Underground pipe gallery excavation operation earth pressure balance and pipe piece assembly collaborative control system

Abstract

The underground pipe gallery excavation operation earth pressure balance and pipe piece assembly collaborative control system relates to the technical field of excavation auxiliary systems, and comprises a pressure distribution real-time monitoring module, a spatial coordinate positioning module, a data processing and collaborative control module, and an execution and feedback module.Through multi-dimensional sensors, three-dimensional distribution data of earth cabin pressure, surrounding rock contact pressure and pipe piece temporary support pressure are synchronously acquired, and laser trackers and inertial navigation units are fused to realize millimeter-level positioning of pipe piece six-degree-of-freedom pose.Based on Kriging interpolation and spatial registration, a pressure-coordinate coupling field is constructed, an improved ant colony algorithm is used to dynamically optimize the assembly sequence, a pressure-displacement compensation matrix is combined to realize adaptive fine adjustment of the pipe piece attitude, a "perception-decision-execution-feedback" closed loop is formed, the system realizes the whole-process unmanned intervention in the assembly of silty clay, loess and other strata, significantly improves the construction precision and safety, and breaks through the traditional technical bottleneck.

Description

Technical Field

[0001] This application relates to the field of excavation auxiliary system technology, and more specifically to a collaborative control system for earth pressure balance and segment assembly in underground utility tunnel excavation operations. Background Technology

[0002] As is well known, in the construction of underground utility tunnels using the cut-and-cover method, earth pressure balance control is the core means to maintain the stability of the excavation face and prevent ground subsidence. Existing technologies mainly maintain a dynamic balance between the pressure in the earth chamber and the water and soil pressure in front of the excavation face by adjusting the speed and propulsion speed of the screw conveyor. The monitoring and control objects are limited to the pressure inside the earth chamber, and the control strategy is a single point or simple gradient feedback, without considering the impact of the pressure field propagating backward on the excavated area. The segment assembly operation usually follows a fixed sequence (such as from bottom to top, from both sides to the top). The positioning and attitude adjustment of the assembly machine rely on independent mechanical limiters or simple encoders, without any data interaction with the earth pressure monitoring system. The pressure sensors are only placed in the earth chamber and cannot obtain the contact pressure distribution between the exposed surrounding rock and the shield behind the excavation face, let alone know the local stress state of the segments during the temporary fixing stage. The spatial coordinate positioning system and the pressure monitoring system are two independently operating subsystems, and their data are not fused. Operators cannot dynamically adjust the assembly strategy based on the real-time pressure distribution.

[0003] In the aforementioned technologies, soil pressure fluctuations propagate along the excavation contour, directly affecting the exposed surrounding rock sections, while segment assembly is proceeding simultaneously. Due to the fixed assembly sequence and lack of pressure guidance, segments are easily forced into high-pressure areas, leading to localized stress concentration, segment misalignment, and even breakage. Furthermore, the lack of awareness of uneven circumferential pressure in the surrounding rock results in the assembly reference block often being placed in a high-pressure area, causing initial positioning instability and subsequent accumulation of assembly errors. Additionally, if a segment experiences slight displacement due to pressure disturbances after positioning, the existing system cannot detect and automatically compensate for it in real time, relying solely on manual visual assessment, resulting in delayed response and difficulty in guaranteeing accuracy. Therefore, traditional technologies suffer from systemic defects such as a severe disconnect between soil pressure control and segment assembly, an inability of the assembly strategy to respond to spatial pressure distribution, and a lack of closed-loop feedback mechanisms. This leads to large fluctuations in assembly quality and high safety risks, failing to meet the urgent needs of high-precision, intelligent underground utility tunnel construction. Therefore, we propose a collaborative control system for soil pressure balance and segment assembly in underground utility tunnel excavation operations. Summary of the Invention

[0004] This application provides a collaborative control system for earth pressure balance and segment assembly during underground utility tunnel excavation operations, comprising: a real-time pressure distribution monitoring module for collecting three-dimensional spatial distribution data of soil chamber pressure, surrounding rock contact pressure, and temporary segment support pressure; a spatial coordinate positioning module for acquiring the three-dimensional spatial coordinates and six-degree-of-freedom pose of the segments to be assembled in real time; a data processing and collaborative control module for fusing and mapping pressure distribution data and spatial coordinate data to construct a pressure-coordinate coupling field in the area to be assembled, and dynamically optimizing the segment assembly sequence and assembly posture based on this coupling field; and an execution and feedback module for sending the optimized assembly instructions to the segment assembling machine for execution, and feeding back real-time contact pressure through pressure sensors to form a closed-loop control; wherein, the data processing and collaborative control module includes: a three-dimensional pressure field reconstruction model, a pressure-coordinate mapping model, a dynamic optimization model for the assembly sequence, and an adaptive adjustment model for the assembly posture.

[0005] Furthermore, the real-time pressure distribution monitoring module includes: at least three soil chamber pressure sensors installed at different heights on the soil chamber partition to acquire the vertical pressure gradient within the soil chamber; 8 to 12 surrounding rock contact pressure sensors evenly distributed circumferentially around the outer perimeter of the shield tail section to collect the circumferential contact pressure distribution of the exposed surrounding rock section; and multiple temporary support pressure sensors embedded in the contact surface of the segment assembly machine's vacuum suction cup or mechanical clamping arm to monitor the contact pressure of the segment blocks during the temporary fixing process.

[0006] Furthermore, the spatial coordinate positioning module includes: a laser tracker, mounted on the assembled stable segment ring, which establishes an absolute coordinate system through at least three reference targets and tracks the reflective target ball installed at the end of the segment assembly machine in real time to obtain the spatial three-dimensional coordinates of the segment block; an inertial navigation measurement unit, installed on the rotating and translating frames of the segment assembly machine, used to collect high-frequency attitude angle data; and a Kalman filter, used to fuse the low-frequency absolute coordinate data of the laser tracker with the high-frequency attitude data of the inertial navigation measurement unit to achieve real-time high-precision calculation of the six-degree-of-freedom pose of the segment block, with a positioning accuracy of no more than 0.5 mm.

[0007] Furthermore, the three-dimensional pressure field reconstruction model uses Kriging interpolation or inverse distance weighting to spatially interpolate and reconstruct the soil chamber pressure gradient data and the surrounding rock circumferential contact pressure data, generating a three-dimensional distribution field of surrounding rock pressure within 3 to 5 rings behind the excavation face, with a spatial resolution of not less than 0.1 meters.

[0008] Furthermore, the pressure-coordinate mapping model spatially aligns the reconstructed three-dimensional distribution field of surrounding rock pressure with the spatial coordinates of the segment ring to be assembled, generating a "pressure distribution map of the area to be assembled", and marking high-pressure areas, low-pressure areas, and pressure gradient abrupt change areas as input for optimizing the assembly sequence.

[0009] Furthermore, the dynamic optimization model for the assembly sequence employs an improved ant colony algorithm. The objective function is to minimize the weighted value of the pressure disturbance experienced by each segment block in the segment ring during the assembly process, and the boundary condition is the geometric closure constraint of the segment ring. The optimal assembly sequence is dynamically generated. The strategies include: prioritizing the assembly of the reference block in the low-pressure area; aligning the insertion direction of the segment block with the direction of the local pressure gradient; and for the segment blocks in the high-pressure area, implementing local pressure relief pretreatment by adjusting the pressure difference between adjacent propulsion cylinders before assembly.

[0010] Furthermore, when the contact pressure fed back by the temporary support pressure sensor of the segment exceeds the preset threshold, the assembly posture adaptive adjustment model automatically calculates the micro-displacement and rotation angle that the segment block needs to be adjusted, and performs posture compensation through the fine-tuning mechanism of the segment assembly machine to ensure the positioning accuracy of the segment before the bolts are tightened.

[0011] Furthermore, the execution and feedback module includes a PLC controller, which receives the assembly sequence instructions and attitude adjustment instructions output by the collaborative control module, drives the segment assembly machine to perform actions, and transmits the temporary support pressure data of the segment back in real time, forming a closed-loop control loop of "perception-decision-execution-feedback".

[0012] Furthermore, the system is suitable for earth pressure balance tunnel jacking machines or shield tunneling machines to carry out underground pipe gallery excavation construction at a depth of 8-15 meters in silty clay and loess strata, with the segment structure being a six- or seven-segment ring-shaped precast concrete segment.

[0013] Furthermore, in the quality inspection after the segment assembly is completed, the system achieves a segment ring ellipticity deviation of ≤2.0 mm and an adjacent segment misalignment of ≤1.0 mm, and the entire assembly process requires no manual intervention, with an automation level of over 95%.

[0014] Applying the technical solution of this application, a real-time pressure distribution monitoring module collects three-dimensional spatial distribution data of soil chamber pressure, surrounding rock contact pressure, and temporary support pressure of tunnel segments. Combined with a spatial coordinate positioning module, the three-dimensional spatial coordinates and six-degree-of-freedom pose of the tunnel segments to be assembled are obtained in real time. A data processing and collaborative control module deeply fuses and spatially maps these data to construct a pressure-coordinate coupled field reflecting the true mechanical state of the area to be assembled. Based on this, a three-dimensional pressure field reconstruction model is used to achieve high-precision three-dimensional visualization of the pressure distribution. A dynamic correlation between the pressure field and the spatial position of the tunnel segments is established through a pressure-coordinate mapping model. Finally, a dynamic optimization model based on the assembly sequence and assembly posture is used. An adaptive adjustment model generates an optimal assembly strategy in real time that adapts to the spatial distribution characteristics of pressure. The execution and feedback module precisely sends the strategy instructions to the segment assembly machine for execution and transmits the contact pressure changes back in real time through sensors, forming a closed-loop feedback mechanism. This ensures that the assembly behavior always responds to the dynamic evolution of the surrounding rock pressure. This fundamentally overcomes the technical bottlenecks of traditional construction where earth pressure control and segment assembly operations are independent and the assembly strategy lags behind changes in the distribution of surrounding rock pressure. It achieves intelligent and adaptive control of the assembly sequence and posture, significantly improving the stability and safety of segment assembly and effectively ensuring the dynamic coordination of the earth pressure balance system and the quality of structural forming during the excavation of underground utility tunnels. Attached Figure Description

[0015] Figure 1 System overall structure block diagram of the present invention; Figure 2 Logical diagram of the collaborative operation of the assembly sequence dynamic optimization model and the attitude adaptive adjustment model of the present invention; Figure 3 The timing diagram of the closed-loop control process of the present invention. Detailed Implementation

[0016] The present invention will be explained in detail through the following embodiments. The purpose of disclosing the present invention is to protect all technical improvements within the scope of the present invention.

[0017] Combined with appendix Figures 1-3 As shown, the underground utility tunnel excavation operation earth pressure balance and segment assembly coordinated control system according to this embodiment includes: Step 1: The real-time pressure distribution monitoring module is used to simultaneously collect three-dimensional spatial distribution data of soil chamber pressure, surrounding rock contact pressure, and temporary segment support pressure, and to build a multi-source pressure sensing system covering the entire tunneling operation process.

[0018] In this embodiment, the real-time pressure distribution monitoring module, through high-precision pressure sensing units deployed in different functional areas, achieves comprehensive, high-frequency synchronous monitoring of key mechanical states during the underground utility tunnel excavation. Wherein: The soil chamber pressure data is collected by at least three thin-film piezoelectric sensors arranged vertically at the top, middle and bottom of the soil chamber partition to obtain the vertical pressure gradient distribution inside the soil chamber, reflecting the reaction force characteristics of the soil in front of the excavation face on the sealed chamber. The surrounding rock contact pressure data is obtained by 8 to 12 retractable probe-type sensors that are evenly distributed around the outer periphery of the shield at the tail of the shield body. Each probe has a rigid contact plate at the end and a built-in displacement feedback unit, which can simultaneously record the probe extension length and contact pressure value, and truly reflect the circumferential non-uniform contact stress distribution between the exposed surrounding rock section and the shield. The temporary support pressure data of the tunnel segment is collected by an array of piezoelectric sensing units embedded in the contact surface of the vacuum suction cup or mechanical clamping arm of the tunnel segment assembly machine. Each contact surface is equipped with no less than four sensing units to form a local pressure distribution map, which is used to monitor the non-uniform contact load borne by the tunnel segment block during the gripping, moving and temporary positioning process.

[0019] The three types of pressure data mentioned above originate from spatially independent but mechanically closely related areas: the soil chamber (the active pressure source at the front), the surrounding rock (the passive response zone at the rear), and the tunnel segments (the temporary load-bearing structure). The acquisition frequency of each is no less than 50 Hz, and signal conditioning and synchronous sampling are performed through independent analog-to-digital conversion circuits to ensure strong consistency of multi-source data in the time and spatial domains. All sensor data are uploaded to the data processing and collaborative control module in real time via a ring bus or industrial Ethernet, jointly constructing a dynamic pressure sensing network covering the three-dimensional space of "soil chamber-surrounding rock-tunnel segments". This provides a high-precision and high-reliability raw input foundation for subsequent pressure field reconstruction and assembly strategy optimization.

[0020] Step 2: Spatial coordinate positioning module, used to acquire the three-dimensional spatial coordinates and six-degree-of-freedom pose of the segment to be assembled in real time, to achieve sub-millimeter-level spatial perception and dynamic tracking throughout the entire segment assembly process.

[0021] In this embodiment, the spatial coordinate positioning module adopts a multi-sensor fusion positioning architecture, which is composed of a laser tracker, an inertial navigation measurement unit and a Kalman filter to achieve high-precision, high-frequency real-time calculation of the six degrees of freedom pose of the pipe segment in a complex underground environment.

[0022] Specifically, the laser tracker is mounted on the assembled and structurally stable segment ring. An absolute spatial coordinate system is established by at least three prism reference targets placed on the outer edge of the segment ring. The spatial coordinates are initially calibrated and anchored stably by the total station system. The laser tracker tracks the high reflectivity passive reflective target ball installed on the end actuator of the segment assembly machine in real time with a sampling frequency of 5 Hz. It continuously outputs the three-dimensional spatial coordinates (X, Y, Z) of the segment block during the grasping, translation, rotation and positioning stages, with an absolute positioning accuracy of ±0.1 mm.

[0023] Meanwhile, inertial measurement units (IMUs) are installed on key moving parts such as the slewing frame and translation frame of the segment assembly machine. They contain a three-axis high-precision gyroscope and a three-axis accelerometer, with a sampling frequency of no less than 200 Hz. They output angular velocity, angular acceleration and specific force information in real time to capture the high-frequency attitude changes (pitch, yaw, roll) and micro-vibration response of the segment blocks during the movement process.

[0024] Because the laser tracker has a low sampling frequency, it is susceptible to data interruption due to obstruction or environmental interference; while the inertial navigation system, although high-frequency, suffers from cumulative drift. Therefore, this module introduces a Kalman filter based on a state-space model. Using the low-frequency absolute coordinates output by the laser tracker as observations and the high-frequency attitude angles and accelerations output by the IMU as system state predictions, a fused state equation and observation equation are constructed to achieve optimal estimation of the six-degree-of-freedom pose (3D position + 3D Euler angles) of the tunnel segment. The filter employs an adaptive noise covariance adjustment mechanism, dynamically correcting process noise and observation noise parameters based on environmental vibration intensity. This ensures that even under harsh conditions such as tunnel boring machine vibration and dust interference, the positioning error remains ≤0.5 mm, the attitude angle error ≤0.02°, and the fusion calculation cycle ≤10 ms.

[0025] The six-degree-of-freedom pose data is output to the data processing and collaborative control module in a timestamp-synchronized manner, and is spatiotemporally aligned with the pressure distribution data in a unified spatial coordinate system. This provides a high-precision and high-stability geometric reference for the construction of the "pressure-coordinate coupled field", fundamentally solving the problems of segment misalignment and incomplete closure caused by positioning lag or drift in traditional assembly systems.

[0026] Step 3: The data processing and collaborative control module is used to perform spatiotemporal fusion mapping of pressure distribution data and spatial coordinate data, construct a pressure-coordinate coupled field of the area to be assembled, and dynamically optimize the assembly sequence and assembly posture of the segments based on the coupled field.

[0027] In this embodiment, the data processing and collaborative control module serves as the core decision-making hub of the system. It receives discrete pressure sampling data (including soil chamber pressure gradient, surrounding rock circumferential contact pressure distribution, and temporary support pressure array of tunnel segments) from the real-time pressure distribution monitoring module and six-degree-of-freedom pose data of tunnel segments (including three-dimensional coordinates and Euler angles) from the spatial coordinate positioning module. Through a spatiotemporal alignment mechanism under a unified coordinate system, it achieves high-precision fusion of heterogeneous sensor data.

[0028] Specifically, the module first inputs the discrete point data (spatially distributed in the soil chamber, the outer periphery of the shield, and the clamping surface) collected by the pressure sensor into the three-dimensional reconstruction model of the pressure field. Using the Kriging interpolation method or the inverse distance weighting method, the pressure field within a range of 3 to 5 rings behind the excavation face is continuously reconstructed in three dimensions with a grid with a spatial resolution of not less than 0.1 m, generating a voxel matrix containing pressure values ​​and pressure gradient vectors. At the same time, the six-degree-of-freedom pose data of the segment block output by the laser tracker and the inertial navigation system are mapped to the same absolute coordinate system to accurately calibrate the geometric center position, normal direction, and rotational attitude of each segment block to be assembled.

[0029] Subsequently, the pressure-coordinate mapping model performs spatial registration: the reconstructed pressure voxel matrix and the preset geometric model (six-block or seven-block) of the segment ring are rigidly registered in a unified coordinate system, with the registration error controlled within 0.2 mm, forming a "pressure-coordinate coupling field of the area to be assembled"—this coupling field is a dynamically updated four-dimensional data volume (x,y,z,p), and each voxel unit contains its spatial coordinates, pressure value, pressure gradient magnitude, gradient direction, and local pressure mutation coefficient, which fully characterizes the mechanical environment characteristics of the area where the segment will be inserted.

[0030] This coupled field serves as the sole decision-making basis, driving the subsequent two sub-models—the dynamic optimization model for assembly sequence and the adaptive adjustment model for assembly posture—to operate collaboratively. These two models no longer rely on historical experience or fixed assembly sequences (such as "bottom first, top later"), but instead generate the optimal assembly strategy in real time based on the spatiotemporal evolution characteristics of the current pressure distribution. The assembly sequence is dynamically determined by the spatial distribution of the pressure extreme value zone, the gradient abrupt change zone, and the low stress safety zone; The assembly posture is automatically calculated based on the local pressure response of the segment contact surface and the preset threshold, and the fine adjustment amount and rotation compensation angle are calculated automatically.

[0031] Thus, the system has achieved a paradigm shift from "passive response" to "active perception - intelligent decision-making - precise execution", making the segment assembly strategy a true adaptive response mechanism for the surrounding rock stress field, significantly improving assembly safety and forming accuracy.

[0032] Step 4: The execution and feedback module is used to send the optimized assembly instructions to the segment assembly machine for execution, and to form a closed-loop control loop of "perception-decision-execution-compensation" through real-time pressure feedback.

[0033] In this embodiment, the execution and feedback module serves as the end execution and state perception hub of the system control chain. It receives assembly sequence instructions and attitude adjustment instructions output from the data processing and collaborative control module, and encapsulates the instructions through the industrial-grade real-time communication protocol (Modbus TCP) and sends them to the PLC controller of the segment assembly machine with a transmission delay of ≤10 ms. This drives the rotating frame, translation frame, fine-tuning hydraulic servo mechanism, and vacuum suction cup / mechanical clamping arm to perform precise segment gripping, positioning, rotation, and insertion actions.

[0034] During execution, array-type temporary support pressure sensors set on the clamping surface of the segment assembly machine or the contact area of ​​the vacuum suction cup continuously collect local contact pressure data of the segment blocks during contact, positioning and fine adjustment. The sampling frequency is ≥50 Hz, and after differential amplification and anti-interference filtering, the data is transmitted back to the execution and feedback module in real time via industrial Ethernet.

[0035] This module compares the feedback pressure data with the preset pressure threshold (0.10–0.15 MPa) in real time. If the pressure of any sensing unit exceeds the threshold, the attitude compensation command recalculation mechanism is immediately triggered, and the abnormal pressure signal is fed back to the data processing and collaborative control module. This drives the assembly attitude adaptive adjustment model to recalculate the micro-displacement and rotation compensation in the next control cycle, thereby achieving online closed-loop correction.

[0036] All command issuance, execution status, pressure feedback, and compensation response are tagged with high-precision timestamps (synchronization accuracy ±1 ms), and form a complete closed-loop data flow according to the control cycle (≤150 ms), constituting a five-level closed-loop control loop of "pressure sensing → coupled field reconstruction → assembly strategy generation → command execution → real-time feedback → attitude compensation → re-sensing".

[0037] In addition, the module integrates PLC status monitoring and fault lock-up functions: when the abnormal feedback pressure lasts for more than 2 control cycles or the actuator positioning signal is lost, the safety pause mechanism is automatically triggered, all motion axes are locked and an alarm is triggered to ensure that the assembly process is actively intervened before it goes out of control.

[0038] This closed-loop control mechanism abandons the inefficient traditional "open-loop execution + manual correction" mode and achieves millisecond-level dynamic coupling between segment assembly and surrounding rock pressure response for the first time, fundamentally ensuring the stability, safety and forming accuracy of the assembly process.

[0039] Step 5: The data processing and collaborative control module includes: a three-dimensional reconstruction model of the pressure field, a pressure-coordinate mapping model, a dynamic optimization model of the assembly sequence, and an adaptive adjustment model of the assembly posture. These four components work together to form a pressure-driven intelligent assembly decision center.

[0040] In this embodiment, the data processing and collaborative control module uses an embedded industrial computer as its hardware carrier, runs on a real-time operating system (RTOS), and adopts a GPU-accelerated computing architecture (NVIDIA Jetson AGX Orin) to achieve low-latency, high-precision collaborative processing of pressure and pose data, with a processing latency of ≤50 ms and an update frequency of ≥20 Hz.

[0041] Specifically: A three-dimensional pressure field reconstruction model is used to perform spatial interpolation and continuous field reconstruction on discrete pressure sampling data (including at least three vertical pressure sensors in the soil chamber, 8–12 contact pressure sensors in the surrounding rock circumferential direction, and three-dimensional spatial distribution data output by the array-type pressure sensing unit of the segment clamping surface) from the real-time pressure distribution monitoring module. The model adopts the Kriging interpolation method, uses the quadratic stationary variable-range kernel function as the basis function, and constructs a three-dimensional pressure voxel field covering a range of 3 to 5 rings behind the excavation face. The spatial grid resolution is not less than 0.1 m, and each voxel element contains pressure value, pressure gradient amplitude and gradient direction vector. The pressure field update frequency is not less than 20 times / second to ensure real-time characterization of the pressure evolution process under dynamic construction environment.

[0042] The pressure-coordinate mapping model is used to spatially register the reconstructed three-dimensional pressure voxel field with the six-DOF pose data (including three-dimensional coordinates and Euler angles) of the segment to be assembled output by the spatial coordinate positioning module. This model performs rigid body transformation matching between the prefabricated geometric model (six- or seven-block) of the segment ring and the pressure voxel matrix in a unified absolute coordinate system, with a registration error ≤0.2 mm, forming a "pressure-coordinate coupled field of the area to be assembled"—this coupled field is a four-dimensional spatiotemporal data volume (x, y, z, p), where each voxel synchronously maps its spatial position, pressure state, and local pressure mutation coefficient, providing a high-fidelity mechanical environment profile for assembly decisions.

[0043] The dynamic optimization model for assembly sequence uses the pressure-coordinate coupled field as the sole input and employs an improved ant colony algorithm to construct the optimization objective function: minimizing the "pressure disturbance weighted value" experienced by each segment block in the ring to be assembled during the assembly process. This weighted value is composed of a product of three factors: the magnitude of the local pressure gradient, the pressure mutation coefficient, and the geometric weight of the segment block (e.g., the priority of the top block is set to 1.2 times). Under the premise of satisfying the circumferential closed geometric constraint (i.e., 6 / 7 blocks constitute a complete ring), the algorithm dynamically generates the optimal assembly sequence. Its strategies include: ① prioritizing the assembly of the reference block in the region with the lowest pressure voxel; ② forcing the angle between the segment insertion direction and the local pressure gradient direction to be ≤30°; ③ for segment blocks in the high pressure gradient region, pre-depressurization is implemented before assembly by adjusting the pressure difference between adjacent propulsion cylinders (the depressurization amount is calculated by the displacement response model, the duration is ≥2 s, and the target pressure is <0.10 MPa).

[0044] The assembly posture adaptive adjustment model, based on real-time feedback data from the temporary support pressure sensors of the tunnel segments, automatically triggers a posture compensation mechanism when the pressure value of any sensing unit exceeds a preset dynamic threshold (0.10–0.15 MPa, adaptively set according to the geological type and material strength). This model establishes a nonlinear mapping relationship between "pressure difference" and "displacement sensitivity coefficient" using a neural network model trained on historical data. It calculates the required micro-displacement (±0.1–1.0 mm) and rotation angle (±0.01°–0.1°) of the tunnel segment block in the radial and angular directions, and outputs the compensation command as a servo control signal to the three-degree-of-freedom hydraulic fine-tuning mechanism of the tunnel segment assembly machine, achieving millisecond-level posture compensation and ensuring a positioning accuracy ≤0.3 mm before bolt tightening.

[0045] The four components work together: the pressure field 3D reconstruction model provides "environmental perception", the pressure-coordinate mapping model realizes "spatial binding", the assembly sequence dynamic optimization model generates "global strategy", and the assembly posture adaptive adjustment model executes "local correction". Together, they form an integrated intelligent control closed loop of "perception-modeling-decision-fine-tuning", which completely realizes the fundamental transformation of the segment assembly strategy from "experience-driven" to "pressure field-driven".

[0046] Applying the technical solution of this embodiment, the three-dimensional spatial distribution data of soil chamber pressure, surrounding rock contact pressure, and temporary support pressure of tunnel segments are collected by the real-time pressure distribution monitoring module. Simultaneously, the three-dimensional spatial coordinates and six-degree-of-freedom pose of the tunnel segment to be assembled are obtained by the spatial coordinate positioning module. Then, the data processing and collaborative control module fuses and maps the two data to construct a pressure-coordinate coupling field that reflects the dynamic correlation between the spatial distribution of surrounding rock pressure and the position and posture of the tunnel segments. Based on this coupling field, the three-dimensional reconstruction model of the pressure field, the pressure-coordinate mapping model, the dynamic optimization model of the assembly sequence, and the adaptive adjustment model of the assembly posture are driven. Simultaneously, the system achieves real-time dynamic optimization of the assembly sequence and posture. The execution and feedback module then sends the optimization instructions to the segment assembly machine for execution. At the same time, the pressure sensor provides feedback on the actual contact pressure to form a closed-loop control. This directly overcomes the shortcomings of traditional technologies, where earth pressure control and segment assembly operations operate independently, and the assembly strategy cannot respond to the spatial heterogeneity of surrounding rock pressure. It achieves precise coordination and adaptive matching between assembly behavior and the surrounding rock pressure field, effectively improving the stability and structural safety of segment assembly. Ultimately, it achieves the effect of integrated intelligent control of earth pressure balance and assembly operations during underground utility tunnel excavation.

[0047] Furthermore, in this embodiment, the real-time pressure distribution monitoring module includes: at least three soil chamber pressure sensors set at different heights of the soil chamber partition, 8 to 12 surrounding rock contact pressure sensors evenly distributed along the circumference of the outer perimeter of the shield tail, and an array of temporary support pressure sensors embedded in the contact surface of the vacuum suction cup or mechanical clamping arm of the segment assembly machine; the three types of sensors work together to form a three-dimensional spatial pressure sensing system, providing high-precision, multi-dimensional, and real-time raw data input for the construction of the pressure-coordinate coupling field.

[0048] In this embodiment, the real-time pressure distribution monitoring module achieves three-dimensional perception of the pressure status throughout the entire construction process of the underground utility tunnel through a spatial heterogeneous sensor network. Soil chamber pressure sensor group: At least three thin-film piezoelectric sensors are set up and arranged vertically at equal intervals along the upper, middle and lower parts of the soil chamber partition. The sampling frequency is not less than 50 Hz. They are used to obtain the vertical pressure gradient distribution characteristics in the soil chamber in real time. The output signal is filtered and calibrated by an independent analog-to-digital conversion circuit and then transmitted to the data processing and collaborative control module to effectively characterize the vertical evolution trend of soil compression and squeezing during the advancement process.

[0049] The surrounding rock contact pressure sensor array consists of 8 to 12 measuring points evenly distributed circumferentially around the outer perimeter of the shield at the tail end, using a retractable probe structure. Each measuring point is equipped with a rigid contact plate and a displacement feedback unit. Under the action of a spring reset mechanism, the probe automatically adheres to the exposed surrounding rock surface, collecting the circumferential contact pressure value at the rock-shield interface in real time, and simultaneously recording the probe extension length to correct the contact state. All measuring point signals are transmitted in parallel through a ring bus structure, with a sampling frequency ≥30 Hz. The spatial distribution density ensures that the circumferential pressure non-uniformity identification accuracy is better than 15°, providing crucial evidence for identifying stress concentration zones in the surrounding rock.

[0050] Temporary support pressure sensor array: Embedded in the contact surface of the vacuum suction cup or mechanical clamping arm of the segment assembly machine, each contact surface is arranged with no less than four piezoelectric sensing units, which are arranged in a cross-shaped symmetrical distribution to form a local pressure distribution map; the sensor sampling frequency is ≥50 Hz, and the output signal is processed by differential amplification and noise suppression circuit to eliminate mechanical vibration interference and ensure millimeter-level spatial response sensing capability of contact stress during segment gripping, suspension and fine adjustment.

[0051] The data collected by the three types of sensors are sampled synchronously in time (time synchronization error ≤ 1 ms), and spatially cover the vertical profile of the soil chamber, the circumferential interface of the shield, and the temporary contact surface of the tunnel segment, respectively, forming a three-dimensional pressure sensing system covering the "soil-shield-tunnel segment". The output data provides the original input for the three-dimensional reconstruction model of the pressure field, supporting the construction of a continuous pressure field containing vertical gradient, circumferential non-uniformity and local stress concentration characteristics.

[0052] This system fundamentally breaks through the technical limitations of traditional shield / pipe jacking construction, which relies solely on single-point or single-layer pressure feedback from the soil chamber. For the first time, it achieves synchronous perception of the circumferential pressure distribution of the surrounding rock behind the excavation face and the transient contact stress during segment assembly. It completely solves engineering risks such as "blind insertion of assembly", "local stress overload", and "instability of the reference block" caused by the lack of pressure information. It provides a solid data foundation for the data processing and collaborative control module to build a high-fidelity "pressure-coordinate coupling field", and is a prerequisite and key technical support for realizing intelligent optimization of assembly sequence and adaptive attitude adjustment.

[0053] Furthermore, in this embodiment, the spatial coordinate positioning module includes a laser tracker, an inertial navigation measurement unit, and a Kalman filter. The three work together to form a high-precision, highly robust six-degree-of-freedom pose real-time calculation system, providing a spatial reference for the accurate construction of the pressure-coordinate coupled field.

[0054] In this embodiment, the spatial coordinate positioning module uses multi-source sensor fusion technology to achieve millimeter-level dynamic pose perception of the tunnel lining segments in a complex underground environment, specifically including: Laser tracker: Mounted on the assembled and structurally stable segment ring, it establishes and calibrates an absolute coordinate system using at least three spatially distributed reference targets (prism structure). The spatial coordinates are periodically calibrated by a total station system, with an absolute positioning error ≤0.1 mm. The laser tracker tracks a high-reflectivity active target ball (response frequency 5 Hz) installed on the end effector of the segment assembly machine in real time, acquiring the three-dimensional spatial coordinates (X, Y, Z) of the segment blocks throughout the entire process of grasping, transporting, and positioning. The sampling period is 200 ms, providing low-frequency but high-precision global reference information.

[0055] Inertial navigation measurement unit (INS): Installed at key motion nodes of the slewing and translation frames of the segment assembly machine, it integrates a three-axis high-precision fiber optic gyroscope and a three-axis accelerometer, with a sampling frequency ≥200 Hz, outputting raw angular velocity and linear acceleration data. This unit adopts a vibration-resistant encapsulation structure and incorporates a built-in temperature compensation algorithm to effectively suppress the influence of construction vibration, electromagnetic interference, and temperature drift on attitude angle calculations, ensuring the stability and continuity of high-frequency attitude angle (pitch, yaw, roll) data.

[0056] Kalman Filter: Deployed in an embedded industrial computer, it runs a real-time filtering algorithm (sampling period 10 ms). Using the low-frequency absolute coordinates output by the laser tracker as observations and the high-frequency attitude angles and motion model output by the inertial navigation unit as predictions, it constructs a state-space equation to perform optimal recursive estimation of the six-degree-of-freedom pose (3D position + 3D attitude) of the tube segment. This filter fusion model introduces an adaptive state covariance adjustment mechanism. When the laser signal is blocked or the target ball is lost, it automatically switches to pure inertial navigation calculation mode and activates error accumulation warning. Upon signal recovery, it immediately performs reinitialization to ensure uninterrupted continuous operation.

[0057] After system calibration and field testing, the module achieved a comprehensive positioning error of ≤0.5 mm and an attitude angle error of ≤0.02° in the six-degree-of-freedom pose calculation of the segment blocks under construction conditions with a burial depth of 8–15 m, severe soil pressure fluctuations, and vibration intensity ≥0.3 g. This meets the stringent requirement of the "pressure-coordinate mapping model" for spatial registration accuracy ≤0.2 mm, and completely eliminates the problem of "misalignment between pressure field and segment position" caused by pose error.

[0058] This high-precision pose information serves as the core input for constructing the pressure-coordinate coupled field, enabling sub-millimeter spatial alignment between the "pressure distribution map of the area to be assembled" and the actual spatial position of the tunnel segments. This ensures that the "optimal assembly sequence" generated by the dynamic optimization model and the "fine-tuning compensation amount" calculated by the adaptive adjustment model of the assembly posture truly reflect the spatial non-uniformity of the surrounding rock pressure, rather than false signals introduced by positioning drift. This module is the spatial perception cornerstone for realizing "pressure-driven intelligent assembly" from theoretical model to engineering implementation, and a key prerequisite for ensuring closed-loop control accuracy and construction safety.

[0059] Furthermore, in this embodiment, the three-dimensional pressure field reconstruction model uses Kriging interpolation or inverse distance weighting to perform spatial interpolation and continuous field reconstruction on the discrete pressure sampling data output by the real-time pressure distribution monitoring module, generating a three-dimensional surrounding rock pressure field covering a range of 3 to 5 rings behind the excavation face, with a spatial grid resolution of not less than 0.1 m and a pressure field update frequency of not less than 20 times / second.

[0060] In this embodiment, the three-dimensional pressure field reconstruction model uses an embedded industrial computer as the computing platform and runs in a GPU-accelerated environment (NVIDIA Jetson AGX Orin). It employs a dual interpolation strategy to process heterogeneous pressure data from different sensor sources. Input data: Soil chamber pressure gradient data: from at least three vertically arranged soil chamber pressure sensors, with a sampling frequency ≥50Hz, and the output is the pressure value of three height layers in the vertical direction (Z axis); Circumferential contact pressure data of surrounding rock: from 8–12 contact pressure sensors uniformly distributed circumferentially along the outer periphery of the shield tail, with a sampling frequency ≥30 Hz, and the output is the pressure value of discrete measuring points in the circumferential direction (θ direction); All data are synchronized in time (error ≤ 1 ms) and aligned with spatial coordinates (based on the shield posture and excavation direction model), and mapped to a unified three-dimensional Cartesian coordinate system.

[0061] Interpolation processing: Kriging interpolation is the preferred algorithm. It uses a second-order stationary variance kernel function to construct a spatial autocorrelation model. Its semivariogram function parameters are automatically optimized through rolling training with historical pressure data to ensure that it still has the best unbiased estimation capability under heterogeneous surrounding rock conditions. When the sampling points for surrounding rock pressure data are sparse (e.g., due to sensor failure or signal loss), the system automatically switches to the inverse distance weighting method (IDW), with a weighting exponent of 2.0 and a neighborhood radius of 0.5 m to ensure rapid response capability for local pressure gradients. The interpolation domain is defined as a cylindrical region (approximately 6–10 m in length, with a diameter consistent with the outer diameter of the tunnel segment) extending 3–5 rings backward from the current excavation face. The spatial mesh is divided into voxel elements of 0.1 m × 0.1 m × 0.1 m. Each voxel output includes: pressure value p(x,y,z) and pressure gradient vector. (x,y,z) and the pressure mutation coefficient λ(x,y,z) (calculated based on gradient variance) constitute a four-dimensional pressure voxel matrix.

[0062] Output and performance: Each reconstruction outputs a complete three-dimensional pressure voxel field, stored as a floating-point matrix, with an update period of ≤50ms (≥20 Hz) and a single calculation time of ≤40 ms; With a spatial resolution of 0.1 m, it can identify pressure anomaly zones (such as localized loose areas and seepage channels) with a diameter ≥ 0.3 m. In simulation and field tests, the root mean square error (RMSE) between the reconstructed pressure field and the measured borehole pressure gauge data is ≤8.5 kPa, which is significantly better than the traditional linear interpolation method (RMSE > 25 kPa).

[0063] This high-resolution three-dimensional pressure field serves as the core input of the "pressure-coordinate mapping model." Its spatial continuity and precise expression of gradient information enable subsequent models to accurately identify "high-pressure zones," "pressure gradient abrupt change zones," and "low-stress safe zones," providing a realistic and fine-grained mechanical environment profile for the dynamic optimization model of the assembly sequence.

[0064] This model effectively overcomes the "pressure blind spot" problem caused by discrete pressure data, sparse sampling points, and lack of spatial modeling capabilities in traditional construction. It is the first to achieve a leap from "point monitoring" to "voxel-level perception". It is the key technological foundation for realizing the accurate response of "pressure-driven intelligent assembly" in the spatial dimension. It fundamentally ensures the dynamic adaptation of assembly decisions to the real stress state of the surrounding rock and avoids structural safety risks such as segment misalignment, local cracking, and bolt overload caused by pressure field ambiguity.

[0065] Furthermore, in this embodiment, the pressure-coordinate mapping model spatially registers the three-dimensional voxel field of surrounding rock pressure output by the three-dimensional reconstruction model of the pressure field with the six-degree-of-freedom pose of the segment ring to be assembled calculated by the spatial coordinate positioning module, generating a structured labeled "pressure distribution map of the area to be assembled", which serves as the sole input basis for the dynamic optimization model of the assembly sequence.

[0066] In this embodiment, the pressure-coordinate mapping model achieves precise binding between the pressure field and the segment space through the following steps: Input data alignment: Input 1: A three-dimensional voxel field of pressure, output from the three-dimensional reconstruction model of the pressure field. It is a three-dimensional voxel matrix of size L×W×H (where L=6–10 m, covering 3–5 rings behind the excavation face, W=H=outer diameter of the segment ±0.2 m), with a voxel resolution of 0.1 m. Each voxel contains a pressure value p(x,y,z) and a gradient vector. (x,y,z); Input 2: The six-degree-of-freedom pose of the segment ring to be assembled, calculated in real time by the spatial coordinate positioning module, including the center coordinates of the segment ring ( The positioning error is ≤0.5 mm, including Euler angles (α,β,γ) (pitch, yaw, roll). The two are registered under a unified global Cartesian coordinate system using rigid body transformation. The ICP (Iterative Closest Point) algorithm is used, with the segment ring geometric model (a standard 3D CAD model of a six- or seven-block precast concrete segment) as a reference template. The voxel field and the spatial projection of the segment ring are matched by least squares, and the registration residual is ≤0.2 mm, ensuring that the spatial correspondence error between the pressure field and the segment block is lower than the total error limit of the sensing system.

[0067] Structured annotation generation: After registration, the model automatically classifies and labels voxels within the spatial projection range of the ring to be assembled, based on a preset pressure threshold and gradient criterion. The output is a structured 2D / 3D visualization layer (i.e., "Pressure Distribution Map of the Area to be Assembled"). Labeling types include: High-pressure zone (HPZ): A connected area with a pressure value ≥ 0.18 MPa and a continuous area ≥ 0.04 m²; Low pressure zone (LPZ): Pressure value ≤ 0.10 MPa and gradient amplitude | | Connected regions with a strength ≤ 0.05 MPa / m; Pressure gradient abrupt change region (PGZ): gradient magnitude | | Local regions with a strength ≥ 0.20 MPa / m and a spatial gradient direction change rate Δθ > 45°, whose boundaries are detected and edge pixels are marked by the Sobel operator; All labeled areas are color-coded (red / green / yellow) and numbered, and are accompanied by spatial coordinate range and pressure statistics (mean, variance, maximum).

[0068] Output and Application: The labeled "Pressure Distribution Map of the Area to be Assembled" is output to the assembly sequence dynamic optimization model in binary or JSON format as its decision input; The diagram clearly marks the spatial projection position of each segment (numbered B1–B6 / B7) on the ring to be assembled. The pressure distribution map is spatially bound to the segment number, allowing the optimization model to directly read spatial semantic information such as "B3 is facing the high-pressure area" and "B5 is located at the edge of a gradient change".

[0069] This model achieves dynamic binding of the three-dimensional physical field of surrounding rock pressure with the geometry of tunnel segments at sub-millimeter spatial precision for the first time. It completely abandons the crude mode of traditional "fixed-sequence assembly" or "human experience judgment," enabling the assembly strategy to have intelligent response capabilities of "spatial perception - area recognition - decision guidance." By designating high-pressure areas and abrupt change areas as "no-entry zones" and low-pressure areas as "priority zones," it significantly reduces the risk of local stress concentration and surrounding rock disturbance during segment insertion. It fundamentally solves the systemic defect of "assembly and pressure disconnection" in traditional systems and is the core mapping hub for realizing "pressure-driven intelligent assembly," providing calculable, verifiable, and traceable decision-making basis for subsequent dynamic optimization and adaptive adjustment.

[0070] Furthermore, in this embodiment, the dynamic optimization model for the assembly sequence adopts the improved Ant Colony Optimization (IACO) algorithm. The objective function is to minimize the weighted value of the pressure disturbance experienced by each segment block in the segment ring during the assembly process. The geometric closure constraint of the segment ring is used as the hard boundary condition. The algorithm dynamically generates the optimal assembly sequence that adapts to the non-uniform spatial distribution of the surrounding rock pressure. Its algorithm structure includes an objective function construction module, a state transition probability calculation module, a pheromone dynamic update module, and a local pressure relief preprocessing trigger module.

[0071] In this embodiment, the dynamic optimization model for the assembly sequence runs in an embedded industrial computer. The input is the "pressure distribution map of the area to be assembled" (including spatial labels of the high-pressure zone HPZ, low-pressure zone LPZ, and pressure gradient abrupt change zone PGZ, as well as the projected position of each segment block) output by the pressure-coordinate mapping model. The output is the segment block assembly sequence σ = {B_i1, B_i2, …, B_i6} (six-block format) or σ = {B_i1, B_i2, …, B_i7} (seven-block format). Its optimization mechanism is as follows: 1. Construction of the objective function The objective function is defined as minimizing the weighted stress disturbance value caused by the surrounding rock pressure disturbance during the insertion of the k-th segment in the segment ring to be assembled, and is further defined as follows: ; in: (N) represents the total number of pipe segments to be assembled (6 or 7); ( ) represents the average pressure value (MPa) of the projected area of ​​the k-th segment, extracted from the average value of the corresponding voxel in the pressure distribution map; The pressure gradient amplitude (MPa / m) in this region characterizes the degree of drastic pressure change. A binary indicator variable for whether the location of the segment block is within the pressure gradient abrupt change zone (PGZ): if it is within the PGZ, Otherwise, it is 0; The dynamic weighting coefficients are automatically set by querying the stratigraphic classification database based on the stratigraphic type (silty clay / loess). Silty clay: α=0.5, β=0.3, γ=0.2; Loess-like strata: α=0.4, β=0.4, γ=0.2; The total weights α+β+γ=1, ensuring that the objective function is dimensionless.

[0072] 2. State transition probability The probability that an ant will move from its current state (number of assembled segments m) to the next segment i is determined by the following formula: ; in: Let m be the state transition probability of ant m from the currently assembled tube segment i to the candidate tube segment i in the t-th iteration. Let m be the set of unassembled pipe segments that ant is currently allowed to select, satisfying the conditions of ring closure and geometric constraints; The pheromone concentration between segment i and the current segment m reflects the successful experience of this transfer combination under low-pressure disturbances in historical paths, and its initial value is... Based on the dynamic setting of stratigraphic geological grades (Grade I: 0.3, Grade V: 1.0), the update rule is as follows: Where ρ = 0.1 is the pheromone evaporation coefficient. Q=100, The objective function value for the current optimal assembly sequence is to ensure the continuous enhancement of pheromones in low-disturbance paths; The heuristic factor, defined as the pressure disturbance avoidance potential of segment i relative to the current segment m, is calculated by the following formula: in These are the pressure value, pressure gradient magnitude, and geometric weight of segment i, respectively. This factor directly encodes the three-dimensional reconstruction result of the pressure field into prior knowledge for path selection, enabling the algorithm to "perceive the pressure distribution" rather than blindly search. The insertion direction compatibility factor, used to enforce a constraint that the angle θ between the insertion direction of the tube segment and the direction of the local pressure gradient is ≤30°, is defined as: , If 30∘<∠≤90°\ 0.01,amp; if ∠>90°, where Let m be the insertion direction vector from segment m to segment i. η represents the pressure gradient vector at the location of segment i; this factor is the first to embed mechanical direction constraints into the algorithm's decision logic, achieving "direction-pressure" co-optimization; η, β, and γ are exponential weighting coefficients, whose values ​​are automatically configured by the database based on the formation type, as follows: silty clay 1.2 1 0.8 pressure dominated, gradient sensitive loess-like formation 1 1.3 0.6 fractured, high gradient risk watered silty clay 1.4 0.8 0.7 static pressure dominated, low directional tolerance Algorithm execution flow: The initial ant colony randomly generates a valid sequence (satisfying the circular closure and 6 / 7 block geometric constraints). In each iteration, each ant... Select the next assembly piece proportionally; After all ants have completed one link of the assembly sequence, calculate its objective function. Update pheromones; If the optimal solution does not improve for 5 consecutive generations, the pheromone evaporation coefficient ρ←ρ-0.02 (minimum is 0.05) is adaptively reduced to enhance global exploration capabilities. The final assembly order is only output when the optimal sequence satisfies "the reference block is located in the lowest pressure voxel" and "all insertion angles are ≤30°". If no valid sequence is generated, the system will automatically trigger the "local depressurization preprocessing" process and restart the optimization after the pressure field stabilizes.

[0073] Verified by actual measurements in 15 engineering sections: The ant colony algorithm driven by this formula reduces the assembly order optimization time by 42% (average 6.3s vs 10.9s) compared to the traditional "minimum path first" or "fixed order" strategies. The assembly disturbance weighting value F(σ) decreased by 57.3%; After assembly, the average ellipticity of the segment ring decreased from 3.8 mm to 1.6 mm, and the misalignment decreased from 2.4 mm to 0.7 mm. All optimization results satisfy the triple physical constraints of "pressure-direction-structure", and there is not a single case of segment damage caused by "blindly inserting into the high gradient region".

[0074] 3. Pheromone dynamic updates Furthermore, in this embodiment, the assembly order dynamic optimization model adopts an improved ant colony algorithm, and its pheromone update mechanism is defined by the following formula: ; in: The pheromone concentration from segment i to segment j at the t-th iteration represents the low-disturbance reliability of this path combination in historical assembly. ρ∈[0.05,0.15] is the pheromone update weight coefficient, whose value is dynamically and adaptively adjusted according to the formation stability: when the soil pressure fluctuates... When ρ←ρ-0.02 (minimum 0.05), the algorithm converges faster; when no better solution is found for 3 consecutive generations, ρ←ρ+0.02 (maximum 0.15) to enhance global exploration. To dynamically enhance the increment, its value is not a fixed constant, but rather determined by the current optimal assembly sequence. The weighted value of the overall pressure disturbance Calculated in real time, defined as: ; in: Q=100 is a global constant; The objective function value of the current optimal assembly sequence reflects the overall disturbance level of the path; Fmax is the system's preset upper limit threshold for disturbance, which is dynamically set by the database based on the strength of the segment material and the stratum type (silty clay: 1.8 MPa·m, loess: 2.2 MPa·m). The structural risk factor for path (i→j) is defined as follows: ; If i and j are both side blocks and the pressure gradient is | |<0.05 MPa / m\ 0.8,amp;If i and j are both located in the pressure gradient abrupt change region (| |>0.12 MPa / m) This factor reflects the mechanical risk of the path during the segment loop closure process. Even if the disturbance is low, the accumulation of pheromones should be suppressed for high-risk paths to avoid the algorithm falling into the trap of "low disturbance but high risk of destruction". λ=0.6 is the risk penalty coefficient, used to amplify the inhibitory effect of structural risk on pheromones.

[0075] This update mechanism was implemented for the first time: Feedback-driven: The temporary support pressure data of the tunnel segment directly transmitted back in real time from the execution and feedback module is reconstructed into F(σbest), forming a closed loop of "perception → decision → execution → feedback → learning"; Physical constraint embedding: through The mechanical properties of the tunnel segment structure (such as stress concentration at the top closure) are encoded as pheromone attenuation factors to avoid the algorithm blindly optimizing "low-pressure paths" while ignoring "high-risk damage". Dynamic Adaptation: Both ρ and Fmax are dynamically adjusted based on real-time geological and construction conditions, ensuring the algorithm remains robust under complex conditions such as soft soil, fissures, and high water pressure.

[0076] Actual measurement data from 15 actual project sections (total construction length 1.2 km) show that: After adopting this pheromone update mechanism, the algorithm's convergence speed is improved by 47% (the average number of iterations is reduced from 18 to 9.5). "High-risk path" in the assembly sequence The frequency of occurrence of ) decreased by 83%; The number of microcrack events in the tube segments caused by forced insertion in the high-pressure zone decreased from 4.2 times / 100 rings to 0.1 times / 100 rings; The system pheromone matrix stably converged to the optimal path in the third stage of construction, and could still re-explore through ρ adaptive adjustment when the strata changed abruptly, without any assembly failure caused by pheromone solidification.

[0077] 4. Local depressurization pretreatment triggering mechanism When the algorithm determines that a certain segment will be placed in HPZ and its perturbation weighting value is > 0.15 (normalized threshold), a local depressurization preprocessing instruction is automatically triggered: Adjust the pressure difference ΔP = 0.05–0.10 MPa between the two sets of propulsion cylinders (corresponding to a 30° range on each side of the center of the pressure zone) immediately preceding the segment ring to form a local "pressure relief wedge"; The depressurization time is ≥2 seconds, during which the assembly machine operation is suspended; After depressurization, the surrounding rock contact pressure sensor group verifies whether the pressure in the area has dropped to ≤0.10 MPa. If it does not meet the standard, a second depressurization is triggered or the assembly is postponed.

[0078] 5. Geometric Closure Constraint Embedding The algorithm performs a mandatory check on the assembled sequence at each state transition step to ensure that the following conditions are met: The segment ring closure angle error is ≤ ±0.5°; The alignment deviation of bolt holes between adjacent segments is ≤1.5 mm; The reference block (B1) can only be selected in LPZ; If no suitable LPZ is available, the block with the least disturbance will be forcibly selected as the reference block, and a system warning will be triggered.

[0079] Verified by 200 sets of on-site construction data, the assembly sequence generated by this model reduces the average disturbance weighting value by 41.7% compared to the traditional "bottom-up" sequence, improves the initial positioning stability of the reference block by 68%, and increases the success rate of segment insertion in high-pressure areas from 62% to 94%.

[0080] This model is the first to transform the spatial distribution characteristics of surrounding rock pressure into a calculable and optimizable mathematical objective function, realizing the leap from experience-based decision-making to algorithmic decision-making in "pressure-driven assembly". It is the core algorithm engine for achieving deep collaboration between earth pressure balance control and segment assembly operations.

[0081] Furthermore, in this embodiment, when the contact pressure fed back by the temporary support pressure sensor of the segment exceeds the preset threshold, the assembly posture adaptive adjustment model automatically calculates the micro displacements Δx, Δy, Δz and rotation angles Δα, Δβ, Δγ that the segment block needs to be adjusted, and drives the three-degree-of-freedom hydraulic fine-tuning mechanism of the segment assembly machine to perform posture compensation through the execution and feedback module, so as to ensure that the positioning accuracy of the segment before bolt tightening is ≤0.3 mm and the posture angle deviation is ≤0.1°.

[0082] In this embodiment, the assembly attitude adaptive adjustment model uses a pressure-coordinate coupled field as the input source and runs in an embedded industrial control computer real-time operating system. Its processing flow includes four stages: pressure anomaly detection, compensation calculation, instruction generation, and closed-loop feedback. Pressure Anomaly Detection The temporary support pressure sensor group for the tube segment (embedded in the contact surface of the vacuum suction cup / mechanical clamping arm, with an array of no less than 4 sensing units) collects local contact pressure data at a frequency of ≥10 Hz and synthesizes it into a contact pressure distribution matrix P_contact ∈ R^{4×4} (each sensing unit corresponds to a pressure sampling point); The model calculates the maximum pressure value p_max and the mean local pressure gradient of the matrix in real time. _avg; A dynamic preset threshold P_th is set, the value of which is determined by a joint query of the formation classification database based on the formation type and the segment material strength: Silty clay (moisture content > 20%): P_th = 0.12 MPa Loess-like strata (dry density ≥ 1.6 g / cm³): P_th = 0.10 MPa For tunnel segment concrete strength grade C50 and above: P_th can be increased to 0.15 MPa. If p_max > P_th and If _avg > 0.08 MPa / m, it is determined to be a "local pressure over-limit event" and the attitude compensation process is triggered.

[0083] Micro-displacement and rotation angle solution model Furthermore, in this embodiment, the assembly attitude adaptive adjustment model adopts a linear compensation mechanism based on physical response, and its attitude fine-tuning command is calculated and generated by the following six-degree-of-freedom control equations: ; in: The three-dimensional translation compensation amount of the tunnel segment block in the global coordinate system (unit: millimeters). The three-dimensional rotation compensation angle of the tube segment about the spatial coordinate axis (unit: milliradians, mrad). This is the measured pressure distribution vector acquired by the pressure sensor array for the temporary support of the tunnel lining segments. n≥4 represents the number of piezoelectric sensing units on the contact surface. Its value is collected in real time by an array of sensors embedded in the contact surface of the vacuum chuck or mechanical clamping arm, with a sampling frequency ≥100 Hz. This is a dynamic pressure threshold vector, where each component Pth,i is dynamically set by the formation-material joint database based on the spatial location of sensor i, the strength grade of the segment material, and the formation type. The value range is 0.10~0.15MPa, and it is automatically increased by 10%~15% in high water pressure areas. This is a six-DOF pressure-displacement compensation gain matrix, where each row corresponds to one displacement / rotation degree of freedom, and each column corresponds to the pressure sensor input. Its elements are obtained through a dual mechanism of offline calibration and online learning. Offline calibration phase: In a laboratory environment, a known pressure distribution (simulated by a hydraulic loading system) is applied to a standard six-section tube segment. A high-precision laser tracker records the corresponding six-degree-of-freedom displacement response, establishing a pressure-displacement response library. The initial pressure is then solved using the least squares method. ; Online learning phase: During construction, the system records the actual arrival error Eresidual = Pfinal − Pth after each compensation execution, and updates Kcomp using the Recursive Least Squares (RLS) algorithm. ; in To learn the gain matrix, the convergence factor λ = 0.98 is used to ensure that adaptive calibration is completed after 3 to 5 compensations. Physical constraint embedding: Off-diagonal terms in matrix Kcomp (such as the effect of Δx on Δβ) are forced to zero to ensure decoupling of each degree of freedom and avoid coupling disturbances; at the same time, the maximum output of rotation compensation terms Δα, Δβ, Δγ is limited to ±1.5 mrad to prevent over-adjustment from causing bolt hole misalignment.

[0084] The execution process of this compensation mechanism is as follows: When any sensor pressure At this time, the system triggers attitude compensation; Collect all current sensor pressure data and construct ; Calculate the deviation vector Pcontact–Pth; The six-degree-of-freedom compensation command is calculated using Kcomp. The command is sent to the three-degree-of-freedom hydraulic servo valve of the segment assembly machine with a millisecond delay, driving the fine-tuning mechanism to perform micro-displacement and micro-rotation; Within 50 ms after compensation, the system resamples the pressure. If the deviation is still > 0.02 MPa, a second compensation is initiated, with a maximum of 3 iterations. If the limit is still exceeded after 3 compensations, the system will automatically suspend assembly, trigger "partial depressurization pretreatment", and record the abnormal event.

[0085] Verification was conducted in 12 engineering sections (burial depth 8-15 m, silty clay and loess strata): The compensation response time is ≤ 120 ms (meeting a 150 ms closed-loop cycle). The pressure fluctuation of the temporary support for the tunnel segments decreased from ±0.08 MPa to ±0.015 MPa; After assembly, the misalignment between adjacent segments decreased from an average of 2.1 mm to 0.7 mm, meeting the requirement of ≤1.0 mm; The number of "segment jamming" incidents caused by sudden pressure changes decreased by 92%; The compensation gain matrix Kcomp converges stably in the third loop of construction and does not require manual recalibration in the subsequent 500 loops.

[0086] Furthermore, in this embodiment, the execution and feedback module includes an industrial-grade PLC controller, which receives assembly sequence instructions and attitude adjustment instructions from the data processing and collaborative control module through a redundant Ethernet interface, and drives the hydraulic servo system and rotation and translation mechanism of the segment assembly machine to perform actions through digital / analog output channels. At the same time, it transmits the contact pressure data of the segment temporary support pressure sensor group in real time through a high-speed acquisition module, forming an end-to-end closed-loop control loop with a control cycle ≤150 ms and an instruction response delay ≤30 ms.

[0087] In this embodiment, the execution and feedback module uses a Siemens S7-1500 series or equivalent industrial PLC, running a real-time operating system (RTOS). Its control logic includes the following four sub-processes:

[0088] Command reception and parsing The PLC controller receives assembly instruction packets from the collaborative control module (deployed in an embedded industrial computer) via the Modbus TCP protocol (port 502). The instruction packet structure is as follows: { "CmdType": "Sequence" or "PoseAdjust", "Timestamp": UNIX timestamp "TargetBlock": B3, "SequenceOrder": [B1, B5, B3, B6, B2, B4], / / Assembly order "Compensation": { "dx": 0.4 mm, "dy": -0.2 mm, "dz": 0.1 mm, "dα": 0.3 mrad, "dβ": -0.1 mrad, "dγ": 0.0 mrad }, "CRC32": 0xABCDE123 } After receiving an instruction, the PLC performs a CRC check. If the check fails, the instruction is discarded and an error is reported. If the check passes, the execution queue is triggered, and the instruction processing delay is ≤10 ms.

[0089] Execution drive and motion control If it is an "assembly sequence instruction", the PLC will enable the vacuum suction cup solenoid valve, rotary servo motor and translation hydraulic cylinder of the assembly machine in sequence according to the instruction sequence, and drive the tube block to move along the specified path. If it is an "attitude adjustment command", the PLC calculates the opening command of each hydraulic servo valve (X / Y / Z three directions) according to the compensation vector (Δx, Δy, Δz, Δα, Δβ, Δγ), and drives the fine-tuning mechanism to perform compensation action through the PID control loop (sampling period 10 ms); All actions are equipped with encoder feedback. When the position error exceeds the limit (>0.5 mm), an emergency stop is automatically triggered and the issue is reported to the collaborative control module.

[0090] Pressure data acquisition and closed-loop transmission The PLC synchronously acquires 8–16 channels of analog signals from the temporary support pressure sensor group of the tunnel segment through a built-in 16-bit high-precision AD module (sampling rate ≥100 Hz), and packages them into pressure data frames after filtering. { "SensorID": [S1, S2, ..., S16], "Pressure": [0.102, 0.118, ..., 0.095] MPa, "MaxPressure": 0.118 MPa "Timestamp": UNIX timestamp "Status": "OK" / "OverLimit" / "Fault" } In each control cycle (≤150 ms), the data frame is actively pushed back to the collaborative control module via Modbus TCP, forming a closed loop of "perception-decision-execution-feedback". The feedback data serves as the input for the next round of optimization, and the closed loop delay is ≤150ms (including communication, processing and execution).

[0091] Fault tolerance and security mechanisms The PLC has a built-in watchdog timer that automatically resets and sends a "system fault" signal when the system malfunctions. When the pressure feedback value exceeds the threshold (>0.15 MPa) for three consecutive cycles and fails to converge, the PLC automatically locks the actuator and triggers the "pause assembly" command, while simultaneously illuminating the on-site audible and visual alarm. All instructions and feedback data are written to a local industrial SSD, supporting power failure protection and post-event rollback. The PLC and the collaborative control module use dual-channel redundant communication. When the main channel fails, it automatically switches to the backup CAN bus channel (baud rate 500 kbps).

[0092] Field tests showed that after 120 hours of continuous operation in silty clay soil at a depth of 12 m, the PLC control closed-loop success rate was ≥99.2%, the command error rate was <0.01%, and the pressure feedback data loss rate was <0.005%, meeting the high reliability control requirements for underground engineering.

[0093] This execution and feedback module is the first to deeply couple the two subsystems of "dynamic earth pressure balance" and "intelligent assembly of tunnel segments" with the physical and logical layers through a standardized industrial control protocol (Modbus TCP) and a real-time closed-loop PLC architecture. This completely eliminates the systemic defects in traditional systems such as the disconnect between "upper computer decision-making and lower computer execution", data loss, and response lag. It is the core execution center for realizing "fully automated, millisecond-level response, and highly reliable closed-loop" intelligent assembly of underground utility tunnels.

[0094] Furthermore, in this embodiment, the system is specifically designed for underground pipe gallery excavation at depths of 8-15 meters in silty clay and loess strata using earth pressure balance pipe jacking machines or tunnel boring machines, and is compatible with six- or seven-segment annular precast concrete tunnel segment structures. Its applicability is based on the following triple technical coupling mechanism: Matching formation characteristics with pressure sensor deployment density Silty clay and loess-like strata have high water content (>20%) and low permeability (k<10). -6 With its high pressure density (cm / s) and strong rheological characteristics, the surrounding rock pressure distribution behind the excavation face exhibits a significant non-uniform circumferential stress gradient (maximum gradient reaching 0.15 MPa / m) within a burial depth range of 8–15 m. Furthermore, pressure fluctuations in the soil chamber tend to propagate rapidly along the rear of the shield. To accurately capture the spatial heterogeneity of pressure in this type of stratum, this system requires that the surrounding rock contact pressure sensor group be uniformly distributed with 8–12 measuring points along the outer circumference of the shield tail (as described in claim 2), and that the sampling frequency be ≥50 Hz to cover the stress variation range of every 30°–45°. If applied to sand or rock strata (with uniform pressure distribution), 8 sensors would suffice. However, in this scenario, fewer than 8 measuring points would result in a pressure field reconstruction error >15%, leading to misjudgments in the assembly strategy. Therefore, this sensor density is a necessary condition for the system to achieve effective control in this stratum.

[0095] Coupling of burial depth range with pressure threshold and control cycle When the burial depth is <8 m, the surface load has a significant impact, and the surrounding rock pressure fluctuates violently and frequently (>0.5 Hz). The system needs to use a higher sampling frequency (>30 Hz) and a shorter control cycle (<100 ms) to cope with transient disturbances. When the burial depth is >15 m, the surrounding rock stress tends to stabilize, but the soil chamber pressure needs to be set at a higher value (>0.25 MPa), which greatly increases the requirements for the pressure regulation capability of the propulsion cylinder, exceeding the safe operating range of the hydraulic system adapted to this system (rated pressure ≤35 MPa). The pressure pretreatment decompression amount (0.05–0.10 MPa), temporary support pressure threshold (0.10–0.15 MPa), and control cycle ≤150 ms adopted by this system are all calibrated based on the measured pressure time-domain characteristics in the burial depth range of 8~15 m (fitted from 200 sets of field data). If the range is exceeded, the assembly stability and safety cannot be guaranteed. Therefore, the burial depth range is a hard boundary constraint for the system parameter design.

[0096] Adaptation of the segment structure form and assembly sequence optimization model Six-segment (one base, two waists, two tops, one center) and seven-segment (one base, two waists, two tops, two adjacent) annular precast concrete tunnel segments have asymmetrical geometry and mechanical characteristics with bolt connections concentrated in the circumferential middle section. Under non-uniform pressure distribution, if eight or more segments are used (such as the 10-segment type commonly used in subway tunnels), the selection space for the reference block is too large, which can easily lead to the assembly sequence optimization model getting trapped in local optima. If five segments are used (traditional pipe jacking structure), the closure angle error accumulates quickly, making it difficult to meet the geometric constraint of ≤0.5°. In this system, the constraints of the dynamic optimization model for the assembly sequence (such as the reference block must be selected from the low-pressure zone, and the angle between the insertion direction and the pressure gradient ≤30°) and the displacement sensitivity coefficient of the attitude compensation model (limited by the neural network training dataset) are all trained based on the geometric model of the six-segment / seven-segment tunnel segments and finite element stress distribution simulation (ABAQUS modeling, concrete C50, elastic modulus 35 GPa), and are only applicable to this type of structure. If applied to other structures (such as steel segments or irregularly shaped segments), the pressure-coordinate mapping will be inaccurate and the compensation vector output will fail.

[0097] Verified in 15 typical engineering sections, under the conditions of silty clay, burial depth of 10–13 m, and seven-piece tunnel segments, this system achieves the entire process of tunnel segment assembly without human intervention, with an automation rate of ≥96.5%, tunnel segment ellipticity ≤1.8 mm, and adjacent misalignment ≤0.8 mm. If forcibly applied to rock strata with a burial depth of 20 m or eight-piece tunnel segments, the assembly failure rate increases to 14.2%, and the frequency of pressure exceeding limits increases by 3.1 times.

[0098] This system is not a general-purpose assembly control system, but a collaborative control system specifically designed for the combination of "soft cohesive strata + shallow to medium burial depth + six / seven-section tunnel segments". All its technical features - sensor density, pressure threshold, algorithm parameters, control cycle, compensation strategy - are collaboratively designed around the physical constraints of this combination, forming an inseparable and irreplaceable technical whole, which solves the three fundamental defects of traditional systems caused by "one-size-fits-all" approach: "no strata sensing, structural mismatch, and no control response".

[0099] Furthermore, in this embodiment, during the quality inspection after the tunnel segments are assembled, the system achieves a segment ellipticity deviation of ≤2.0 mm and an adjacent segment misalignment of ≤1.0 mm, and the entire assembly process requires no manual intervention, with an automation level of ≥95%. This quality indicator is necessarily achieved through the collaborative action of the pressure distribution real-time monitoring module, the spatial coordinate positioning module, the data processing and collaborative control module, and the execution and feedback module. The technical implementation path is as follows: Mechanism for achieving ellipticity deviation ≤ 2.0 mm The ellipticity of the segment ring is defined as the difference between the maximum and minimum diameters within the ring. Traditional systems, due to their fixed assembly sequence and delayed attitude adjustment, often experience localized stress concentration during ring closure, leading to "flattening" or "bulging," with ellipticity typically reaching 3.5–5.0 mm.

[0100] This system achieves high-precision ring formation through the following mechanism: Pressure-coordinate coupled field drives precise positioning of the reference block: The assembly sequence dynamic optimization model prioritizes the installation of the reference block (B1) in the low pressure zone (pressure gradient < 0.05 MPa / m), and ensures that the angle between its normal direction and the local principal stress direction is ≤ 15°, so as to ensure the initial positioning force balance; Adaptive attitude compensation suppresses cumulative error: After each segment is in place, the assembly attitude adaptive adjustment model automatically calculates and performs micro-displacement (Δx, Δy, Δz ≤ 2.0 mm) and micro-rotation (Δα, Δβ, Δγ ≤ 0.5 mrad) compensation based on the local stress difference fed back by temporary support pressure sensors (≥4 points / contact surface), ensuring that the positioning error of each segment before bolt tightening is ≤0.3 mm; Closed-loop control suppresses circumferential accumulation: After each ring is assembled, the system automatically collects the spatial coordinates of all segments of that ring and calculates the actual ellipticity; if it is >1.5 mm, the "intra-ring fine-tuning mode" is activated, and prestress compensation is applied to the ring structure by adjusting the pressure difference (±0.05 MPa) between adjacent propulsion cylinders until the ellipticity is ≤2.0 mm.

[0101] In actual measurements across 15 construction sections, this mechanism resulted in an average segment ring ellipticity of 1.6 mm and a maximum of 1.9 mm, with over 98% of rings meeting the requirement of ≤2.0 mm.

[0102] Mechanism for achieving adjacent segment misalignment ≤ 1.0 mm Misalignment refers to the circumferential or axial displacement between adjacent segments. Traditional methods, lacking real-time pressure feedback, often result in misalignment exceeding the limit (2.0–4.0 mm) within 2–3 hours after assembly due to the release of surrounding rock stress.

[0103] This system achieves stable control through the following mechanism: Temporary support pressure closed-loop control: The vacuum suction cup / clamping arm of the segment assembly machine is equipped with an array of pressure sensors (≥4 units / contact surface) to monitor the local stress on the segment contact surface in real time; when the pressure at any measuring point is >0.15 MPa, the system immediately triggers attitude compensation to eliminate "warping" or "deflection" caused by uneven pressure. Assembly sequence avoids stress abrupt change zone: When planning the path, the dynamic optimization model of assembly sequence prohibits placing adjacent segments (such as B2 and B3) in high pressure gradient zone (such as pressure gradient > 0.12 MPa / m) at the same time, so as to avoid local deformation caused by "two blocks being under pressure at the same time"; Bolt tightening pressure stability judgment: The system will only allow the bolt tightening procedure to be started if the pressure fluctuation of the temporary support of the tunnel segment is <±5% and the position error is stable at ≤0.3 mm within 3 consecutive control cycles (≤450 ms).

[0104] The measured data shows that the average misalignment between adjacent segments in this system is 0.7 mm, the maximum value is 0.9 mm, and 100% of the misalignment meets the requirement of ≤1.0 mm.

[0105] Implementation mechanism with an automation level of ≥95% The degree of automation is defined as the percentage of assembled blocks that automatically complete the entire process of "positioning, handling, placement, posture adjustment, fastening, and feedback" without human intervention.

[0106] This system ensures a high degree of automation through the following architecture: End-to-end autonomous decision-making chain without human intervention: from pressure perception → field reconstruction → sequence optimization → attitude calculation → PLC execution → pressure feedback → model re-optimization, forming an end-to-end autonomous decision-making chain with no human intervention points; Automatic anomaly handling: When anomalies such as soil chamber pressure fluctuation > ±15%, sensor failure, or positioning loss occur, the system automatically pauses assembly and starts the "pressure stabilization pretreatment process" (adjusting oil cylinders, micro-grouting, and buffer zone activation). After recovery, the assembly process is automatically resumed without manual reset. The operation interface is limited to confirmation: the operator only needs to confirm the "start command" via the touch screen before each assembly (the system automatically records the operator's identity and timestamp), and the rest of the process is completely autonomous.

[0107] During the 120-hour continuous operation test, a total of 187 rings were assembled, with a total of 1122 segments. There were only 5 manual interventions (all for equipment maintenance), and the automation rate was (1122–5) / 1122 = 99.55%.

[0108] In summary, the "ellipticity ≤ 2.0 mm, misalignment ≤ 1.0 mm, and automation rate ≥ 95%" achieved by this system are not accidental performance results, but rather the inevitable technical effect derived from the synergistic effect of pressure-coordinate dynamic coupling modeling, multi-source sensing closed-loop feedback, adaptive optimization algorithm, and industrial-grade PLC execution architecture. Its accuracy and automation level are significantly better than traditional systems (ellipticity ≥ 3.5 mm, misalignment ≥ 2.0 mm, automation rate < 70%), providing a quantifiable, verifiable, and replicable technical paradigm for high-precision, fully autonomous construction of underground utility tunnels.

[0109] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those described herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0110] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A coordinated control system for earth pressure balance and segment assembly during underground utility tunnel excavation, characterized in that, include: The real-time pressure distribution monitoring module is used to collect three-dimensional spatial distribution data of soil chamber pressure, surrounding rock contact pressure, and temporary support pressure of tunnel segments; The spatial coordinate positioning module is used to acquire the three-dimensional spatial coordinates and six-degree-of-freedom pose of the tube segments to be assembled in real time. The data processing and collaborative control module is used to fuse and map the pressure distribution data with the spatial coordinate data, construct a pressure-coordinate coupled field for the area to be assembled, and dynamically optimize the segment assembly sequence and assembly posture based on the coupled field. The execution and feedback module is used to send the optimized assembly instructions to the segment assembly machine for execution, and to provide feedback on the real-time contact pressure through the pressure sensor to form a closed-loop control. The data processing and collaborative control module includes: a three-dimensional pressure field reconstruction model, a pressure-coordinate mapping model, a dynamic optimization model for assembly sequence, and an adaptive adjustment model for assembly posture.

2. The collaborative control system according to claim 1, characterized in that, The real-time pressure distribution monitoring module includes: At least three soil pressure sensors are installed at different heights on the soil chamber partition to obtain the vertical pressure gradient inside the soil chamber. Eight to twelve surrounding rock contact pressure sensors are evenly distributed along the circumference of the shield at the tail of the shield body to collect the circumferential contact pressure distribution of the exposed surrounding rock section. Multiple temporary support pressure sensors embedded in the contact surface of the vacuum suction cup or mechanical clamping arm of the segment assembly machine are used to monitor the contact pressure of the segment blocks during the temporary fixing process.

3. The collaborative control system according to claim 1, characterized in that, The spatial coordinate positioning module includes: The laser tracker is mounted on the assembled stable segment ring. It establishes an absolute coordinate system through at least three reference targets and tracks the reflective target ball installed at the end of the segment assembly machine in real time to obtain the spatial three-dimensional coordinates of the segment block. The inertial navigation measurement unit is installed on the rotary frame and translation frame of the segment assembly machine and is used to collect high-frequency attitude angle data; A Kalman filter is used to fuse low-frequency absolute coordinate data from the laser tracker with high-frequency attitude data from the inertial navigation measurement unit to achieve real-time high-precision calculation of the six-degree-of-freedom pose of the tube segment, with a positioning accuracy of no more than 0.5 mm.

4. The collaborative control system according to claim 1, characterized in that, The pressure field three-dimensional reconstruction model uses Kriging interpolation or inverse distance weighting to spatially interpolate and reconstruct the soil chamber pressure gradient data and the surrounding rock circumferential contact pressure data, generating a three-dimensional distribution field of surrounding rock pressure within a range of 3 to 5 rings behind the excavation face, with a spatial resolution of not less than 0.1 meters.

5. The collaborative control system according to claim 1, characterized in that, The pressure-coordinate mapping model spatially aligns the reconstructed three-dimensional distribution field of surrounding rock pressure with the spatial coordinates of the segment ring to be assembled, generating a "pressure distribution map of the area to be assembled", and marking high-pressure areas, low-pressure areas and pressure gradient abrupt change areas as input for optimizing the assembly sequence.

6. The collaborative control system according to claim 1, characterized in that, The dynamic optimization model for the assembly sequence employs an improved ant colony algorithm. The objective function is to minimize the weighted value of the pressure disturbance experienced by each segment in the segment ring during assembly. The boundary condition is the geometric closure constraint of the segment ring. The model dynamically generates the optimal assembly sequence, and its strategy includes: Prioritize assembling the reference blocks in low-pressure areas; Align the insertion direction of the tube segment with the direction of the local pressure gradient; For high-pressure pipe segments, local pressure relief pretreatment is carried out by adjusting the pressure difference between adjacent propulsion cylinders before assembly.

7. The collaborative control system according to claim 1, characterized in that, When the assembly posture adaptive adjustment model detects that the contact pressure fed back by the temporary support pressure sensor of the segment exceeds the preset threshold, it automatically calculates the micro-displacement and rotation angle that the segment block needs to be adjusted, and performs posture compensation through the micro-adjustment mechanism of the segment assembly machine to ensure the positioning accuracy of the segment before the bolts are tightened.

8. The collaborative control system according to claim 1, characterized in that, The execution and feedback module includes a PLC controller, which receives assembly sequence instructions and attitude adjustment instructions output by the collaborative control module, drives the segment assembly machine to perform actions, and transmits temporary support pressure data of the segments back in real time, forming a closed-loop control loop of "perception-decision-execution-feedback".

9. The collaborative control system according to claim 1, characterized in that, The system is suitable for underground pipe gallery excavation at a depth of 8-15 meters in silty clay and loess strata using earth pressure balance pipe jacking machines or shield machines. The segment structure consists of six or seven ring-shaped precast concrete segments.

10. The cooperative control system according to any one of claims 1 to 9, characterized in that, In the quality inspection after the tunnel segments are assembled, the system achieves a tunnel segment ellipticity deviation of ≤2.0 mm and an adjacent segment misalignment of ≤1.0 mm, and the entire assembly process requires no manual intervention, with an automation level of over 95%.

Images