A real-time monitoring and early warning method for deformation of surrounding rock of a subway shield tunnel

Abstract

The present application relates to the technical field of subway shield tunnel engineering safety monitoring, in particular to a kind of real-time monitoring and early warning method of surrounding rock deformation of subway shield tunnel, comprising the following steps: S1 is arranged in shield tunnel by the multi-source fusion monitoring array of distributed optical fiber sensor, point high-precision sensor, machine vision unit is formed, and shield tunneling parameter acquisition interface is accessed simultaneously;S2, surrounding rock deformation data, environmental data, shield construction data are synchronously collected;The present application is distributed optical fiber+point sensing+machine vision fusion layout, realizes surrounding rock continuous, global, multi-dimensional deformation perception, completely eliminates monitoring blind area, and the monitoring accuracy reaches ±0.1mm~±0.5mm, uses edge local processing, and data response delay is ≤100ms, far more than traditional cloud processing mode, ensure that early warning real-time, early warning threshold is adjusted with stratum, driving, environment dynamically, false alarm rate is reduced by 70%, false negative rate is reduced by 90%, adapt to all complex stratum.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for subway shield tunnel engineering, specifically a method for real-time monitoring and early warning of surrounding rock deformation in subway shield tunnels. Background Technology

[0002] Subway shield tunnels often traverse urban core areas, with surrounding rock primarily consisting of soft soil, silty soil, sand and gravel, and water-rich fractured strata. These rock formations are characterized by poor self-stability, rapid deformation rates, and sensitivity to tunneling disturbances. Uncontrolled deformation of the surrounding rock can easily trigger major safety accidents such as surface subsidence, building cracking, pipeline damage, and tunnel segment instability. Existing surrounding rock deformation monitoring and early warning technologies have fundamental technical deficiencies and cannot meet the high safety, high precision, and high real-time requirements of subway shield tunneling construction.

[0003] The monitoring methods are limited and the spatiotemporal coverage is insufficient.

[0004] Traditional monitoring relies mainly on manual leveling instruments and convergence meters for measurement, while automated monitoring only uses single-point sensors. This results in problems such as monitoring lag, low sampling rate, and limited coverage, making it impossible to achieve continuous deformation perception of the entire surrounding rock area and prone to monitoring blind spots and missed alarms.

[0005] Data from multiple sources is isolated and has low fusion and utilization rates.

[0006] The deformation of the surrounding rock is affected by multiple factors such as geological conditions, shield thrust, soil pressure, grouting parameters, and groundwater. Existing technologies have not achieved deep integration of sensor data, geological data, tunneling parameters, and environmental data, resulting in one-sided deformation analysis and poor early warning accuracy.

[0007] The warning threshold is fixed and its adaptability is extremely poor.

[0008] Traditional early warning systems use fixed thresholds and do not take into account differences in strata, dynamic changes in construction, seasonality, and the influence of groundwater. Soft strata are prone to false alarms, while hard strata are prone to missed alarms, making them unsuitable for the complex and ever-changing dynamic deformation patterns of strata.

[0009] Lack of predictive ability, passive response

[0010] It can only monitor the current deformation value and cannot make advance predictions on future deformation trends and sudden risks. The early warning response is delayed, and construction personnel do not have enough time to take disposal measures.

[0011] Separated from shield tunneling, without closed-loop linkage

[0012] The monitoring system and the shield control system are independent of each other. When the surrounding rock deformation is abnormal, the tunneling parameters cannot be adjusted in real time, which can easily lead to increased disturbance and continuous expansion of deformation.

[0013] Data processing lag and low level of intelligence

[0014] Data is uploaded to the cloud for processing, but the latency is high. In harsh construction environments, network instability can easily lead to data loss. It lacks edge computing real-time analysis capabilities and cannot meet the needs for millisecond-level real-time early warning.

[0015] While existing patents and standards propose some tunnel monitoring solutions, none have formed a complete technical system integrating multi-source fusion perception, edge real-time computing, AI trend prediction, dynamic hierarchical early warning, and tunneling closed-loop linkage. This system cannot truly solve the core pain point of monitoring surrounding rock deformation in subway shield tunnels. Therefore, a real-time monitoring and early warning method for surrounding rock deformation in subway shield tunnels is proposed. Summary of the Invention

[0016] In view of this, the present invention provides a method for real-time monitoring and early warning of surrounding rock deformation in subway shield tunnels, so as to solve or alleviate the technical problems existing in the prior art, and at least provide a beneficial option.

[0017] The technical solution of this invention is implemented as follows: A method for real-time monitoring and early warning of surrounding rock deformation in subway shield tunnels, comprising the following steps:

[0018] S1 is equipped with a multi-source fusion monitoring array consisting of distributed fiber optic sensors, point-type high-precision sensors, and machine vision units inside the shield tunnel, and is simultaneously connected to the shield tunneling parameter acquisition interface.

[0019] S2 synchronously collects surrounding rock deformation data, environmental data, and shield tunneling construction data, and achieves low-latency real-time transmission through an edge computing gateway;

[0020] S3 performs cleaning, spatiotemporal alignment, and noise reduction on the collected multi-source heterogeneous data, and extracts feature parameters such as cumulative deformation, deformation rate, deformation acceleration, and strain gradient.

[0021] S4 is based on a mechanical inversion model and a multi-source data fusion model to perform real-time inversion calculations of the surrounding rock arch subsidence, horizontal convergence, vertical settlement, and global deformation field.

[0022] S5 establishes a dynamic early warning threshold by combining stratum type, burial depth, tunneling parameters and water environment conditions, and uses ST-GCN spatiotemporal graph convolutional network to predict the deformation trend of surrounding rock in the next 30 min to 24 h.

[0023] S6 triggers four levels of early warning based on cumulative deformation value, deformation rate, and predicted risk value: blue alert, yellow warning, orange alert, and red emergency alert, and displays them in 3D visualization on the BIM platform.

[0024] Based on the warning level, S7 automatically sends excavation parameter control commands to the shield control system to achieve closed-loop linkage of monitoring, early warning and control.

[0025] S8 stores all monitoring data, early warning records, and control logs into a spatiotemporal database, and performs incremental learning and iterative updates on the AI ​​prediction model and dynamic thresholds.

[0026] More preferably, the deployment rule of the multi-source fusion monitoring array in step S1 is as follows:

[0027] A monitoring section is set up every 20m along the tunnel axis;

[0028] Each cross section is equipped with distributed optical fibers, static level, tilt sensor and convergence sensor at the arch top, left and right arch waists and arch bottom.

[0029] One machine vision unit is deployed every 100m to capture the displacement of the surrounding rock surface and the expansion of cracks;

[0030] Simultaneously collect data on the total thrust of the tunnel boring machine, cutterhead torque, soil chamber pressure, propulsion speed, synchronous grouting volume, and groundwater pressure.

[0031] More preferably, in step S2, the data acquisition frequency is adaptively adjustable from 1Hz to 10Hz, and dual-redundant transmission of industrial Ethernet and 5G is adopted, with a transmission delay of ≤50ms. The edge computing gateway realizes local real-time analysis, and supports local caching and automatic resume transmission after network disconnection.

[0032] More preferably, in step S4, the surrounding rock deformation inversion adopts a combination of distributed fiber strain inversion and point sensor calibration, and the final deformation monitoring accuracy is controlled within ±0.1mm~±0.5mm.

[0033] More preferably, the formula for calculating the dynamic early warning threshold in step S5 is:

[0034] Dynamic threshold = baseline threshold × formation correction factor × tunneling disturbance factor × water environment correction factor;

[0035] The baseline threshold is determined based on the subway engineering monitoring and measurement specifications, and the correction coefficient is adaptively and iteratively optimized according to the on-site geological and construction conditions.

[0036] More preferably, in step S5, the input data of the ST-GCN spatiotemporal graph convolutional network includes historical deformation sequences, stratum type, burial depth, shield tunneling parameters, groundwater, and surface load, and outputs the deformation prediction values ​​and instability risk probabilities for the next 30 minutes, 2 hours, 6 hours, 12 hours, and 24 hours.

[0037] More preferably, the four-level early warning determination criteria in step S6 are:

[0038] Blue indicator: Deformation rate ≤ 50% of the specification limit, trend is stable;

[0039] Yellow alert: Deformation rate is 50%–100% of the standard limit, with an increasing trend;

[0040] Orange alert: Deformation rate exceeds the standard limit by 100% to 150%, indicating a risk of instability;

[0041] Red emergency alert: Deformation rate exceeds the standard limit by more than 150%, which immediately endangers construction safety;

[0042] Early warning information is simultaneously pushed through on-site audio and visual displays, platform pop-ups, SMS, and APP.

[0043] A further preferred closed-loop linkage control strategy in step S7 is:

[0044] Yellow alert: Reduce the advance speed, stabilize the soil pressure, and fine-tune the grouting volume;

[0045] Orange alert: Suspend tunneling, increase grouting pressure and volume, and strengthen support;

[0046] Red emergency alert: Immediately shut down the machine, close the working face, and organize the evacuation of personnel;

[0047] After adjustment, the deformation recovery is continuously monitored to form a closed-loop feedback.

[0048] Further preferred, in step S8 the system has self-learning capability, continuously optimizing the AI ​​prediction model and dynamic threshold coefficient through actual monitoring data, and is suitable for soft soil, gravel, fractured rock, water-rich, and complex strata with soft upper and hard lower layers.

[0049] Furthermore, this method is applicable to the monitoring of surrounding rock deformation throughout the entire life cycle of subway shield tunnels during construction and operation, and is compatible with earth pressure balance shield tunnels and slurry balance shield tunnels. The monitoring data supports BIM+GIS three-dimensional visualization.

[0050] The embodiments of the present invention have the following advantages due to the adoption of the above technical solutions:

[0051] I. This invention utilizes a distributed optical fiber + point sensor + machine vision fusion deployment to achieve continuous, full-area, and multi-dimensional deformation perception of surrounding rock, completely eliminating monitoring blind spots. The monitoring accuracy reaches ±0.1mm~±0.5mm. It adopts local processing at the edge, with a data response latency of ≤100ms, far exceeding the traditional cloud processing mode, ensuring real-time early warning. The early warning threshold is dynamically adjusted according to the strata, tunneling, and environment, reducing the false alarm rate by 70% and the missed alarm rate by 90%, and is adaptable to all complex strata.

[0052] Second, the ST-GCN model of this invention can accurately predict the deformation trend in the next 24 hours, provide early warning of sudden change risks, and provide sufficient time for construction and disposal. The monitoring system is directly linked with the tunnel boring machine, and automatically adjusts parameters when abnormalities occur to suppress the expansion of surrounding rock deformation from the source. The four-level early warning system (blue, yellow, orange, and red) is matched with standardized disposal plans, which facilitates rapid response by construction personnel and reduces the difficulty of safety management.

[0053] Third, the system of this invention continuously learns engineering data, constantly optimizes the model and thresholds, adapts to different geological and engineering conditions, has strong versatility, and is applicable to all common subway strata such as soft soil, gravel, fractured rock, water-rich soil, and soft upper and hard lower strata, covering the entire life cycle of construction and operation.

[0054] The above overview is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments, and features described above, further aspects, embodiments, and features of the invention will become readily apparent from the accompanying drawings and the following detailed description. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a structural diagram of the present invention. Detailed Implementation

[0057] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0058] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0059] like Figure 1 As shown in the figure, this invention provides a method for real-time monitoring and early warning of surrounding rock deformation in subway shield tunnels, including the following steps:

[0060] S1 is equipped with a multi-source fusion monitoring array consisting of distributed fiber optic sensors, point-type high-precision sensors, and machine vision units inside the shield tunnel, and is simultaneously connected to the shield tunneling parameter acquisition interface.

[0061] S2 synchronously collects surrounding rock deformation data, environmental data, and shield tunneling construction data, and achieves low-latency real-time transmission through an edge computing gateway;

[0062] S3 performs cleaning, spatiotemporal alignment, and noise reduction on the collected multi-source heterogeneous data, and extracts feature parameters such as cumulative deformation, deformation rate, deformation acceleration, and strain gradient.

[0063] S4 is based on a mechanical inversion model and a multi-source data fusion model to perform real-time inversion calculations of the surrounding rock arch subsidence, horizontal convergence, vertical settlement, and global deformation field.

[0064] S5 establishes a dynamic early warning threshold by combining stratum type, burial depth, tunneling parameters and water environment conditions, and uses ST-GCN spatiotemporal graph convolutional network to predict the deformation trend of surrounding rock in the next 30 min to 24 h.

[0065] S6 triggers four levels of early warning based on cumulative deformation value, deformation rate, and predicted risk value: blue alert, yellow warning, orange alert, and red emergency alert, and displays them in 3D visualization on the BIM platform.

[0066] Based on the warning level, S7 automatically sends excavation parameter control commands to the shield control system to achieve closed-loop linkage of monitoring, early warning and control.

[0067] S8 stores all monitoring data, early warning records, and control logs into a spatiotemporal database, and performs incremental learning and iterative updates on the AI ​​prediction model and dynamic thresholds.

[0068] In one embodiment, the deployment rule for the multi-source fusion monitoring array in step S1 is as follows:

[0069] A monitoring section is set up every 20m along the tunnel axis;

[0070] Each cross section is equipped with distributed optical fibers, static level, tilt sensor and convergence sensor at the arch top, left and right arch waists and arch bottom.

[0071] One machine vision unit is deployed every 100m to capture the displacement of the surrounding rock surface and the expansion of cracks;

[0072] Simultaneously collect data on the total thrust of the tunnel boring machine, cutterhead torque, soil chamber pressure, propulsion speed, synchronous grouting volume, and groundwater pressure.

[0073] In one embodiment, the data acquisition frequency in step S2 is adaptively adjustable from 1Hz to 10Hz, and industrial Ethernet and 5G dual-redundant transmission are adopted. The transmission delay is ≤50ms, and the edge computing gateway realizes local real-time analysis. In the event of network outage, it supports local caching and automatic resume transmission after network connection.

[0074] In one embodiment, the surrounding rock deformation inversion in step S4 adopts a combination of distributed fiber strain inversion and point sensor calibration, and the final deformation monitoring accuracy is controlled within ±0.1mm~±0.5mm.

[0075] In one embodiment, the formula for calculating the dynamic early warning threshold in step S5 is:

[0076] Dynamic threshold = baseline threshold × formation correction factor × tunneling disturbance factor × water environment correction factor;

[0077] The baseline threshold is determined based on the subway engineering monitoring and measurement specifications, and the correction coefficient is adaptively and iteratively optimized according to the on-site geological and construction conditions.

[0078] In one embodiment, the ST-GCN spatiotemporal graph convolutional network in step S5 inputs historical deformation sequences, stratum type, burial depth, shield tunneling parameters, groundwater, and surface load, and outputs predicted deformation values ​​and instability risk probabilities for the next 30 minutes, 2 hours, 6 hours, 12 hours, and 24 hours.

[0079] In one embodiment, the criteria for determining the fourth-level early warning in step S6 are as follows:

[0080] Blue indicator: Deformation rate ≤ 50% of the specification limit, trend is stable;

[0081] Yellow alert: Deformation rate is 50%–100% of the standard limit, with an increasing trend;

[0082] Orange alert: Deformation rate exceeds the standard limit by 100% to 150%, indicating a risk of instability;

[0083] Red emergency alert: Deformation rate exceeds the standard limit by more than 150%, which immediately endangers construction safety;

[0084] Early warning information is simultaneously pushed through on-site audio and visual displays, platform pop-ups, SMS, and APP.

[0085] In one embodiment, the closed-loop linkage control strategy in step S7 is as follows:

[0086] Yellow alert: Reduce the advance speed, stabilize the soil pressure, and fine-tune the grouting volume;

[0087] Orange alert: Suspend tunneling, increase grouting pressure and volume, and strengthen support;

[0088] Red emergency alert: Immediately shut down the machine, close the working face, and organize the evacuation of personnel;

[0089] After adjustment, the deformation recovery is continuously monitored to form a closed-loop feedback.

[0090] In one embodiment, the system in step S8 has self-learning capabilities and continuously optimizes the AI ​​prediction model and dynamic threshold coefficient through actual monitoring data, making it suitable for complex strata such as soft soil, gravel, fractured rock, water-rich soil, and soft upper and hard lower strata.

[0091] In one embodiment, the method is applicable to the monitoring of surrounding rock deformation throughout the entire life cycle of subway shield tunnel construction and operation, and is compatible with earth pressure balance shield tunnels and slurry balance shield tunnels. The monitoring data supports BIM+GIS three-dimensional visualization.

[0092] In one embodiment,

[0093] 1. Project Overview

[0094] A certain subway tunnel section, with a total length of 1200m, passes through a composite stratum of silty clay and fine sand with a soft upper layer and a hard lower layer. The overburden thickness is 3.5m to 8m. It is rich in water and surrounded by dense residential areas and municipal pipelines. The monitoring level is Level 1.

[0095] 2. Monitoring Deployment

[0096] Distributed optical fiber: laid along the entire length of the arch crown, arch waist, and arch base;

[0097] Monitoring sections: 1 section every 20m, for a total of 60 sections;

[0098] Sensors: 60 sets of hydrostatic level, 60 sets of tilt sensor, and 60 sets of convergence sensor;

[0099] Visual units: 1 set per 100m, 12 sets in total;

[0100] Edge gateway: 1 unit per 500m, 3 units in total;

[0101] 3. Implementation Steps

[0102] The system powers on and the multi-source array acquires data synchronously at a sampling frequency of 5Hz.

[0103] Real-time preprocessing at the edge to extract deformation rate and cumulative deformation;

[0104] The inversion yielded the crown subsidence, horizontal convergence, and surface subsidence.

[0105] The dynamic threshold is automatically set based on the soft soil strata:

[0106] Daily deformation rate ≤ 0.5mm (blue)

[0107] 0.5~1.0mm (yellow)

[0108] 1.0~1.5mm (orange)

[0109] >1.5mm (red)

[0110] The ST-GCN model predicts the deformation trend over the next 6 hours.

[0111] When the tunneling reached the 382nd ring, the rate of subsidence of the arch crown was monitored to be 0.9 mm / h;

[0112] The system triggered a yellow alert and automatically issued the following instructions:

[0113] The propulsion speed was reduced from 40 mm / min to 20 mm / min

[0114] Grouting volume increased by 15%

[0115] The pressure in the earthwork increased by 0.2 bar.

[0116] After adjustment, the deformation rate dropped to 0.2 mm / h, and the risk was eliminated;

[0117] Data is entered into the database, and the model is updated with early warning parameters for soft soil sections;

[0118] 4. Implementation Results

[0119] Monitoring accuracy: ±0.3mm;

[0120] Warning delay: ≤80ms;

[0121] Early warning accuracy rate: 98.5%;

[0122] No deformation or instability incidents occurred;

[0123] Compared with traditional methods, it saves 40% on monitoring costs and improves safety by 90%.

[0124] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in the present invention, and these should all be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for real-time monitoring and early warning of surrounding rock deformation in subway shield tunnels, characterized in that: Includes the following steps: S1 is equipped with a multi-source fusion monitoring array consisting of distributed fiber optic sensors, point-type high-precision sensors, and machine vision units inside the shield tunnel, and is simultaneously connected to the shield tunneling parameter acquisition interface. S2 synchronously collects surrounding rock deformation data, environmental data, and shield tunneling construction data, and achieves low-latency real-time transmission through an edge computing gateway; S3 performs cleaning, spatiotemporal alignment, and noise reduction on the collected multi-source heterogeneous data, and extracts feature parameters such as cumulative deformation, deformation rate, deformation acceleration, and strain gradient. S4 is based on a mechanical inversion model and a multi-source data fusion model to perform real-time inversion calculations of the surrounding rock arch subsidence, horizontal convergence, vertical settlement, and global deformation field. S5 establishes a dynamic early warning threshold by combining stratum type, burial depth, tunneling parameters and water environment conditions, and uses ST-GCN spatiotemporal graph convolutional network to predict the deformation trend of surrounding rock in the next 30 min to 24 h. S6 triggers four levels of early warning based on cumulative deformation value, deformation rate, and predicted risk value: blue alert, yellow warning, orange alert, and red emergency alert, and displays them in 3D visualization on the BIM platform. Based on the warning level, S7 automatically sends excavation parameter control commands to the shield control system to achieve closed-loop linkage of monitoring, early warning and control. S8 stores all monitoring data, early warning records, and control logs into a spatiotemporal database, and performs incremental learning and iterative updates on the AI ​​prediction model and dynamic thresholds.

2. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: The deployment rules for the multi-source fusion monitoring array in step S1 are as follows: A monitoring section is set up every 20m along the tunnel axis; Each cross section is equipped with distributed optical fibers, static level, tilt sensor and convergence sensor at the arch top, left and right arch waists and arch bottom. One machine vision unit is deployed every 100m to capture the displacement of the surrounding rock surface and the expansion of cracks; Simultaneously collect data on the total thrust of the tunnel boring machine, cutterhead torque, soil chamber pressure, propulsion speed, synchronous grouting volume, and groundwater pressure.

3. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: In step S2, the data acquisition frequency is adaptively adjustable from 1Hz to 10Hz, and industrial Ethernet and 5G dual-redundant transmission are adopted. The transmission delay is ≤50ms. The edge computing gateway realizes local real-time analysis and supports local caching and automatic resume transmission after network disconnection.

4. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: In step S4, the surrounding rock deformation inversion adopts a combination of distributed fiber strain inversion and point sensor calibration, and the final deformation monitoring accuracy is controlled within ±0.1mm~±0.5mm.

5. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: The formula for calculating the dynamic early warning threshold in step S5 is as follows: Dynamic threshold = baseline threshold × formation correction factor × tunneling disturbance factor × water environment correction factor; The baseline threshold is determined based on the subway engineering monitoring and measurement specifications, and the correction coefficient is adaptively and iteratively optimized according to the on-site geological and construction conditions.

6. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: In step S5, the ST-GCN spatiotemporal graph convolutional network input data includes historical deformation sequences, stratum type, burial depth, shield tunneling parameters, groundwater, and surface load. It outputs the deformation prediction values ​​and instability risk probabilities for the next 30 minutes, 2 hours, 6 hours, 12 hours, and 24 hours.

7. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: The criteria for determining the Level 4 early warning in step S6 are as follows: Blue indicator: Deformation rate ≤ 50% of the specification limit, trend is stable; Yellow alert: Deformation rate is 50%–100% of the standard limit, with an increasing trend; Orange alert: Deformation rate exceeds the standard limit by 100% to 150%, indicating a risk of instability; Red emergency alert: Deformation rate exceeds the standard limit by more than 150%, which immediately endangers construction safety; Early warning information is simultaneously pushed through on-site audio and visual displays, platform pop-ups, SMS, and APP.

8. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: The closed-loop linkage control strategy in step S7 is as follows: Yellow alert: Reduce the advance speed, stabilize the soil pressure, and fine-tune the grouting volume; Orange alert: Suspend tunneling, increase grouting pressure and volume, and strengthen support; Red emergency alert: Immediately shut down the machine, close the working face, and organize the evacuation of personnel; After adjustment, the deformation recovery is continuously monitored to form a closed-loop feedback.

9. The method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: In step S8, the system has self-learning capabilities and continuously optimizes the AI ​​prediction model and dynamic threshold coefficient through actual monitoring data. It is suitable for soft soil, gravel, fractured rock, water-rich, and complex strata with soft upper and hard lower layers.

10. A method for real-time monitoring and early warning of surrounding rock deformation in a subway shield tunnel according to claim 1, characterized in that: This method is applicable to the whole life cycle monitoring of surrounding rock deformation during the construction and operation of subway shield tunnels. It is compatible with earth pressure balance shield tunnels and slurry balance shield tunnels, and the monitoring data supports BIM+GIS three-dimensional visualization.

Images