Karst stratum shield tunneling parameter prediction method and system based on federated learning

Abstract

The invention provides a karst stratum shield tunneling parameter prediction method and system based on federated learning, and relates to the technical field of shield tunneling, and the method comprises the steps: collecting multi-modal data, and carrying out the preprocessing; constructing a karst cave three-dimensional distribution model and extracting microscopic feature indexes through the constructed deep learning classification model in combination with multi-modal data and federated learning technology; analyzing the extracted microscopic characteristic indexes through a deep learning classification model to obtain the karst cave type; inputting the quantized microscopic characteristic index and the karst cave type into a parameter mapping model, and establishing an internal relationship between the microscopic characteristic and the tunneling parameter; according to the internal relation, the karst cave type and the distribution complexity, a tunneling parameter adjusting instruction is obtained; and the tunneling parameters of the shield tunneling machine are dynamically and accurately adjusted in real time through the tunneling parameter adjusting instruction. And in combination with a multi-modal sensor, edge computing equipment and a federated learning framework, accurate prediction and real-time adjustment of shield tunneling parameters of the karst stratum are realized.

Description

Technical Field

[0001] The present application relates to the field of shield tunneling technology, and in particular to a method and system for predicting shield tunneling parameters in karst strata based on federated learning. Background Art

[0002] At present, shield machines face significant technical challenges in the construction process of karst formations. Karst formations are widely distributed in the tunnel construction area, and their caves are unevenly distributed and highly concealed, posing a major threat to the shield excavation process. The distribution of caves and the mud and water inrush problems they cause are particularly prominent, which not only affects construction efficiency, but also poses serious safety hazards.

[0003] One existing method is to use finite element numerical simulation technology to establish a simulation model of shield tunneling based on geological exploration data. For example, patent CN202211102890.8 proposes a technical solution to predict the range of tunneling parameters through orthogonal experimental design and discrete element coupling model. This method obtains the optimal range of tunneling speed, cutter head speed and screw machine speed by simulating the adjustment of tunneling parameters under different excavation conditions. However, this method mainly relies on offline analysis and static data, and cannot dynamically respond to sudden changes in geological conditions during tunneling, and has poor adaptability.

[0004] Another method is to combine slag monitoring and image recognition technology to detect the occurrence of caves in real time during the tunneling process of the shield machine. For example, patent CN202410288627.5 uses the YOLO model to train images of cave fillings, and monitors the torque changes of the cutterhead and the distribution of cave fillings through a slag detection camera to achieve real-time identification and area calculation of caves. This method has certain advantages in identifying caves, but it is mainly aimed at monitoring after tunneling, and cannot optimize the tunneling parameter settings in advance, and has insufficient response capabilities to dynamic changes during tunneling.

[0005] The above methods have their own focuses, but they still have the following shortcomings in predicting the excavation parameters of karst strata: First, the data collection is single. The existing geological radars and sensors are mainly used for the detection of macroscopic stratigraphic information, while the monitoring accuracy of microscopic characteristics such as cave distribution and filling composition is low, resulting in a lag in the adjustment of excavation parameters. Secondly, the prediction accuracy of traditional models is not high, and it is difficult to effectively cope with the highly heterogeneous geological environment in karst strata, especially the complex distribution of filled caves and hollow caves. During shield tunneling, sudden cave problems (such as water gushing and mud bursting) are prone to cause major safety hazards, and the existing technologies are mostly offline analysis, resulting in weak real-time response capabilities, unable to adjust parameters in real time during dynamic excavation, and lack of dynamic response capabilities for sudden geological conditions. Finally, the existing technologies usually focus on a single project, resulting in weak collaboration between multiple projects, and unable to improve the generalization performance of the model through cross-project data sharing. Summary of the invention

[0006] The purpose of the present invention is to provide a method and system for predicting shield tunneling parameters in karst strata based on federated learning in order to solve the problem that the existing method for predicting karst strata tunneling parameters cannot detect microscopic features in real time during dynamic tunneling and has poor prediction accuracy.

[0007] The above-mentioned purpose of the present application is achieved through the following technical solutions: S1: Collect multimodal data of karst strata and perform preprocessing; multimodal data includes: high-resolution point cloud data of caves, pressure change information, temperature information, humidity information and hydrological data; S2: Use convolutional neural network (CNN) and Transformer algorithm to build a deep learning classification model; S3: Through deep learning classification model, combined with multimodal data and federated learning technology, a three-dimensional distribution model of karst caves is constructed and microscopic characteristic indicators are extracted; S4: The extracted microscopic characteristic indicators are analyzed through the deep learning classification model to obtain the cave type; S5: Construct parameter mapping model; input the quantified microscopic characteristic indexes and cave types into the parameter mapping model, and establish the intrinsic relationship between microscopic characteristics and shield tunneling parameters through multivariate regression analysis; S6: Divide the karst cave area of ​​the 3D karst cave distribution model into different geological zones, and obtain excavation parameter adjustment instructions according to internal relationships, karst cave types and distribution complexity; S7: Adjust the excavation parameters of each geological zone through excavation parameter adjustment instructions, control the hydraulic pump, cutter head drive motor, water injection pump and valve of each project, and complete the real-time dynamic and precise adjustment of the shield machine's excavation parameters.

[0008] Optionally, step S1 includes: the preprocessing includes: data cleaning, feature extraction and standardization operations.

[0009] Optionally, step S3 includes: Mineral spectrum data obtained by spectral analysis instruments; Through mineral spectral data and multimodal data, combined with deep learning classification models, a three-dimensional distribution model of caves is constructed and microscopic characteristic indicators are extracted; Microscopic characteristic indicators include: wall roughness , crack density , Mineral Composition Index and moisture index .

[0010] Optionally, step S3 further includes: The surface roughness of the wall is calculated by fitting the high-resolution point cloud data of the cave. ; The deep learning model is used to identify and count the number of cracks per unit area to obtain the crack density. ; Through mineral spectrum data, combined with deep learning algorithms, the proportion of each mineral component is identified to obtain the mineral composition index ; The moisture content index is calculated through temperature information, humidity information and hydrological data. , the formula is as follows:

[0011] in, The water content measured by the water content sensor is the hydrological data; is the relative humidity, which is calculated from the temperature information and humidity information; and is the weight coefficient, satisfying .

[0012] Optionally, step S4 further includes: The extracted microscopic characteristic indicators are analyzed by a deep learning classification model to obtain the cave types; the cave types include: filled caves and cavity caves; The output of the deep learning classification model is the probability distribution of cave types. Defined as:

[0013] in, represents the input multimodal sensor data, It is a feature extraction and modeling function that combines CNN and Transformer.

[0014] Optionally, step S6 includes: The specific steps for calculating the distribution complexity of caves are as follows: S61: Extract the geometric features of the cave through the high-resolution point cloud data of the cave; the geometric features include: shape, size and volume; S62: Connectivity analysis based on high-resolution point cloud data of the cave, and calculation of connectivity , including: Set distance threshold; If the distance between two caves is less than the distance threshold, they are considered to be connected and have a connection relationship; With each cave as a vertex and connectivity as an edge, an undirected graph is constructed to calculate connectivity ; S63: Based on the geometric characteristics and connectivity of the caves , the distribution complexity is calculated as follows:

[0015] in, is the distribution complexity, is the cave volume, is the length of the cave, For cave connectivity.

[0016] Optionally, step S7 includes: the excavation parameter adjustment instruction includes: thrust increase or decrease, cutter head speed adjustment and water injection volume control; excavation parameters: including thrust , Cutter speed and water injection volume .

[0017] A karst stratum shield tunneling parameter prediction system based on federated learning, the system comprising: a multimodal sensor module, an edge computing device, a communication module, a shield machine control unit and a central server; The multimodal sensor module is connected to the edge computing device; the edge computing device is connected to the communication module; the communication module is connected to the central server; the shield machine control unit is connected to the edge computing device; The multimodal sensor module includes: geological radar, pressure sensor, temperature and humidity sensor and moisture content detector; The geological radar is arranged at the front end of the shield machine and is used to scan the distribution and geometric characteristics of the karst caves in the karst strata in real time, obtain high-resolution point cloud data of the caves, and generate the spatial position and structural information of the caves; The pressure sensor is arranged on the cutter head and screw conveyor of the shield machine to monitor the dynamic change of surrounding rock pressure during excavation and obtain pressure change information; The temperature and humidity sensors are arranged around the cutterhead of the shield machine and are used to monitor the temperature and humidity information in the cave environment; The water content detector is arranged in the tunneling chamber of the shield machine and is used to monitor the hydrological data in the karst cave environment; The edge computing device includes: a high-performance computing unit GPU, a data storage unit and a communication interface module; The multimodal sensor module is used to collect multimodal data; the multimodal data includes: the spatial location of the cave, the structural information of the cave, the pressure change information, the temperature information, the humidity information and the hydrological data; The edge computing device is used to preprocess and edge compute the received multimodal data to obtain local model parameters; the preprocessing includes: data cleaning, feature extraction and standardization operations; The communication module includes: a 5G module and a cache memory; The communication module is used to transmit the local model parameters to the central server; The central server is used to build a deep learning classification model based on convolutional neural network CNN and Transformer algorithm; The central server is used to obtain globally optimized model parameters based on local model parameters, deep learning classification model and federated learning technology; The shield machine control unit includes: PLC controller and actuator drive module; The edge computing device is also used to receive and process the globally optimized model parameters, obtain the tunneling parameter adjustment instructions and send them to the shield machine control unit; The shield machine control unit includes: hydraulic pump, cutter head drive motor, water injection pump and valve; The shield machine control unit is used to control the hydraulic pump, cutterhead drive motor, water injection pump and valve through tunneling parameter adjustment instructions to complete real-time dynamic and precise adjustment of the shield machine's tunneling parameters.

[0018] A computer-readable storage medium stores instructions, and when the instructions are executed, a method for predicting shield tunneling parameters in karst strata based on federated learning is executed.

[0019] The beneficial effects of the technical solution provided by this application are: 1. Application of multimodal data fusion technology: The system integrates multiple sensors, such as geological radar, pressure sensor, temperature and humidity sensor and moisture content detector, to collect environmental data of karst strata in real time. After preprocessing, cleaning and standardization by edge computing equipment, the data is fused by convolutional neural network CNN and Transformer algorithm to realize accurate modeling and key feature extraction of cave distribution, and generate a real-time updated three-dimensional distribution model of cave.

[0020] 2. Application of the federated learning framework: The system relies on the federated learning framework, adopts homomorphic encryption technology and differential privacy protection mechanism, uploads the local model update parameters of different projects to the central server, realizes data collaborative optimization through global model training, and ensures that data privacy between projects is not leaked.

[0021] 3. Dynamic excavation parameter prediction and real-time adjustment: Based on the 3D model of the cave and real-time sensor data and according to the classification model design logic of filled caves and hollow caves, the system dynamically predicts key excavation parameters such as thrust, cutterhead speed and water injection volume, and combines the real-time feedback mechanism to achieve automatic optimization of the parameters.

[0022] 4. Collaboration between edge computing and communication modules: By deploying high-performance computing units at the edge of the shield machine, the system realizes real-time processing of sensor data, model reasoning, and instruction generation, avoiding the delay and uncertainty caused by the transmission of large amounts of data to the central server. The communication module plays a role in transfer and storage in the process of uploading data and issuing models.

[0023] 5. Closed-loop feedback mechanism for the whole system: The system realizes the complete workflow of data collection, model optimization, parameter adjustment and feedback through the closed-loop feedback mechanism. The sensor module continuously collects real-time data during the construction process and transmits feedback information back to the edge computing device for further optimization of the excavation parameter prediction model. This closed-loop design ensures the prediction accuracy and response speed of the system, enabling it to better cope with the complex and changeable geological conditions of karst strata. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present application will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 is a system structure diagram in an embodiment of the present application; Figure 2 It is a flow chart of three-dimensional modeling of a cave in an embodiment of the present application; Figure 3 is a dynamic excavation parameter prediction flow chart in an embodiment of the present application; Figure 4 This is a diagram of the blade sensor setting in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to have a clearer understanding of the technical features, purposes and effects of the present application, the specific implementation methods of the present application are now described in detail with reference to the accompanying drawings.

[0026] An embodiment of the present application provides a method for predicting shield tunneling parameters in karst strata based on federated learning.

[0027] Please refer to Figure 1 , Figure 1 : is a system structure diagram of a method for predicting shield tunneling parameters in karst strata based on federated learning in an embodiment of the present application, including: S1: Collect multimodal data of karst strata and perform preprocessing; multimodal data includes: high-resolution point cloud data of caves, pressure change information, temperature information, humidity information and hydrological data; S2: Use convolutional neural network (CNN) and Transformer algorithm to build a deep learning classification model; S3: Through deep learning classification model, combined with multimodal data and federated learning technology, a three-dimensional distribution model of karst caves is constructed and microscopic characteristic indicators are extracted; S4: The extracted microscopic characteristic indicators are analyzed through the deep learning classification model to obtain the cave type; S5: Construct parameter mapping model; input the quantified microscopic characteristic indexes and cave types into the parameter mapping model, and establish the intrinsic relationship between microscopic characteristics and shield tunneling parameters through multivariate regression analysis; The intrinsic relationship refers to the correlation between the microscopic geometric characteristics of the cave (such as diameter, depth, volume, shape parameters, etc.) and the shield tunneling parameters (cutterhead speed, thrust, water injection volume). This relationship is constructed through the following model:

[0028] Among them, P is the preliminary excavation parameter adjustment suggestion, including thrust F, cutter head speed and water injection volume Q. D is the depth of the cave, R is the diameter of the cave, V is the volume of the cave, and shape characteristics include aspect ratio, irregularity coefficient, etc.

[0029] Through the above model, the system takes microscopic characteristics as input and outputs preliminary suggestions for adjusting the excavation parameters. This preliminary suggestion will be further integrated with the cave type and distribution complexity to ultimately generate a complete excavation parameter adjustment result.

[0030] S6: Divide the karst cave area of ​​the 3D karst cave distribution model into different geological zones, and obtain excavation parameter adjustment instructions according to internal relationships, karst cave types and distribution complexity; S7: Adjust the excavation parameters of each geological zone through excavation parameter adjustment instructions, control the hydraulic pump, cutter head drive motor, water injection pump and valve of each project, and complete the real-time dynamic and precise adjustment of the shield machine's excavation parameters.

[0031] As an embodiment, in view of the complex distribution of caves, the system adopts a partition processing and adaptive adjustment strategy.

[0032] Step S1 includes: the preprocessing includes: data cleaning, feature extraction and standardization operations.

[0033] Step S3 includes: Mineral spectrum data obtained by spectral analysis instruments; Through mineral spectral data and multimodal data, combined with deep learning classification models, a three-dimensional distribution model of caves is constructed and microscopic characteristic indicators are extracted; In one embodiment of the present application, Figure 2As shown in the figure, based on the multimodal data collected by geological radar and sensors, the convolutional neural network (CNN) is combined to extract the cave features, and the global dependencies in the stratigraphic data are captured through the Transformer algorithm to generate a real-time updated three-dimensional distribution model of the cave. This model not only includes the spatial position and geometric morphology of the cave, but also refines the microscopic features such as wall roughness, crack density, mineral composition ratio and water content, providing comprehensive stratigraphic information to support the precise adjustment of shield tunneling parameters.

[0034] In one embodiment of the present application, Figure 3 As shown in the figure, with the three-dimensional distribution model of the cave and real-time multimodal data as input, multimodal data fusion technology is used to predict key parameters such as shield machine thrust, cutter head speed and water injection volume in real time, and dynamically adjust the excavation strategy according to abnormal data feedback.

[0035] Microscopic characteristic indicators include: wall roughness , crack density , Mineral Composition Index and moisture index .

[0036] Step S3 also includes: The surface roughness of the wall is calculated by fitting the high-resolution point cloud data of the cave. ; The deep learning model is used to identify and count the number of cracks per unit area to obtain the crack density. ; Through mineral spectrum data, combined with deep learning algorithms, the proportion of each mineral component is identified to obtain the mineral composition index ; As an embodiment, deep learning algorithms include: convolutional neural network (CNN), Transformer, random forest, support vector machine (SVM), etc.

[0037] As an example, the real-time acquisition of mineral composition index relies on the combination of a spectrometer and a deep learning model: a mineral spectrometer is used to collect mineral composition spectral data in real time, and then a pre-trained CNN model is used to classify and proportion the spectral data. The analysis of spectral data is usually completed within a few seconds, so it can be obtained in real time.

[0038] The moisture content index is calculated through temperature information, humidity information and hydrological data. , the formula is as follows:

[0039] in, The water content measured by the water content sensor is the hydrological data; is the relative humidity, which is calculated from the temperature information and humidity information; and is the weight coefficient, satisfying ; Step S4 also includes: The extracted microscopic characteristic indicators are analyzed by a deep learning classification model to obtain the cave types; the cave types include: filled caves and cavity caves; The output of the deep learning classification model is the probability distribution of cave types. Defined as:

[0040] in, represents the input multimodal sensor data, It is a feature extraction and modeling function that combines CNN and Transformer.

[0041] As an example, according to the output of the deep learning classification model, caves are divided into two types: filled caves and empty caves. Filled caves usually have higher density and different mineral composition distribution, while empty caves are hollow structures and may contain water or other fluids.

[0042] Step S6 includes: The specific steps for calculating the distribution complexity of caves are as follows: S61: Extract the geometric features of the cave through the high-resolution point cloud data of the cave; the geometric features include: shape, size and volume; S62: Connectivity analysis based on high-resolution point cloud data of the cave, and calculation of connectivity , including: Set distance threshold; If the distance between two caves is less than the distance threshold, they are considered to be connected and have a connection relationship; With each cave as a vertex and connectivity as an edge, an undirected graph is constructed to calculate connectivity , the specific calculation formula is as follows:

[0043] S63: Based on the geometric characteristics and connectivity of the caves , the distribution complexity is calculated as follows:

[0044] in, is the distribution complexity, is the cave volume, is the length of the cave, For cave connectivity.

[0045] In one embodiment of the present application, the cave area is divided into three categories: simple, complex and highly complex according to the level of distribution complexity to guide the optimization of excavation parameters and finally obtain excavation parameter adjustment instructions.

[0046] Step S7 includes: The excavation parameter adjustment instructions include: thrust increase and decrease, cutter head speed adjustment and water injection control; excavation parameters: including thrust , Cutter speed and water injection volume .

[0047] As an example, filling a cave area: Since the filling material may cause local instability, the system will increase the water injection volume To improve soil stability or adjust thrust To adapt to the density of different materials. At the same time, the cutter head speed It also needs to be properly adjusted according to the properties of the filler material to optimize cutting efficiency and reduce mechanical wear.

[0048] The specific adjustment formula is as follows:

[0049]

[0050]

[0051] in, It represents the filling of caves; It is the adjustment coefficient of the cave type, and different adjustment strategies are set according to different cave types; Indicates the thrust, cutter head speed and increased water injection volume before adjustment; Indicates the adjusted thrust, cutter head speed and increased water injection; Indicates adjustment factor; wall roughness , crack density , Mineral Composition Index and moisture index ; As an example, in the cavity cave area: special attention should be paid to the strengthening of the support structure, and the system will appropriately reduce the thrust To avoid excessive pressure on the cave wall, strengthen monitoring of water pressure changes and adjust water injection volume in time. At the same time, the blade speed It also needs to be properly adjusted according to the properties of the filler material to optimize cutting efficiency and reduce mechanical wear.

[0052] The specific adjustment formula is as follows:

[0053]

[0054]

[0055] in, Represents a hollow cave.

[0056] A karst stratum shield tunneling parameter prediction system based on federated learning, the system comprising: a multimodal sensor module, an edge computing device, a communication module, a shield machine control unit and a central server; The multimodal sensor module is connected to the edge computing device; the edge computing device is connected to the communication module; the communication module is connected to the central server; the shield machine control unit is connected to the edge computing device; The multimodal sensor module includes: geological radar, pressure sensor, temperature and humidity sensor and moisture content detector; The geological radar is arranged at the front end of the shield machine and is used to scan the distribution and geometric characteristics of the karst caves in the karst strata in real time, obtain high-resolution point cloud data of the caves, and generate the spatial position and structural information of the caves; The pressure sensor is arranged on the cutter head and screw conveyor of the shield machine to monitor the dynamic change of surrounding rock pressure during excavation and obtain pressure change information; The temperature and humidity sensors are arranged around the cutterhead of the shield machine and are used to monitor the temperature and humidity information in the cave environment; The water content detector is arranged in the tunneling chamber of the shield machine and is used to monitor the hydrological data in the karst cave environment; The edge computing device includes: a high-performance computing unit GPU, a data storage unit and a communication interface module; The multimodal sensor module is used to collect multimodal data; the multimodal data includes: the spatial location of the cave, the structural information of the cave, the pressure change information, the temperature information, the humidity information and the hydrological data; The edge computing device is used to preprocess and edge compute the received multimodal data to obtain local model parameters; the preprocessing includes: data cleaning, feature extraction and standardization operations; As an example, all sensors aggregate data to the edge computing device through an interface to complete real-time data collection. The edge computing device is the core hardware module of the system, with a built-in high-performance computing unit (such as GPU), data storage unit and communication interface module, which can quickly complete sensor data processing and model reasoning. The device receives multimodal data from the sensor module through a standardized interface, performs data cleaning, feature extraction and standardization operations, and transmits the processing results to the central server through the communication module.

[0057] The communication module includes: a 5G module and a cache memory; The communication module is used to transmit the local model parameters to the central server; As an embodiment, the communication module consists of a wireless communication device (such as a 5G module) and a cache memory, and plays a bridge role in the system. On the one hand, the communication module receives the local model parameters uploaded by the edge computing device and transmits them to the central server; on the other hand, the central server sends the globally optimized model parameters to the edge computing device through the communication module.

[0058] The central server is used to build a deep learning classification model based on convolutional neural network CNN and Transformer algorithm; The central server is used to obtain globally optimized model parameters based on local model parameters, deep learning classification model and federated learning technology; The shield machine control unit includes: PLC controller and actuator drive module; The edge computing device is also used to receive and process the globally optimized model parameters, obtain the tunneling parameter adjustment instructions and send them to the shield machine control unit; The shield machine control unit includes: hydraulic pump, cutter head drive motor, water injection pump and valve; The shield machine control unit is used to control the hydraulic pump, cutterhead drive motor, water injection pump and valve through tunneling parameter adjustment instructions to complete real-time dynamic and precise adjustment of the shield machine's tunneling parameters.

[0059] As an example, the shield machine control unit is an important component of the adjustment strategy generated by the execution system, which is mainly composed of a PLC controller and an actuator drive module. The shield machine control unit receives the excavation parameter adjustment instructions sent by the edge computing device, including thrust increase and decrease, cutter head speed adjustment and water injection control, and drives the hydraulic system, cutter head drive motor, water injection pump and valve through the internal program logic to complete the parameter adjustment.

[0060] As an embodiment, the hydraulic system achieves precise control of the propulsion force by adjusting the output flow of the hydraulic pump; the cutter disc drive motor achieves dynamic control of the cutter disc speed by adjusting the motor speed through the inverter; the water injection pump and valve achieve dynamic adjustment of the water injection volume by adjusting the pump power or the valve opening.

[0061] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure.

[0062] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A method for predicting shield tunneling parameters in karst strata based on federated learning, characterized in that: The method comprises the following steps: S1: Collect multimodal data of karst strata and perform preprocessing; multimodal data includes: high-resolution point cloud data of caves, pressure change information, temperature information, humidity information and hydrological data; S2: Use convolutional neural network (CNN) and Transformer algorithm to build a deep learning classification model; S3: Through deep learning classification model, combined with multimodal data and federated learning technology, a three-dimensional distribution model of karst caves is constructed and microscopic characteristic indicators are extracted; S4: The extracted microscopic characteristic indicators are analyzed through the deep learning classification model to obtain the cave type; S5: Construct parameter mapping model; input the quantified microscopic characteristic indexes and cave types into the parameter mapping model, and establish the intrinsic relationship between microscopic characteristics and shield tunneling parameters through multivariate regression analysis; S6: Calculate the distribution complexity of the karst caves; divide the karst cave area of ​​the 3D karst cave distribution model into different geological zones, and obtain excavation parameter adjustment instructions according to the internal relationship, karst cave type and distribution complexity; S7: Adjust the excavation parameters of each geological zone through excavation parameter adjustment instructions, control the hydraulic pump, cutter head drive motor, water injection pump and valve of each project, and complete the real-time dynamic and precise adjustment of the shield machine's excavation parameters.

2. A method for predicting shield tunneling parameters in karst strata based on federated learning as claimed in claim 1, characterized in that: Step S1 includes: the preprocessing includes: data cleaning, feature extraction and standardization operations.

3. The method for predicting shield tunneling parameters in karst strata based on federated learning according to claim 1, characterized in that: Step S3 includes: Mineral spectrum data obtained by spectral analysis instruments; Through mineral spectral data and multimodal data, combined with deep learning classification models, a three-dimensional distribution model of caves is constructed and microscopic characteristic indicators are extracted; Microscopic characteristic indicators include: wall roughness , crack density , Mineral Composition Index and moisture index .

4. A method for predicting shield tunneling parameters in karst strata based on federated learning as claimed in claim 3, characterized in that: Step S3 also includes: The surface roughness of the wall is calculated by fitting the high-resolution point cloud data of the cave. ; The deep learning model is used to identify and count the number of cracks per unit area to obtain the crack density. ; Through mineral spectrum data, combined with deep learning algorithms, the proportion of each mineral component is identified to obtain the mineral composition index ; The moisture content index is calculated through temperature information, humidity information and hydrological data. , the formula is as follows: in, The water content measured by the water content sensor is the hydrological data; is the relative humidity, which is calculated from the temperature information and humidity information; and is the weight coefficient, satisfying .

5. The method for predicting shield tunneling parameters in karst strata based on federated learning according to claim 1, characterized in that: Step S4 also includes: The extracted microscopic characteristic indicators are analyzed by a deep learning classification model to obtain the cave types; the cave types include: filled caves and cavity caves; The output of the deep learning classification model is the probability distribution of cave types. Defined as: in, represents the input multimodal sensor data, It is a feature extraction and modeling function that combines CNN and Transformer.

6. The method for predicting shield tunneling parameters in karst strata based on federated learning according to claim 1, characterized in that: Step S6 includes: The specific steps for calculating the distribution complexity of caves are as follows: S61: Extract the geometric features of the cave through the high-resolution point cloud data of the cave; the geometric features include: shape, size and volume; S62: Connectivity analysis based on high-resolution point cloud data of the cave, and calculation of connectivity , specifically including: Set distance threshold; If the distance between two caves is less than the distance threshold, they are considered to be connected and have a connection relationship; With each cave as a vertex and connectivity as an edge, an undirected graph is constructed to calculate connectivity ; S63: Based on the geometric characteristics and connectivity of the caves , the distribution complexity is calculated as follows: in, is the distribution complexity, is the cave volume, is the length of the cave, For cave connectivity.

7. The method for predicting shield tunneling parameters in karst strata based on federated learning according to claim 1, characterized in that: Step S7 includes: The excavation parameter adjustment instructions include: thrust increase and decrease, cutter head speed adjustment and water injection control; excavation parameters: including thrust , Cutter speed and water injection volume .

8. A karst stratum shield tunneling parameter prediction system based on federated learning, used to implement a karst stratum shield tunneling parameter prediction method based on federated learning as described in any one of claims 1 to 7, characterized in that: The system includes: a multimodal sensor module, an edge computing device, a communication module, a shield machine control unit and a central server; The multimodal sensor module is connected to the edge computing device; the edge computing device is connected to the communication module; the communication module is connected to the central server; the shield machine control unit is connected to the edge computing device; The multimodal sensor module includes: geological radar, pressure sensor, temperature and humidity sensor and moisture content detector; The geological radar is arranged at the front end of the shield machine and is used to scan the distribution and geometric characteristics of the karst caves in the karst strata in real time, obtain high-resolution point cloud data of the caves, and generate the spatial position and structural information of the caves; The pressure sensor is arranged on the cutter head and screw conveyor of the shield machine to monitor the dynamic change of surrounding rock pressure during excavation and obtain pressure change information; The temperature and humidity sensors are arranged around the cutterhead of the shield machine and are used to monitor the temperature and humidity information in the cave environment; The water content detector is arranged in the tunneling chamber of the shield machine and is used to monitor the hydrological data in the karst cave environment; The edge computing device includes: a high-performance computing unit GPU, a data storage unit and a communication interface module; The multimodal sensor module is used to collect multimodal data; the multimodal data includes: the spatial location of the cave, the structural information of the cave, the pressure change information, the temperature information, the humidity information and the hydrological data; The edge computing device is used to preprocess and edge compute the received multimodal data to obtain local model parameters; the preprocessing includes: data cleaning, feature extraction and standardization operations; The communication module includes: a 5G module and a cache memory; The communication module is used to transmit the local model parameters to the central server; The central server is used to build a deep learning classification model based on convolutional neural network CNN and Transformer algorithm; The central server is used to obtain globally optimized model parameters based on local model parameters, deep learning classification model and federated learning technology; The shield machine control unit includes: PLC controller and actuator drive module; The edge computing device is also used to receive and process the globally optimized model parameters, obtain the tunneling parameter adjustment instructions and send them to the shield machine control unit; The shield machine control unit includes: hydraulic pump, cutter head drive motor, water injection pump and valve; The shield machine control unit is used to control the hydraulic pump, cutterhead drive motor, water injection pump and valve through tunneling parameter adjustment instructions to complete real-time dynamic and precise adjustment of the shield machine's tunneling parameters.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 7 is executed.

Images