Since the values of the optimization variables such as the cutterhead rotating speed, total thrust, cutterhead torque, and cutterhead penetration are in one-to-one correspondence with the optimal values of the objectives (tunneling speed and tunneling specific energy), the corresponding cutterhead rotating speed, total thrust, cutterhead torque, and cutterhead penetration are displayed in the digital twin system, and the construction parameters are controlled within the optimization range.

[0105] Further, the first, second, and third shield construction risk early warning modules are used to give early warning results for various construction risks in the construction process. The construction risk early warning module includes: geological risk early warning, equipment risk early warning, and environmental risk early warning.

[0106] Geological risk early warning. When the shield machine is physically constructed and excavated, the shield machine digital twin is simultaneously performed in the digital space. Since the three-dimensional geological information of the construction area has been imported into the shield digital twin system during modeling, the geological information in front of the excavation face can be intuitively displayed in the digital twin system. Based on the prediction results of the stratum parameter prediction module and the visual display of the geological characteristics in front of the excavation face, early warning information is given before the geological conditions change. For example, when weak surrounding rock and fault fracture zones are encountered, an alarm is given in advance to avoid construction accidents.

[0107] Equipment risk early warning. Based on the cutter wear prediction module, the cutter state is monitored, faults and abnormalities are predicted, and measures are taken in advance to avoid risks. At the same time, the shield digital twin model integrates real-time data of the shield machine. Based on the visualization module, the running state of key equipment such as main bearings and hydraulic systems can be monitored. When the equipment parameters exceed the rated value or are not working abnormally, the equipment risk early warning module gives an early warning signal and gives the abnormal position.

[0108] Environmental risk early warning. It includes ground settlement early warning and toxic gas early warning. The ground settlement early warning is based on the ground settlement deformation module, which can realize the settlement value prediction of the shield before, during, and after the crossing for a long time. The ground settlement value trend at this position is predicted in advance during construction. If it exceeds the specified value, an early warning signal is given, and construction adjustment (such as grouting reinforcement) is performed to avoid excessive settlement in the later period and reduce the impact on the surrounding environment. The harmful gas early warning function relies on the gas monitoring sensor carried by the shield machine. Real-time analysis of sensor information is performed to focus on the concentration of harmful gases such as methane, carbon monoxide, and hydrogen sulfide. When the concentration changes abnormally, reasonable ventilation and exhaust measures are taken to ensure construction safety and personnel health.

[0109] Safety accident warning, shield entity and supporting system are equipped with monitoring equipment, combined with on-site monitoring video, when a safety accident occurs, the accident is located according to the monitoring information, timely rescue and reduce the loss of the accident. A large amount of video or image data is collected through the monitoring camera, and these data are labeled, especially the injured scene and the posture of the personnel. Through data enhancement (such as rotation, scaling, adjusting brightness, etc.) to enrich the database, ensure that the model has better robustness to different environments. The human pose estimation technology OpenPose is adopted, and the key points of human skeleton are detected to judge whether there is posture abnormality, such as falling or twisting. After the model training is completed, it can be deployed in the monitoring system to analyze the monitoring picture in real time and detect whether someone is injured. When the system detects abnormal behavior or injury, it can trigger the alarm mechanism to notify the relevant personnel to take emergency measures. The alarm conditions can be set according to the posture abnormality, behavior abnormality and other characteristics. At the same time, the system needs to continuously collect new data for model fine-tuning to adapt to the changing monitoring environment and improve the detection accuracy. In order to reduce false positives, the different perspectives of multiple monitoring cameras can be combined to enhance the judgment ability of the system.

[0110] Further, the first, second and third automatic control modules, the automatic control module relies on the Internet of Things to realize the control of the shield digital twin system on the shield entity, in order to fully utilize the tunneling parameter optimization module function, the automatic control module converts the optimization strategy into control instructions, the control instructions are in a form that can be recognized by the shield machine, such as plc control, pid control, which are issued to the controller of the shield entity through the Internet of Things, to realize accurate control and optimized operation of the shield entity, real-time adjustment of construction parameters and improvement of construction efficiency.

[0111] Further, the first, second and third visualization modules are used to display the established shield digital twin model, which is closely integrated with all the above-mentioned function modules, and the real-time collected construction parameters and the previously constructed shield digital twin model parameters are displayed in the form of charts, and the prediction and optimization results of each function module are visualized.

[0112] In order to fully utilize the construction experience of different units and improve the diversity of data, the concept of federated learning is introduced, which learns the model parameters and experience of each party under the premise of protecting the privacy of each party. The federated learning system mainly includes a plurality of shield digital twin models established on the local of the participating parties and the corresponding edge control system, and a central server.

[0113] The participants establish unique shield digital twin models and systems according to respective construction characteristics, geological conditions, shield parameters and the like. The shield digital twin system collects data from the edge control system and has the above-mentioned data storage, tunneling parameter prediction, tunneling parameter optimization and the like.

[0114] The federal learning method is a continuous and dynamic learning and updating process between the participants and the central server. First, the participants transmit the digital twin model update parameters and gradients to the central processor through homomorphic encryption. Next, the central processor processes the encrypted data by decryption and aggregation and updates the global model, and transmits the new parameters and gradient information to the participants through homomorphic encryption. Finally, the participants adjust the local digital twin system according to the latest information and compare and analyze with the measured values. The above process is repeated, and through the joint participation of multiple parties in federal learning, the construction of the shield digital twin model is optimized, and the prediction and simulation capabilities of the digital twin system of each party are continuously enhanced, and the tunneling efficiency of the shield machine is continuously improved.

[0115] As shown in Figure 5 , it is understood that the S103: the shield machine manufacturer client, the shield tunnel construction client, the third party client and the central server jointly constitute a federal learning system; the federal learning system is established, the central server is set, the participants of the federal learning are determined, and the connection between the participants and the central server is constructed.

[0116] The participants who do not have direct construction data such as scientific research institutions and shield machine manufacturers can sign a legal and compliant smart contract with the construction unit, determine the data transmission protocol, encrypt the required data through the data encryption module in the shield digital twin system and transmit it.

[0117] The participants establish unique shield digital twin models and systems according to respective construction characteristics, geological conditions, shield parameters and the like. The shield digital twin system collects data from the shield tunnel construction client and has the above-mentioned data storage, tunneling parameter prediction, tunneling parameter optimization and the like.

[0118] Further, the shield machine manufacturer client, the shield tunnel construction client and the third party client are all participants, and each participant sends the model parameters of the respective shield construction risk early warning model to the central server, including: the participants transmit the model update and gradient to the central server through homomorphic encryption.

[0119] Further, as shown in Figure 1As shown, S104: the central server performs an averaging operation on the parameters of the three models, and distributes the average values of the parameters to each participant, including:

[0120] The central server is responsible for receiving and decrypting the model update parameters and gradients from each participant, verifying the integrity and correctness of the data, and avoiding data damage or malicious tampering during transmission. The central server uses a weighted average aggregation method to give higher weights to model parameters with larger data volume and higher prediction accuracy, constructs a new global model, and distributes the new parameters and gradient information to each participant through homomorphic encryption processing.

[0121] Further, each participant updates its own model according to the average value of the parameters, and realizes the shield construction risk early warning based on the updated model, including:

[0122] The participant receives and decrypts the latest model data and gradient information distributed by the central server, and integrates the information into the local shield construction risk early warning model. After multiple debugging and testing, the accuracy of the local shield construction risk early warning model is verified.

[0123] S103-S104 is a continuous cycle process. Through the joint participation of multiple parties in federated learning, the construction of the shield digital twin model is optimized, the prediction and simulation capabilities of each party's digital twin system are continuously enhanced, and the tunneling efficiency of the shield machine is continuously improved. When the accuracy of the local digital twin system of the participant reaches the expected value, it can exit on its own, and other participants can continue to learn the task.

[0124] The edge computing system provides data support for the shield digital twin system. The digital twin system uses the transmitted data for model updating and data learning to provide online decision guidance for on-site construction.

[0125] The shield digital twin system can analyze the data from the edge computing system in real time, deeply mine key parameters such as tunneling speed, shield posture, and settlement deformation, predict equipment state and ground conditions through various machine learning and deep learning algorithms, and timely repair and replace cutters and reasonably adjust construction strategies. At the same time, the digital twin system has a parameter prediction and optimization module, which makes the shield machine work according to the optimized parameters through the automatic control module, providing more scientific and efficient tunneling construction schemes.

[0126] The federated learning system can break through the limitation of data silos, strengthen the connection between shield machine manufacturers, tunnel construction parties, research institutions and colleges and universities, fully utilize the parameter information of each digital twin model, improve the prediction accuracy and model generalization ability of each participant's local digital twin system, and promote the scientific, efficient and intelligent development of shield construction.

[0127] Embodiment Two

[0128] The embodiment provides a shield tunnel risk early warning system based on digital twinning and federated learning;

[0129] The shield tunnel risk early warning system based on digital twinning and federated learning comprises a shield tunnel construction party client, a shield tunnel construction party client, a third party client and a central server.

[0130] The shield tunnel construction party client is an edge computing terminal, and obtains geological parameters and construction parameters; the shield tunnel construction party client transmits the obtained data to the shield machine production party client and the third party client after homomorphic encryption processing; the shield machine production party client transmits shield equipment parameters to the shield tunnel construction party client and the third party client after homomorphic encryption processing; the geological parameters, the construction parameters and the shield equipment parameters are collectively referred to as basic parameters.

[0131] The shield machine production party client, the shield tunnel construction party client and the third party client all construct first, second and third shield digital twinning systems respectively according to the basic parameters; a shield construction risk early warning model is arranged in each shield digital twinning system.

[0132] The shield machine production party client, the shield tunnel construction party client, the third party client and the central server jointly constitute a federated learning system; the shield machine production party client, the shield tunnel construction party client and the third party client are participants, and each participant sends model parameters of the shield construction risk early warning model of the participant to the central server.

[0133] The central server performs an average value operation on the parameters of the three models, and sends the average values of the parameters to each participant; each participant updates the model of the participant according to the average values of the parameters, and realizes shield construction risk early warning based on the updated model.

[0134] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A shield tunnel risk early warning method based on digital twinning and federated learning, characterized in that, The shield tunnel construction party client, as an edge computing terminal, obtains geological parameters and construction parameters. The shield tunnel construction party client transmits the obtained data to the shield machine manufacturer client and the third party client after homomorphic encryption processing; the shield machine manufacturer client transmits shield equipment parameters to the shield tunnel construction party client and the third party client after homomorphic encryption processing; and the geological parameters, the construction parameters and the shield equipment parameters are collectively referred to as basic parameters. The shield machine manufacturer client, the shield tunnel construction party client and the third party client each construct a first, a second and a third shield digital twin system according to the basic parameters; and each shield digital twin system is provided with a shield construction risk early warning model. The shield machine manufacturer client, the shield tunnel construction party client, the third party client and the central server jointly constitute a federated learning system; the shield machine manufacturer client, the shield tunnel construction party client and the third party client are participants, and each participant sends model parameters of the shield construction risk early warning model to the central server. The central server averages the parameters of the three models, and sends the average values of the parameters to each participant; each participant updates the model according to the average values of the parameters, and realizes shield construction risk early warning based on the updated model. The shield machine manufacturer client, the shield tunnel construction party client and the third party client each construct a first, a second and a third shield digital twin system according to the basic parameters, specifically including: the shield machine manufacturer client constructs a first shield digital twin system; the shield tunnel construction party client constructs a second shield digital twin system; and the third party client constructs a third shield digital twin system.

2. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 1, characterized in that, The first shield digital twin system includes: a first data storage module, a first data encryption module, a first stratum parameter prediction module, a first ground settlement prediction module, a first cutter wear prediction module, a first shield attitude deviation prediction module, a first tunneling parameter prediction module, a first tunneling parameter optimization module, a first shield construction risk early warning module, a first automatic control module and a first visualization module. The second shield digital twin system includes: a second data storage module, a second data encryption module, a second stratum parameter prediction module, a second ground settlement prediction module, a second cutter wear prediction module, a second shield attitude deviation prediction module, a second tunneling parameter prediction module, a second tunneling parameter optimization module, a second shield construction risk early warning module, a second automatic control module and a second visualization module. The third shield digital twin system includes: a third data storage module, a third data encryption module, a third stratum parameter prediction module, a third ground settlement prediction module, a third cutter wear prediction module, a third shield attitude deviation prediction module, a third tunneling parameter prediction module, a third tunneling parameter optimization module, a third shield construction risk early warning module, a third automatic control module and a third visualization module. ​ 3. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second and third data encryption modules adopt a homomorphic encryption algorithm to encrypt data required by a federal learning contract party, and the homomorphic encryption supports calculation and learning on the encrypted data; the data encryption module is also configured to decrypt the encrypted data.

4. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second and third stratum parameter prediction modules establish a database according to geological survey data and shield tunneling machine historical tunneling data, the geological survey data are used to obtain stratum category labels and surrounding rock grade labels, and the cutter head rotating speed and penetration degree tunneling parameters are selected for model training, and the trained model is used to realize stratum category prediction and surrounding rock grade prediction. The stratum category prediction is realized based on the trained support vector machine, in the training process of the support vector machine, input values of the support vector machine are cutter head torque, cutter head rotating speed, thrust force, penetration degree, tunneling speed, torque cut depth index and field cut depth index value, and rock uniaxial compressive strength, and output values of the support vector set are stratum categories; the stratum categories specifically include: super-soft stratum, relatively homogeneous soft rock, fully weathered soft and hard uneven stratum, strongly weathered and moderately weathered soft and hard uneven stratum, medium-soft stratum, soft and hard uneven partial soft stratum, soft and hard uneven partial hard stratum, medium-hard stratum, relatively homogeneous hard rock and super-hard stratum; when a loss function value of the support vector machine decreases to a minimum value or an iteration number reaches a set number of times, the training is stopped, and the trained support vector machine is obtained; The surrounding rock grade prediction is realized based on the trained multilayer perceptron, in the training process, input values of the multilayer perceptron are disc rotating speed, penetration degree, support shoe pressure, support shoe pump pressure, support shoe pitch angle, control pump pressure, advancing speed and support shoe rolling angle; output values of the multilayer perceptron are surrounding rock grades I, II, III, IV and V; when a loss function value of the multilayer perceptron decreases to a minimum value or an iteration number reaches a set number of times, the training is stopped, and the trained multilayer perceptron is obtained.

5. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second and third ground surface settlement prediction modules realize ground surface settlement prediction through the trained CNN-BiLSTM model; The CNN-BiLSTM model comprises: a CNN network and a BiLSTM network connected in sequence; In the training process, input values of the CNN-BiLSTM model are bearing capacity, natural unit weight, tunnel depth, distance between a monitoring point and a cutter head, penetration degree, thrust force, incision pressure, working cabin pressure, cutter head rotating speed and cutter head torque; output values of the CNN-BiLSTM model are ground surface settlement prediction results; when a loss function value of the CNN-BiLSTM model decreases to a minimum value or an iteration number reaches a set number of times, the training is stopped, and the trained CNN-BiLSTM model is obtained.

6. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second and third cutter wear prediction modules predict cutter wear conditions based on the trained BP neural network. In the training process, the BP neural network relies on the historical tunneling parameters of the shield, the stratum parameter information, and the cutter wear information to establish a database, selects the cutter torque, the cutter rotating speed, the thrust force, the tunneling speed, the penetration, the cutter diameter, the installation radius, the cutter spacing, the rock uniaxial compressive strength, the stratum category, and the surrounding rock grade as the input values; the output value of the BP neural network is the cutter wear amount; when the loss function value of the BP neural network decreases to the minimum value or the iteration number reaches the set number, the training is stopped, and the trained BP neural network is obtained.

7. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second, and third shield posture deviation prediction modules are based on the trained long short-term memory neural network model LSTM to realize the prediction of the shield posture deviation; in the training process of the model, the input values of the long short-term memory neural network model LSTM are the cutter torque, the cutter rotating speed, the tunneling speed, the thrust force, the difference between the up and down cylinder thrust forces, the cohesion, the internal friction angle, the lateral pressure coefficient, and the tunnel depth; The output values are the horizontal and vertical deviations of the cut, the horizontal and vertical deviations of the shield tail; when the loss function value of the model decreases to the minimum value or the iteration number reaches the set number, the training is stopped, and the trained long short-term memory neural network model LSTM is obtained.

8. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second, and third tunneling parameter prediction modules are based on the trained XGBoost model to realize the tunneling parameter prediction; In the training process, the input variables of the XGBoost model are the cutter extrusion force, the penetration, the cut pressure, the tunneling speed, the grouting pressure, the grouting flow, the grouting discharge flow, the stratum category, the surrounding rock grade, the torque cut depth index, and the field cut depth index value, and the output variables of the XGBoost model are the cutter torque, the cutter rotating speed, and the total thrust force; when the loss function value of the XGBoost model decreases to the minimum value or the iteration number reaches the set number, the training is stopped, and the trained XGBoost model is obtained.

9. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second, and third tunneling parameter optimization modules are based on the multi-objective optimization theory, select the cutter rotating speed, the cutter torque, the total thrust force, and the penetration as the optimization variables, and realize the multi-objective optimization of the maximum tunneling speed and the minimum tunneling specific energy by using the NSGA-II algorithm. The multi-objective optimization framework is as follows: ； ； ； ； ； ； ； wherein, is the excavation specific energy, n is the cutterhead rotational speed, T is the cutterhead torque, F is the total thrust, R is the shield machine cutterhead radius, is the excavation speed, P is the shield cutterhead penetration.

10. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second, and third shield construction risk early warning modules are used to give early warning results for various construction risks in the construction process. The construction risk early warning module includes: geological risk early warning, equipment risk early warning, and environmental risk early warning; the geological risk early warning, when the shield machine entity construction tunnels, the shield machine digital twin body synchronously tunnels in the digital space. Since the shield digital twin system has imported the three-dimensional geological information of the construction area during modeling, the geological information in front of the excavation face can be intuitively displayed in the digital twin system. Based on the prediction results of the stratum parameter prediction module and the visual display of the geological characteristics in front of the excavation face, early warning information is given before the geological conditions change; The device risk early warning is based on a tool wear prediction module, monitors a tool state, predicts faults and abnormalities, and takes measures in advance to avoid risks. Meanwhile, the shield digital twin system integrates real-time data of the shield machine, monitors running states of the main bearing and the hydraulic system based on a visualization module, and gives an early warning signal and an abnormal position when device parameters exceed rated values or are not working.

11. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second, and third automatic control modules rely on the Internet of Things to realize control of the shield digital twin system on the shield machine entity. In order to fully utilize the tunneling parameter optimization module function, the automatic control module converts control instructions according to the optimization strategy, the control instructions are issued to the controller of the shield machine entity through the Internet of Things, and accurate control and optimized operation of the shield machine entity are realized, and construction parameters are adjusted in real time.

12. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 2, characterized in that, The first, second, and third visualization modules are used to display the established shield digital twin model, and are closely integrated with all the above-mentioned function modules. The real-time collected construction parameters and the previously constructed shield digital twin model parameters are displayed in the form of a chart, and the prediction and optimization results of each function module are visually displayed.

13. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 1, characterized in that, The central server averages the parameters of the three models, and issues the average values of the parameters to each participant, including: the central server is responsible for receiving and decrypting the model update parameters and gradients from each participant, verifying the integrity and correctness of the data, and avoiding damage or malicious tampering of the data during transmission.

14. The shield tunnel risk early warning method based on digital twinning and federated learning according to claim 13, characterized in that, Each participant updates its own model according to the average value of the parameters, and realizes shield construction risk early warning based on the updated model, including: the participant receives and decrypts the latest model data and gradient information issued by the central server, and integrates the information into the local shield construction risk early warning model. After multiple debugging and testing, the accuracy of the local shield construction risk early warning model is verified.

15. A shield tunnel risk early warning system based on digital twinning and federated learning, characterized in that, The system includes: A shield tunnel construction party client, a shield machine production party client, a third party client, and a central server; The shield tunnel construction party client is an edge computing terminal that obtains geological parameters and construction parameters. The shield tunnel construction party client transmits the obtained data to the shield machine production party client and the third party client after homomorphic encryption processing. The shield machine production party client transmits shield device parameters to the shield tunnel construction party client and the third party client after homomorphic encryption processing. The geological parameters, construction parameters, and shield device parameters are collectively referred to as basic parameters; The shield machine production party client, the shield tunnel construction party client, and the third party client each construct a first, second, and third shield digital twin system according to the basic parameters; Each shield digital twin system is provided with a shield construction risk early warning model. The shield machine manufacturer client, the shield tunnel construction party client, the third party client and the central server jointly constitute a federated learning system; the shield machine manufacturer client, the shield tunnel construction party client and the third party client are all participants, and each participant sends the model parameters of the respective shield construction risk early warning model to the central server; The central server performs an average value operation on the parameters of the three models, and sends the average values of the parameters to each participant; each participant updates the model according to the average values of the parameters, and realizes shield construction risk early warning based on the updated model.

Images