Prefabricated pipe gallery joint waterproof construction method for U-shaped shield integrated construction

Abstract

This invention discloses a waterproofing construction method for prefabricated pipe gallery joints used in the integrated construction of U-shaped shields, relating to the field of waterproofing construction technology for prefabricated components in underground engineering. Existing technologies employ multi-cylinder hydraulic systems for synchronous jacking, but the alternating soft and hard underground soil layers cause abrupt changes in the friction coefficient, leading to severe distortion in thrust estimation. Furthermore, during pipe gallery movement, attitude imbalances can cause the sealing gasket on the deflected side to be instantly crushed, while the side furthest from the deflection side experiences insufficient compression. This construction method includes: installing the sealing gasket within the sealing groove of the first pipe gallery segment; collecting multi-source state data of the real-time jacking action of the U-shaped shield; calculating the target reward value and target learning rate of the current jacking action, updating the PPO network accordingly, and outputting the optimal sidewall friction coefficient and optimal bottom friction coefficient; calculating the target jacking force of any cylinder; adjusting the target jacking force of any cylinder to generate the target jacking force of the next jacking action cylinder and executing it. It can adapt to geological changes and dynamically correct deviations.

Description

Technical Field

[0001] This invention relates to the field of waterproof construction technology for prefabricated components in underground engineering, and specifically to a waterproof construction method for prefabricated pipe gallery joints used in the integrated construction of U-shaped shields. Background Technology

[0002] With the large-scale construction of urban underground utility tunnels, the U-shaped shield integrated construction technology has been widely used due to its high efficiency and low disturbance. This process uses a multi-cylinder hydraulic system for synchronous jacking, achieving "dynamic jacking and one-time forming" assembly of tunnel segments. However, this process causes the joint sealing gasket to be subjected to instantaneous compression during assembly, which may result in eccentric loading, and secondary adjustments are not possible, placing stringent requirements on the reliability of the sealing gasket.

[0003] Currently, waterproofing of precast pipe gallery joints largely draws on shield tunneling technology, with commonly used sealing gasket cross-sections including rectangular, multi-lip, wedge, and ordinary horseshoe shapes. However, these traditional cross-sections all have significant defects under the dynamic jacking conditions of U-shaped shields: rectangular cross-sections lack guidance and are prone to stress concentration; multi-lip cross-sections are prone to bending and damage at the lip and are sensitive to compression; wedge cross-sections are prone to slippage of the sealing line under eccentric loading; and although ordinary horseshoe cross-sections have cavity buffers, stress concentration still easily forms at the cavity edge under eccentric loading, and their guiding capacity is limited.

[0004] In actual utility tunnel assembly scenarios, the reliability of prefabricated utility tunnel joint sealing and waterproofing is highly dependent on the precise control of the jacking force. In actual U-shaped shield dynamic jacking conditions, thrust control faces complex disturbances from both macroscopic geology and microscopic structure. On the one hand, the alternating softness and hardness of underground soil layers can easily lead to a precipitous change in the friction coefficient. Existing thrust prediction models based on reinforcement learning often fail to activate their exploration capabilities in time due to the solidification of historical learning rates, resulting in serious distortion of thrust estimation. On the other hand, the movement of the utility tunnel inevitably generates attitude eccentricity, and the sealing gasket has strong nonlinear stress hardening characteristics after the internal cavity is closed. Existing open-loop evenly distributed thrust strategies cannot cope with this complex coupled disturbance, making it difficult to issue precise single-cylinder compensation thrust in real time, and even more difficult to absorb the mechanical lag at the bottom layer. This can easily lead to the sealing gasket on the deflection side being crushed instantly or insufficient compression on the side away from it.

[0005] Therefore, there is an urgent need for a new type of sealing gasket structure and supporting construction method that can adapt to the characteristics of U-shaped shield frame technology and ensure uniform compression and reliable sealing during dynamic one-time molding. Summary of the Invention

[0006] To address the aforementioned problems, this invention aims to provide a prefabricated pipe gallery joint waterproofing construction method for integrated construction of U-shaped shields, which can adapt to the characteristics of U-shaped shield technology and ensure uniform compression and reliable sealing during dynamic one-time molding.

[0007] The main ideas of the technical solution adopted in this invention are as follows: First, a geological abrupt change intensity index with nonlinear saturation constraints is constructed to adaptively linearly compensate the PPO network, giving the system the ability to quickly break out of the deadlock, thereby reversing the unknown geological friction noise. Second, at the thrust space decoupling level, the resistance is scientifically distributed to each cylinder, and a lateral eccentric mapping mechanism is introduced. By deeply combining the viscoelastic constitutive model of the sealing gasket, an adaptive off-center load compensation amount is constructed for cylinders under different stress gradients, which can counteract local nonlinear stress hardening resistance and achieve active attitude correction in milliseconds. Finally, in order to suppress the hysteresis of the underlying mechanical execution, a fuzzy PID closed loop based on the measured contact stress deviation is constructed to ensure the high-quality forming and assembly of the prefabricated pipe gallery joints.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: Install the sealing gasket in the sealing groove of the first pipe rack segment; Calculate the target thrust of each cylinder of the U-shaped shield; Each cylinder of the U-shaped shield performs the target jacking force to push the second pipe gallery segment toward the first pipe gallery segment, compressing the sealing gasket to achieve waterproof treatment of the prefabricated pipe gallery joints. The calculation of the target thrust of each cylinder of the U-shaped shield includes: The observation of thrust, attitude parameters, soil stress parameters, and sealing gasket mechanical parameters under the real-time jacking action of the U-shaped shield were collected. Based on the observed thrust and soil stress parameters under the current jacking operation of the U-shaped shield and several historical jacking operations, a multi-dimensional time series feature matrix is ​​constructed. The multi-dimensional time series feature matrix is ​​input into the PPO network to output the estimated sidewall friction coefficient and the estimated bottom friction coefficient. The theoretical thrust is calculated by combining the estimated sidewall friction coefficient, the estimated bottom friction coefficient, and the soil stress parameters under the current jacking action of the U-shaped shield. The target bonus value is calculated based on the deviation between the theoretical thrust and the observed thrust under the current jacking action of the U-shaped shield. Calculate the intensity of geological environment change under the current jacking action of the U-shaped shield, use the intensity of geological environment change to perform gain compensation on the base learning rate of the PPO network to obtain the target learning rate; update the PPO network by combining the target reward value and the target learning rate, and output the optimal sidewall friction coefficient and the optimal bottom friction coefficient. For any cylinder of the U-shaped shield, the basic friction force of the cylinder is calculated based on the optimal sidewall friction coefficient, the optimal bottom friction coefficient, and the soil force parameters under the current jacking operation of the U-shaped shield; the adaptive off-center load compensation amount of the cylinder is calculated based on the attitude parameters; and the target jacking force of the cylinder is calculated by combining the adaptive off-center load compensation amount, the basic friction force, and the mechanical parameters of the sealing gasket under the current jacking operation of the U-shaped shield.

[0009] Furthermore, the attitude parameter is the attitude deflection angle of the U-shaped shield during its current pushing action; The soil stress parameters include the soil resistance in front of the shield, the lateral active earth pressure, and the current contact length between the U-shaped shield and the lateral soil. The mechanical parameters of the sealing gasket include its nonlinear elastic modulus, bearing area, equivalent stiffness coefficient, initial thickness, damping coefficient, compressive strain, real-time compression amount, compression rate, measured contact stress, and normalized transverse contact stress difference.

[0010] Furthermore, a multi-dimensional time-series feature matrix is ​​constructed based on the observed thrust and soil stress parameters under the current jacking action of the U-shaped shield and several historical jacking actions. This multi-dimensional time-series feature matrix is ​​input into the PPO network, outputting estimated sidewall friction coefficients and estimated bottom friction coefficients, including: Based on the observed thrust and soil stress parameters under the current jacking operation of the U-shaped shield and several historical jacking operations, a multi-dimensional time series feature matrix is ​​constructed. The multidimensional temporal feature matrix is ​​input into the sequence feature extraction layer of the PPO network, and the extracted temporal feature vector is mapped to the probability distribution parameters of the continuous action space through the fully connected layer of the PPO network. The physical boundary constraint layer of the PPO network generates estimated sidewall friction coefficients and estimated bottom friction coefficients based on probability distribution parameters.

[0011] Furthermore, the theoretical thrust is calculated by combining the estimated sidewall friction coefficient, the estimated bottom friction coefficient, and the soil stress parameters under the current jacking action of the U-shaped shield. The target bonus value is then calculated based on the deviation between the theoretical thrust and the observed thrust under the current jacking action of the U-shaped shield, including: The estimated sidewall friction force is obtained by calculating the product of the lateral active earth pressure, the current contact length between the U-shaped shield and the lateral soil, the estimated sidewall friction coefficient, and the number of sidewalls. The estimated bottom friction force is obtained by calculating the product of the bottom friction coefficient and the weight of the U-shaped shield. The theoretical thrust is obtained by summing the soil resistance in front of the shield, the estimated sidewall friction, and the estimated bottom friction. Calculate the absolute value of the difference between the observed thrust and the theoretical thrust under the current jacking action of the U-shaped shield, divide this absolute value by the observed thrust under the current jacking action of the U-shaped shield, and take the negative of the quotient to obtain the target reward value.

[0012] Furthermore, the calculation of the intensity of geological environment abrupt change under the current jacking action of the U-shaped shield includes: Calculate the standard deviation of the observed thrust under the current jacking action of the U-shaped shield and several historical jacking actions, and use it as the thrust fluctuation rate; The absolute value of the difference between the observed thrust and the theoretical thrust under the current jacking action of the U-shaped shield is calculated as the thrust deviation. Calculate the ratio of thrust deviation to thrust fluctuation rate, and take the negative of this ratio as the exponential term. Calculate the exponential function value of the exponential term with the natural constant as the base to obtain the abrupt decay term. The difference between a constant 1 and the abrupt decay term is used as the intensity of geological environmental abrupt changes.

[0013] Furthermore, the base learning rate of the PPO network is compensated for using the intensity of geological environmental abrupt changes to obtain the target learning rate; the PPO network is updated by combining the target reward value and the target learning rate to output the optimal sidewall friction coefficient and the optimal bottom friction coefficient, including: Calculate the target learning rate using the following formula: ; In the formula, This represents the target learning rate for the current top-level action; Indicates the base learning rate; This indicates the severity of sudden changes in the geological environment under the current top-down action; Indicates the gain coefficient; The PPO network uses the target learning rate as the step size and combines the target reward value to perform gradient descent and PPO network weight updates. After convergence and stabilization, it outputs the optimal sidewall friction coefficient and the optimal bottom friction coefficient.

[0014] Furthermore, for any cylinder of the U-shaped shield, the basic friction force of the cylinder is calculated based on the optimal sidewall friction coefficient, the optimal bottom friction coefficient, and the soil stress parameters; including: Extract the effective control area of ​​any cylinder of the U-shaped shield frame corresponding to the side wall group or bottom group and divide it by the total area of ​​the corresponding group to obtain the force weight of the side wall and bottom of the cylinder respectively. The average frontal resistance is calculated by dividing the soil resistance in front of the shield by the total number of hydraulic cylinders on the U-shaped shield. The optimal sidewall friction coefficient, lateral active earth pressure, current contact length between the U-shaped shield and the lateral soil, and the product of the number of sidewalls are calculated as the sidewall friction force, and multiplied by the sidewall force weight as the distributed sidewall friction force. The product of the optimal bottom friction coefficient and the self-weight of the U-shaped shield is calculated as the bottom friction force, and multiplied by the bottom force weight as the distributed bottom friction force. The basic friction force of the hydraulic cylinder is obtained by summing the average frontal resistance, the distributed sidewall friction force, and the distributed bottom friction force.

[0015] Furthermore, the expression for the adaptive off-center load compensation amount is: ; In the formula, Indicates the current top-level push action. The adaptive off-center load compensation amount for each hydraulic cylinder; This represents the normalized lateral contact stress difference of the current jacking action of the U-shaped shield. This indicates the attitude deflection angle of the U-shaped shield during its current pushing action; This indicates the current jacking action of the U-shaped shield. The lateral eccentricity ratio of each cylinder; This represents the thrust conversion factor for eccentric load compensation. Represents a symbolic function; Represents the natural logarithm function; Represents the sine function; This indicates taking the absolute value.

[0016] Furthermore, the expression for the target top thrust is: ; In the formula, This indicates the current jacking action of the U-shaped shield. The target thrust of each hydraulic cylinder; Represents the nonlinear elastic modulus; Indicates the pressure-bearing area; This indicates the current jacking action of the U-shaped shield. The compressive strain of each hydraulic cylinder; Indicates the equivalent stiffness coefficient; This indicates the current jacking action of the U-shaped shield. Real-time compression of the sealing gasket corresponding to each hydraulic cylinder; Indicates the damping coefficient; This indicates the current jacking action of the U-shaped shield. The compression rate corresponding to each hydraulic cylinder; This indicates the current jacking action of the U-shaped shield. The basic friction force of each oil cylinder.

[0017] Furthermore, the hydraulic cylinders of the U-shaped shield execute the target jacking force to advance the second tunnel segment toward the first tunnel segment, including: Set the target contact stress for the sealing gasket; The deviation between the target contact stress and the measured contact stress, as well as the rate of change of this deviation, are calculated and input into the fuzzy controller. The proportional coefficient, integral coefficient, and derivative coefficient of the PID controller are then tuned. The product of the deviation between the target contact stress and the measured contact stress and the proportionality coefficient is calculated as the proportional adjustment term; The integral of the deviation between the target contact stress and the measured contact stress over time, multiplied by the integral coefficient, is used as the integral adjustment term. The product of the derivative of the deviation between the target contact stress and the measured contact stress with respect to time and the derivative coefficient is used as the differential adjustment term; The target thrust of the hydraulic cylinder is summed with the proportional, integral, and derivative adjustment terms to obtain the target thrust of each hydraulic cylinder of the U-shaped shield for the next thrust action.

[0018] The beneficial effects of this invention are: This invention first addresses the limitation of traditional PPO networks, which are prone to learning rate fixation when encountering abrupt geological changes. It constructs a geological change severity index with nonlinear saturation constraints, and applies adaptive linear gain compensation to the PPO network, enabling the system to quickly escape stalemates and thus revealing unknown geological friction noise. Second, at the thrust space decoupling level, the resistance is scientifically distributed to each cylinder, and a lateral eccentric mapping mechanism is introduced. This is combined with the viscoelastic constitutive model of the sealing gasket to construct adaptive off-center load compensation for cylinders under different stress gradients, counteracting local nonlinear stress hardening resistance and achieving active attitude correction in milliseconds. Finally, to suppress the hysteresis of underlying mechanical execution, a fuzzy PID closed loop based on measured contact stress deviation is constructed, ensuring high-quality forming and assembly of prefabricated pipe gallery joints. Attached Figure Description

[0019] Figure 1 A schematic diagram of the cross-sectional structure of the double-eared horseshoe-shaped sealing gasket provided by the present invention; Figure 2 for Figure 1 The stress cloud diagram of the sealing gasket under a 5mm compression state is shown in the finite element simulation. Figure 3 for Figure 1 The stress cloud diagram of the sealing gasket under a 10mm compression state is shown in the finite element simulation. Figure 4 for Figure 1 The stress contour plot of the sealing gasket under a 15mm compression condition is shown in the finite element simulation. Figure 5 for Figure 1 The stress contour plot of the sealing gasket under a 20mm compression state is shown in the finite element simulation. Figure 6 for Figure 1 The stress cloud diagram of the sealing gasket under a 25mm compression state is shown in the finite element simulation. Figure 7 This diagram illustrates the assembly and construction process of a pipe gallery joint using the sealing gasket of this invention. Figure 1 ; Figure 8 This diagram illustrates the assembly and construction process of a pipe gallery joint using the sealing gasket of this invention. Figure 2 ; Figure 9This diagram illustrates the assembly and construction process of a pipe gallery joint using the sealing gasket of this invention. Figure 3 ; Figure 10 This is a schematic diagram of the optimized gasket cross-sectional structure in Example 2; Figure 11 This is a schematic diagram showing the connection state of the two pipe racks when the sealing gasket described in Embodiment 2 is used for sealing; Figure 12 This is a schematic diagram of the displacement sensor measuring point arrangement for compression monitoring in the construction method of Example 3. Figure 1 ; Figure 13 This is a schematic diagram of the displacement sensor measuring point arrangement for compression monitoring in the construction method of Example 3. Figure 2 ; Figure 14 This is a structural schematic diagram of utility tunnel 1; Figure 15 This is a structural schematic diagram of the utility tunnel II; Figure 16 This is a schematic diagram of a sealing test conducted after the connection between utility tunnel 1 and utility tunnel 2. Figure 17 A flowchart of a prefabricated pipe gallery joint waterproofing construction method for integrated construction of U-shaped shields provided by the present invention; Figure 18 A schematic diagram of the force balance analysis for the integrated jacking of the U-shaped shield; Figure 19 A diagram showing the geological abrupt change and the evolution of observed thrust; Figure 20 This is a diagram showing the evolution of attitude eccentric loading and stress difference; Figure 21 A dynamic comparison and verification diagram of the target thrust force of the i-th hydraulic cylinder; The components are: 1. Sealing body; 2. Ear protrusion; 3. Through hole; 4. Fitting part; 5. Pipe gallery one; 6. Pipe gallery two; 7. Grouting pipe; 8. Conical groove; 9. Receiving part; 10. Concrete pin; 11. Sealing gasket; 12. Integrated pipe gallery. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Example 1: A prefabricated pipe gallery joint sealing gasket suitable for integrated construction of U-shaped shields.

[0022] See Figures 1-21This application discloses a prefabricated pipe gallery joint sealing gasket suitable for integrated construction of U-shaped shields. The sealing gasket 11 has a cross-sectional configuration of a double-eared horseshoe shape, which includes a horseshoe-shaped sealing body 1. Multiple through holes 3 are provided on the sealing body 1. The through holes 3 are arranged in three rows, and the number of through holes 3 from top to bottom is arranged in 2-3-3. A pair of symmetrically arranged ear protrusions 2 are provided on the upper part of the sealing body 1.

[0023] The material hardness of the ear protrusion 2 is lower than that of the sealing body 1. For example, the ear hardness is IRHD50±5, and the body hardness is IRHD57±5.

[0024] Preferably, the optimal compression range of the sealing gasket 11 is 20.0 mm to 22.5 mm (corresponding to a compression rate of 40% to 45%). Within this compression range, the average contact stress generated between the sealing gasket 11 and the concrete trench interface is 1.5 MPa to 2.2 MPa.

[0025] Preferably, the sealing gasket 11 is integrally molded from EPDM rubber.

[0026] Example 2: Further optimization of a prefabricated pipe gallery joint sealing gasket suitable for integrated construction of U-shaped shields.

[0027] like Figure 10 As shown, based on Embodiment 1, a fitting part 4 is provided below the sealing body 1. High-density rubber is used. When installing the sealing gasket 11, the fitting part 4 is snapped into the receiving part 9 opened on the integrated pipe gallery 12, so that it is not easy to loosen during installation.

[0028] Figure 11 This diagram illustrates the connection state of two integrated utility tunnels 12 when using the sealing gasket 11 described in Embodiment 2 for sealing. The fitting part 4 is secured inside the integrated utility tunnel 12 on the left side by the receiving part 9, preventing loosening during installation. The structure of this embodiment can be applied during construction to improve the stability of the sealing body 1.

[0029] Example 3: A method for waterproofing precast pipe gallery joints in U-shaped shield construction This embodiment provides a waterproofing construction method for prefabricated pipe gallery joints used in U-shaped shield construction, employing the sealing gasket 11 described in Embodiment 1. (Refer to...) Figure 12 The following diagrams, using six utility tunnels as examples, illustrate the assembled configuration of multiple utility tunnels in this embodiment: utility tunnel A, utility tunnel B, utility tunnel C, utility tunnel D, utility tunnel E, and utility tunnel F.

[0030] The method includes the following steps: S1: Install the sealing gasket 11 in the sealing groove of section 5 of the pipe gallery; S2: Using the multi-cylinder hydraulic synchronous jacking system of the U-shaped shield, the second 6th section of the pipe gallery is pushed towards the first 5th section of the pipe gallery, so that the sealing gasket 11 is compressed. S3: During the jacking process, the joint width or jacking displacement is monitored in real time by a displacement sensor, and the compression of the sealing gasket 11 is controlled within the target range of 20.0mm to 22.5mm; the displacement sensor is positioned as follows... Figure 13 ,in, Figure 13 In the diagram, a1, a2, a3, a4, a5, a6, a7, and a8 represent the layout of multiple displacement sensor monitoring points on the pipe gallery assembly surface. S4: Push the tube rack into place and maintain pressure to complete the one-time assembly of the tube rack segments and the sealing of the joints.

[0031] In the jacking process, this application adopts an adaptive calculation system for the jacking force of prefabricated pipe gallery joints, which can dynamically and accurately adjust the magnitude of the jacking force for integrated construction of U-shaped shields.

[0032] First, an adaptive calculation system for the jacking force of precast pipe gallery joints is constructed for the integrated construction of U-shaped shields. This system is used to control the multi-cylinder hydraulic synchronous jacking system of the U-shaped shields in real time. The adaptive calculation system for the jacking force of precast pipe gallery joints works in conjunction with the sealing gasket 11 to form an intelligent feedback sealing system. This system includes: a data acquisition module, a target reward value calculation module, a friction coefficient adaptive identification module, a single-cylinder jacking force decoupling calculation module, and a jacking force closed-loop adjustment module.

[0033] Reference Figure 17 Correspondingly, an adaptive calculation method for the jacking force of precast pipe gallery joints in the integrated construction of U-shaped shields is proposed, including the following steps: A1 collects multi-source status data on the real-time jacking action of the U-shaped shield.

[0034] It should be noted that the U-shaped shield integrated construction adopts a dynamic jacking and one-time molding process. During the process of multi-cylinder hydraulic synchronous propulsion of prefabricated pipe gallery segments, the pipe gallery system is in a complex dynamic mechanical equilibrium environment. First, from the perspective of external geological resistance, the resistance of the pipe gallery system depends not only on the macroscopic thrust feedback, but also on the front cutting resistance, lateral earth pressure and geological features. If these environmental boundary parameters cannot be fully collected, the mechanical thrust model cannot be accurately reconstructed. Secondly, from the perspective of system motion posture and displacement, the pipe gallery segments are prone to spatial posture deflection during movement; Meanwhile, the actual elongation of each hydraulic cylinder directly determines the compression process of the sealing gasket 11. It is necessary to monitor the attitude deflection angle and the thrust displacement of each cylinder to provide basic data for subsequent dual displacement control. Finally, from the perspective of the micro-stress structure of the joint, when the pipe gallery is subjected to a slight eccentric load, the sealing gasket 11 on the deflection side will be excessively squeezed instantly. It is necessary to penetrate into the micro-interface of the joint to capture the lateral eccentric load stress difference between the left and right sides of the pipe gallery in real time, as well as the actual contact stress of the entire cross section of the sealing gasket 11. Therefore, it is necessary to collect multi-dimensional dynamic data, including geological environment, machinery, spatial attitude, displacement and micro-stress, to provide data support for subsequent analysis of single-cylinder compensated thrust.

[0035] Specifically, the integrated sensor network at the U-shaped shield construction site is used to synchronously and automatically collect multi-source status data of the real-time jacking operation. The specific execution process is as follows: Firstly, advanced ground-penetrating radar or drilling-while-drilling sensing systems mounted in front of the U-shaped shield are used to acquire soil stress parameters during the current jacking operation in real time. Specifically, earth pressure sensors arranged in front of and on the side walls of the shield shell are used to acquire the soil resistance and lateral active earth pressure in front of the shield during the current jacking operation, and the stroke sensors of the multi-cylinder hydraulic synchronous jacking system read the current contact length between the U-shaped shield and the lateral soil (i.e., the current shield contact length).

[0036] Secondly, high-precision dual-axis tilt sensors or inertial measurement units (IMUs) are deployed at the characteristic corner points of the prefabricated pipe gallery segments. With the pipe gallery design axis as the reference, the attitude parameters of the pipe gallery during the dynamic jacking process are collected at high frequency, and the attitude deflection angle of the current jacking operation is extracted in real time.

[0037] Third, during the manufacturing stage of the prefabricated double-eared horseshoe-shaped sealing gasket 11, a micro-thin film pressure sensor array is symmetrically embedded in the circumferentially spaced interior of its pressure-bearing surface to obtain the mechanical parameters of the sealing gasket 11. The mechanical parameters of the sealing gasket 11 include the nonlinear elastic modulus, pressure-bearing area, equivalent stiffness coefficient, initial thickness, damping coefficient, compressive strain, real-time compression amount, compression rate, measured contact stress, and normalized transverse contact stress difference of the sealing gasket 11.

[0038] Among them, the measured contact stress of the sealing gasket 11 under the current pushing action is collected in real time during the extrusion process; The displacement sensor integrated inside each cylinder of the U-shaped shield is used to synchronously collect the real-time compression amount of the sealing gasket 11 corresponding to each cylinder in the current jacking operation; Extract nonlinear elastic modulus, bearing area, equivalent stiffness coefficient, initial thickness, and damping coefficient from the industrial control computer database; The equivalent stiffness coefficient is calculated by treating the sealing gasket 11 as an equivalent spring. The calculation formula is Hooke's Law, which states that within the elastic limit, the spring force F is proportional to the spring elongation or compression x, expressed as F=kx. From this, the formula for calculating the equivalent stiffness coefficient k is derived as k=F / x. Therefore, the equivalent stiffness coefficient of the sealing gasket 11 is calculated using the elastic force and compression of the sealing gasket 11.

[0039] Calculate the ratio of the real-time compression of the sealing gasket 11 to the initial thickness for each cylinder under the current jacking action, and obtain the compressive strain for each cylinder under the current jacking action.

[0040] Calculate the ratio of the real-time compression amount of the sealing gasket 11 corresponding to each cylinder under the current jacking operation to the time interval, and obtain the compression rate corresponding to each cylinder under the current jacking operation; Simultaneously, miniature earth pressure cells embedded in the left and right walls of the pipe gallery joint are used to collect real-time data on the lateral contact compressive stress during the current jacking operation. By calculating the absolute difference in contact compressive stress fed back by the earth pressure cells on both sides of the joint and dividing it by standard atmospheric pressure, the normalized lateral contact stress difference during the current jacking operation is obtained.

[0041] Meanwhile, by reading the pressure sensor data of the main pump station and each cylinder of the multi-cylinder hydraulic synchronous jacking system, the real-time observed thrust of the current jacking operation to overcome the formation resistance is calculated.

[0042] Thus, dynamic operational characteristic data of the prefabricated pipe gallery assembly and construction process were obtained.

[0043] A2, calculate the target reward value for the current push action.

[0044] It should be noted that in the dynamic jacking of the U-shaped shield, the total thrust needs to overcome the direct resistance of the front shield and the frictional resistance of the sidewalls and bottom that are invisible due to geological changes. Since solving for two independent unknown friction coefficients in reverse using only the real-time single total thrust is an underdetermined equation problem that traditional analytical algorithms cannot solve, this step introduces a Proximal Policy Optimization (PPO) reinforcement learning network as an online mechanical parameter identifier to explore and infer unknown friction characteristics.

[0045] To ensure that the parameters output by the network have real engineering physical meaning rather than being purely mathematical fitting, a physical evaluation bridge that strictly follows Newton's laws must be established: the candidate friction coefficients estimated by the network are substituted into the mechanical formula to reconstruct the theoretical thrust, and the target reward value with a pure negative penalty is constructed based on the relative error between the theoretical thrust and the observed thrust. Thus, by using the underlying driving mechanism of reinforcement learning to pursue the maximization of reward, that is, minimizing the error of physical thrust, the network gradient is forced to converge in the direction that is closest to the current real geomechanical boundary, thereby providing a basis for determining the optimal friction parameters (i.e., the optimal sidewall friction coefficient and the optimal bottom friction coefficient).

[0046] Specifically, set a length of The historical sliding window will display the current top action and past actions. The soil stress parameter sequence of the top thrust and the observed thrust sequence are combined to construct a multi-dimensional temporal feature matrix, which serves as the environmental state space input of the PPO network.

[0047] It should be added that, This value is used to define the temporal memory depth for the reinforcement learning network to extract the mechanical state of the environment. If this value is too large, the state matrix will be filled with outdated and invalid long-sequence historical data, making the network extremely sluggish in responding to abrupt changes in the current stratum boundary. If this value is too small, the network will be unable to capture the implicit evolution trend of the stratum frictional resistance in the time dimension, resulting in short-sighted thrust prediction and violent oscillations in model weights. Therefore, considering the upper limit of the computing power load of the edge control computer at the engineering site and the actual characteristics of the dynamic push for millisecond-level real-time calculation, its value ranges from 10 to 50. In this invention, it is set to 20 to ensure that the policy network can fully extract stable temporal evolution patterns without causing a computing power disaster due to excessively high state space dimensions. Implementers can adjust this value according to the frequency of stratum boundary changes and hardware computing power requirements in the actual geological survey report.

[0048] Since frictional resistance exhibits a history-dependent stick-slip dynamic process, the multidimensional temporal feature matrix is ​​first input into the sequence feature extraction layer of the PPO network. The sequence feature extraction layer contains a long short-term memory network to capture the implicit evolution of mechanical state and geological parameters in the time dimension. Subsequently, the extracted high-dimensional temporal feature vector is mapped to the probability distribution parameters of the continuous action space through a fully connected layer.

[0049] The action sampling layer of the PPO network samples actions based on a probability distribution and outputs a set of candidate friction coefficient action vectors. These action vectors are essentially the network's physical prediction of the friction characteristics of the current pushing action. To ensure the physical rationality of the predicted value, the candidate vector is then input into the physical boundary constraint layer. Using activation mapping and truncation functions, it is denormalized and restricted to the range of extreme friction coefficient values ​​allowed by the current geological survey report. Finally, the estimated sidewall friction coefficient and estimated bottom friction coefficient of the current jacking operation are generated, which serve as the actual action space for the decision-making process.

[0050] Substituting the estimated sidewall friction coefficient, estimated bottom friction coefficient, soil resistance in front of the shield, lateral active earth pressure, current contact length between the U-shaped shield and the lateral soil, and shield self-weight into the friction mechanics formula, the theoretical thrust of the current jacking operation is reconstructed. The relative error rate between the theoretical thrust and the observed thrust under the current jacking operation of the U-shaped shield is calculated to obtain the target reward value of the current jacking operation; the target reward value satisfies the expression: ; In the formula, This indicates the target reward value for the current top-level action; This indicates the observed thrust of the current top-push operation; This indicates the soil resistance in front of the current top-pushing action; , These represent the estimated sidewall friction coefficient and the estimated bottom friction coefficient of the current pushing action, respectively; This indicates the lateral active earth pressure exerted by the current top push; This indicates the current contact length between the U-shaped shield and the lateral soil. This indicates the weight of the U-shaped shield. This indicates taking the absolute value; 2 indicates the number of sidewalls; This represents the theoretical thrust of the current top-driven action.

[0051] in, This reflects the estimated sidewall friction force generated by the shield walls overcoming the lateral active earth pressure under the current jacking operation. The estimated bottom sliding friction force reflects the total self-weight of the shield system as the normal pressure source under the current jacking operation; the theoretical thrust is obtained by scalar summing the soil resistance in front of the shield, the estimated sidewall friction resistance, and the estimated bottom friction resistance under the current jacking operation, based on the friction coefficient estimated by the network. This reflects the relative matching error rate between the reconstructed theoretical thrust and the observed thrust under the current top-push operation.

[0052] Since the optimization goal of reinforcement learning is to maximize the reward value, the expression adopts a pure negative penalty mechanism: the closer the relative error is to 0, the smaller the penalty and the larger the target reward value. That is, the closer the negative value is to the theoretical maximum value of 0, the more the network is constrained in each iteration. This means that the friction features estimated by the network at the current time are more realistic at the physical level, thereby achieving high-precision online identification of unknown friction parameters.

[0053] For example, Figure 18 This is a schematic diagram of the force balance analysis of the integrated U-shaped shield jacking, which intuitively shows the dynamic and impedance relationship on the horizontal axis during the dynamic jacking process of the utility tunnel. The total thrust is the only driving force of the system and moves forward along the axis. The soil resistance in front of the shield, the sidewall friction, and the bottom friction constitute the environmental impedance, all of which strictly point in the opposite direction of the jacking. The diagram clearly distinguishes the force source and the resistance component: the lateral active earth pressure indicated by the dashed line and the vertically downward shield self-weight are the normal pressure sources. Through the interface friction effect, they are coupled and transformed into equivalent tangential friction resistance parallel to the axis.

[0054] At this point, the target reward value for the current top push action has been obtained.

[0055] A3 updates the PPO network by combining the target reward value and the target learning rate calculated for the current top push action, and outputs the optimal friction coefficient.

[0056] It should be noted that the target reward value for constructing mechanical evaluation is the learning rate, which is the core hyperparameter that determines whether the network can quickly adapt to the new strata in the real-time backpropagation training of industrial edge computing devices. In traditional reinforcement learning, the basic learning rate will decrease unidirectionally with the push time to ensure eventual convergence. However, when the U-shaped shield encounters a cliff-like change in the physical properties of the strata, the old friction parameters fail, causing the thrust error to soar. If the base learning rate has decayed to an extremely low level at this point, the network will lose its ability to correct itself and fall into the parameter solidification trap. Considering the limitations of computing power and the absolute real-time requirements of the underlying PLC or industrial control computer at the engineering site, this step abandons the time-consuming probabilistic optimization and introduces a linear gain compensation mechanism. By acquiring the abnormal fluctuation of thrust and using it to characterize the intensity of sudden changes in the geological environment, and using it as a direct feedback signal, the attenuated base learning rate is amplified in real time and linearly, giving the system a large step size for exploration when encountering sudden changes, thereby quickly extracting the optimal friction coefficient under the new stratum with extremely low time complexity.

[0057] Specifically, for the current top-down action, calculate its comparison with the past... The standard deviation of the observed thrust of each jacking operation is used as the thrust fluctuation rate of the stratum where the current jacking operation is located; and the absolute value of the difference between the observed thrust and the theoretical thrust of the current jacking operation is obtained as the thrust deviation of the current jacking operation; the intensity of geological environmental change under the current jacking operation is calculated based on the above indicators; the intensity of geological environmental change satisfies the expression: ; In the formula, This indicates the severity of sudden changes in the geological environment under the current top-down action; This indicates the thrust deviation of the current top-push operation; This represents the thrust volatility of the stratum where the current top thrust is located; This represents the natural exponential function.

[0058] in, This reflects the dimensionless anomaly deviation of the instantaneous thrust prediction error relative to the system's inherent historical mechanical noise. A larger ratio indicates that the current actual thrust error has significantly exceeded the system's normal physical fluctuation boundary, meaning the tunnel boring machine encountered a precipitous change in geotechnical properties during the jacking process, such as plunging from soft soil into hard rock. The previously learned and solidified frictional characteristic experience has become completely invalid, requiring immediate forced reconstruction of model parameters. Furthermore, based on... The nonlinear saturation constraint mapping mechanism has the following advantages: In the normal fluctuation range with small thrust deviation, the Taylor expansion of the function is used to approximate linear characteristics to maintain the system's high sensitivity to small formation changes; while when encountering extreme thrust anomalies with no physical upper limit, the decay ceiling characteristic of the natural exponent is used to forcibly and smoothly compress and clamp the deviation that tends to infinity within the safe range of [0,1).

[0059] Furthermore, the base learning rate set by the PPO network according to the conventional decay strategy is obtained. This base learning rate is then compensated for by the intensity of the geological environment change under the current jacking action, yielding the target learning rate for the current jacking action. The target learning rate satisfies the expression: ; In the formula, This represents the target learning rate for the current top-level action; Indicates the base learning rate; This indicates the severity of sudden changes in the geological environment under the current top-down action; This represents the gain coefficient.

[0060] During the normal and stable tunneling phase, When the base learning rate approaches zero, a stable fine-tuning of the base learning rate is maintained; however, in the event of sudden changes such as hard rock, When the multiplier approaches 1, it instantly linearly amplifies the learning rate. The gain factor provides a sufficient gradient update step size; the gain coefficient is rigorously derived based on the ratio between the initial exploration state of the deep learning network and the decay limit, obtaining the maximum initial learning rate of the PPO network in the early stages of training. When the algorithm becomes stuck in convergence, the base learning rate is used. To ensure that it can be instantly restored to its initial peak exploration capability in the event of extreme geological changes, the following conditions should be met: Therefore, the gain coefficient value is derived as follows: .

[0061] Finally, the PPO network directly uses the linearly compensated target learning rate as the step size, combined with the target reward value in A2, to perform deterministic fast gradient descent and network weight updates. The high-intensity learning rate forces the network to quickly discard old stratum experience and reconstruct the actual stress state of the new stratum. The output friction parameters gradually approach the true values, and the theoretical thrust error is reduced. When the target reward value reaches its maximum again and converges stably, it proves that the network has captured the physical resistance law of the current new stratum. At this time, the action vector finally stably output by the policy network at the physical boundary cutoff layer is the optimal sidewall friction coefficient and the optimal bottom friction coefficient of the geology under the current thrust action.

[0062] Thus, the optimal friction coefficient of the geology under the current top-pushing operation has been obtained.

[0063] A4, calculate the target thrust of any cylinder.

[0064] It should be noted that in the integrated construction of the U-shaped shield, the theoretical thrust is physically borne by the multi-cylinder hydraulic system symmetrically arranged in the tail space. If the theoretical thrust is simply distributed arithmetically among the cylinders, the physical premise is that the U-shaped shield, the prefabricated pipe gallery and the multi-cylinder thrust axis are in an absolutely ideal straight alignment state. However, in actual construction scenarios, due to uneven ground hardness or asymmetrical elongation of the hydraulic cylinder, the lateral attitude deviation of the pipe gallery axis is inevitable. The double-eared horseshoe-shaped sealing gasket 11 used at the joint of the precast pipe gallery is not a simple linear spring when compressed. Instead, due to the presence of the internal cavity, it exhibits severe nonlinear stress hardening constitutive characteristics before and after closure. Once the pipe gallery deviates, the sealing gasket 11 corresponding to the hydraulic cylinder on the deflection side will be excessively squeezed, instantly generating a large microscopic local rigid resistance. If the multi-cylinder hydraulic system still blindly executes the average output strategy at this time, it will not only fail to correct the attitude, but will also cause the yaw to be aggravated due to insufficient power on the deflection side, and may even instantly crush the rubber sealing gasket 11 and cause tunnel leakage. Therefore, it is necessary to analyze the degree of eccentricity of the hydraulic cylinder relative to the central axis based on the specific physical coordinates of the cylinder on the cross section of the shield body, and introduce this spatial position mapping mechanism to decouple the overall macroscopic attitude adjustment torque, so as to ensure that the hydraulic cylinders under different lateral force gradients can adaptively obtain the off-center load compensation, thereby accurately counteracting the local nonlinear hardening impedance and achieving active attitude correction in milliseconds.

[0065] Specifically, from the pre-design drawing of the U-shaped shield, the wall groups of any cylinder and the effective control area of ​​the cylinder on the corresponding wall are extracted. Each wall group includes: side wall group and bottom group. The effective control area is divided by the total area of ​​the corresponding group to obtain the force weight of any cylinder on each wall of the shield.

[0066] Substitute the optimal sidewall friction coefficient, lateral active earth pressure, and current contact length between the U-shaped shield and the lateral soil into the friction force calculation formula to obtain the sidewall friction force of the current jacking operation; Similarly, by substituting the optimal bottom friction coefficient of the current jacking operation and the self-weight of the U-shaped shield into the friction calculation formula, the bottom friction force of the current jacking operation can be obtained; Based on the ratio of the soil resistance in front of the shield to the total number of hydraulic cylinders during the current jacking operation, and the force weight of each hydraulic cylinder on each wall of the shield, the frictional forces of each wall are weighted to obtain the basic frictional force of any hydraulic cylinder during the current jacking operation. The basic frictional force satisfies the expression: ; In the formula, The first step in the current top-level promotion The basic friction force of each hydraulic cylinder; This indicates the soil resistance in front of the current top-pushing action; This represents the total number of hydraulic cylinders on the U-shaped shield. , These represent the optimal sidewall friction coefficient and the optimal bottom friction coefficient for the current pushing action, respectively. This indicates the lateral active earth pressure exerted by the current top push; This indicates the weight of the U-shaped shield. , These represent the first and second actions of the current top push. The force weight of each hydraulic cylinder on the side wall of the shield and the force weight at the bottom.

[0067] Furthermore, extract the lateral physical coordinates of any cylinder currently performing the jacking operation on the cross-section of the pipe rack, calculate the absolute horizontal distance from this coordinate point to the vertical centerline of the pipe rack cross-section, and divide it by the maximum half-width of the pipe rack cross-section to obtain the lateral eccentricity ratio. Based on the lateral eccentricity ratio, the attitude deflection angle of the current jacking operation, and the normalized lateral contact stress difference, calculate the adaptive off-center load compensation amount for any cylinder currently performing the jacking operation; the adaptive off-center load compensation amount satisfies the expression: ; In the formula, Indicates the current top-level push action. The adaptive off-center load compensation amount for each hydraulic cylinder; This represents the normalized lateral contact stress difference of the current top-pushing action; Indicates the attitude deflection angle of the current push operation; Indicates the current top-level push action. The lateral eccentricity ratio of each cylinder; This represents the thrust conversion factor for eccentric load compensation. Represents a symbolic function; Represents the natural logarithm function; Represents the sine function; This indicates taking the absolute value.

[0068] in, This reflects the exponentially increasing rigid resistance characteristics exhibited by the current jacking action sealing gasket 11 due to the cavity compaction caused by off-center loading. The logarithmic decay mechanism avoids the divergence of control commands caused by extreme contact stress differences. Map the attitude deflection to an axial geometric feature vector; This reflects the overall off-center load moment of the current jacking operation towards the current jacking operation. The spatial physical allocation process of each hydraulic cylinder A larger value indicates that the cylinder is at the farthest lever arm of the deflection and compression, meaning that the maximum proportion of dynamic compensation force can be allocated to counteract local distortion. It should be noted that the off-center load compensation thrust conversion coefficient has dimensions of... Its physical meaning is: to map the dimensionless attitude deviation and stress deviation into the actual output compensation thrust amplitude of the hydraulic cylinder. In order to prevent the precast pipe gallery segment from being damaged due to excessive correction thrust, the off-center load compensation thrust conversion coefficient is equal to the product of the current rated maximum thrust of the cylinder and the off-center load safety limit coefficient. The rated maximum thrust of the cylinder and the off-center load safety limit coefficient are read from the shield hydraulic system design drawings or the rated nameplate parameters of the hydraulic pump station.

[0069] Based on the compressive strain, basic friction force, adaptive off-center load compensation, nonlinear elastic modulus, bearing area, equivalent stiffness coefficient, real-time compression and damping coefficient, and compression rate of each cylinder in the current jacking operation, calculate the target jacking force of each cylinder in the current jacking operation; the target jacking force satisfies the expression: ; In the formula, Indicates the current top-level push action. The target thrust of each hydraulic cylinder; Represents the nonlinear elastic modulus; Indicates the pressure-bearing area; Indicates the current top-level push action. The compressive strain of each hydraulic cylinder; Indicates the equivalent stiffness coefficient; Indicates the current top-level push action. Real-time compression of the sealing gasket 11 corresponding to each hydraulic cylinder; Indicates the damping coefficient; Indicates the current top-level push action. The compression rate corresponding to each hydraulic cylinder; Indicates the current top-level push action. The basic friction force of each hydraulic cylinder; Indicates the current top-level push action. The adaptive off-center load compensation amount for each hydraulic cylinder.

[0070] in, This constitutes the Kelvin-Voigt generalized viscoelastic resistance model, which reflects the transient material internal stress that the cylinder must overcome during dynamic extrusion. The thrust distribution ensures that it includes the basic noise energy required for propulsion; while the active off-center load compensation force for forced attitude adjustment... This grants active control over environmental disturbances; the initial target thrust expression decouples the complex dynamic thrust requirement into three orthogonal dimensions: microscopic material impedance, macroscopic geological noise, and dedicated attitude adjustment driving force, ultimately yielding... The larger the value, the more intense the soil friction or the more severe the deflection and compression of the cylinder on that side. Under this mechanism, the system will directly bypass the traditional average distribution strategy and issue an initial hydraulic command that is far greater than the average value in real time, thereby independently counteracting impedance distortion and completing active correction in milliseconds.

[0071] At this point, the target thrust of each cylinder in the current jacking operation is obtained.

[0072] A5 adjusts the target jacking force of any cylinder to generate the target jacking force of the next jacking cylinder, so as to achieve dynamic jacking for waterproofing of precast pipe gallery joints.

[0073] It should be noted that the single-cylinder target thrust is a feedforward theoretical command. After the command is issued to the hydraulic servo system, the actual thrust acting on the pipe gallery will inevitably have a slight lag and attenuation due to the pressure loss along the pipeline fluid, mechanical friction gaps, and system response dead zone. If only open-loop control is relied upon, it is very easy to cause over-compression or under-compression of the sealing gasket 11 at the end of the jacking process. Furthermore, in the construction of prefabricated pipe gallery, the waterproofing quality depends on the final contact state of the double-eared horseshoe-shaped sealing gasket 11. For this type of sealing gasket 11, its optimal working stress range is 1.5MPa to 2.2MPa. Therefore, it is necessary to construct a fuzzy PID control closed loop with the absolute goal of controlling the full-section contact stress to steadily enter this optimal working range in order to suppress the underlying mechanical and physical execution errors.

[0074] Specifically, the median of the optimal working range is extracted from the waterproof rating design database of the industrial control computer as the target contact stress. The deviation between the target contact stress and the measured contact stress of the sealing gasket 11 under the current pushing operation, as well as the rate of change of the deviation, are calculated and input into the fuzzy controller. Based on the built-in membership function and engineering experience rule library, the fuzzy controller tunes the proportional coefficient of the PID controller in real time online. Integral coefficient With differential coefficients Subsequently, the target thrust of each cylinder in the next jacking operation is calculated using an adaptive fuzzy PID algorithm; the target thrust of the next jacking operation satisfies the expression: ; In the formula, Indicates the next push action The target thrust of each hydraulic cylinder; Indicates the current top-level push action. The target thrust of each hydraulic cylinder; These are the proportional, integral, and differential coefficients after dynamic tuning, respectively.

[0075] Finally, under fuzzy PID control, the hydraulic system fine-tunes the output of each cylinder in real time and smoothly until the structural fasteners at the joint are anchored, thus completing the dynamic jacking and one-time forming of the waterproofing of the precast pipe gallery joint.

[0076] The aforementioned fuzzy controller is a commonly used adaptive fuzzy PID controller in this field. In this embodiment, it is used to achieve specific quantitative calculations: The subsets are divided into 7 fuzzy subsets: {Negative Large (NB), Negative Medium (NM), Negative Small (NS), Zero (ZO), Positive Small (PS), Positive Medium (PM), Positive Large (PB)}. The membership function is a trigonometric function.

[0077] Based on the aforementioned common membership function and well-known engineering experience rule base, the fuzzy controller defuzzifies using the centroid method, allowing implementers to directly tune and output the proportional, integral, and derivative coefficients of the PID controller in real time online.

[0078] Example 4: Verification Experiment A simulation experiment was conducted based on the above system, and the results are shown below. Figures 19-21 .

[0079] Figure 19 This study demonstrates the mapping relationship between changes in microscopic friction parameters of the strata and the evolution of macroscopic thrust during the jacking process of a U-shaped shield tunneling machine. In the early stages of the jacking operation (approximately 0-55 steps), the U-shaped shield is situated in relatively soft or homogeneous soil layers, with the optimal sidewall friction coefficient remaining at a low and stable level, and the overall macroscopic thrust exhibiting gentle fluctuations. When the jacking progresses to around 60 steps (approximately 55-68 steps), the optimal sidewall friction coefficient experiences a sharp, smooth increase, indicating that the U-shaped shield has encountered a geological abrupt change (such as soft soil penetrating hard rock). Influenced by this change in objective geological characteristics, the macroscopic thrust experiences a real surge. This process reflects the multi-source state acquisition and PPO network reconstruction mechanism of this invention, which can realistically, objectively, and accurately capture the implicit subsurface impedance changes caused by abrupt strata changes.

[0080] Figure 20 This study demonstrates the evolution of spatial attitude deviation and microscopic lateral contact stress difference at the joints during the jacking process of the utility tunnel system. The curves show that in the 0-118 step range, the tunnel's attitude deflection angle is extremely small, the stress on both sides of the joint is relatively uniform, and the normalized lateral contact stress difference is low. However, around step 120 (118-128 step range), the attitude deflection angle suddenly and rapidly increases and remains at a high level. This spatial attitude distortion leads to a severe imbalance in the microscopic stress on the sealing gaskets 11 on both sides of the joint, manifested as a simultaneous surge in the normalized lateral contact stress difference. This clearly reflects a spatial deviation during the tunnel's movement. Without millisecond-level active correction and thrust redistribution, the deflected sealing gasket 11 faces an extremely high risk of stress hardening and even crushing.

[0081] Figure 21This paper presents a multi-dimensional comparison of the target thrust for independently controlled cylinder i under conditions of abrupt formation changes and uneven loading. The baseline for "theoretical thrust distribution based on the optimal friction coefficient" is derived by substituting the optimal friction coefficient dynamically identified by the PPO network into the theoretical calculation formula to obtain the total theoretical thrust to be provided, which is then directly distributed equally among the cylinders. The comparison shows that traditional methods, due to fixed parameters and blind reliance on observed thrust distribution, deviate significantly from the theoretical distribution baseline during 60-step formation changes, resulting in insufficient power. In contrast, this invention employs a formula that extracts the effective control area and divides it by the total area of ​​the group to calculate the force weight of the corresponding wall surface. Its target thrust curve (black solid line) closely follows the distribution baseline and achieves precise spatial allocation based on the actual force-bearing area ratio of the cylinder. The more significant advantage is reflected in the 120-step spatial off-center loading. The theoretical average baseline and the traditional method are completely unaware of the off-center angle. However, the method of this invention calculates the instantaneous adaptive off-center loading compensation by combining the lateral eccentricity ratio and the deflection angle, and actively sends out the high peak power required for correction.

[0082] Example 5: Three-layer waterproofing system To ensure that the pipe rack 15 and pipe rack 26 can be aligned during the jacking process, a concrete pin 10 can be installed in pipe rack 15. The concrete pin 10 is wrapped with rubber, which also serves as a waterproofing measure.

[0083] Fork openings are also provided at the top of pipe rack 15 and pipe rack 26, and the fork openings are sealed with two-component high-elasticity epoxy sealant.

[0084] This embodiment has a total of three waterproof systems: (1) applying two-component high-elasticity epoxy sealant to the fork joint for caulking, (2) pin, and (3) sealing gasket 11.

[0085] Figure 14 For the pipe gallery 15, which is a jacking pipe gallery, concrete pins 10 are installed in the pipe gallery 15 and wrapped with rubber on the outside, which also serves as a waterproof function. Figure 15 For the second type of pipe gallery 6, which is a load-bearing pipe gallery, a conical groove 8 is provided at the position corresponding to the concrete pin 10. The conical groove 8 is adapted to the concrete pin 10. During the jacking process, the concrete pin 10 is inserted into the conical groove 8 to achieve precise positioning.

[0086] The conical groove 8 can also be used for sealing tests. A grouting pipe 7 is installed inside pipe gallery 2 6, with one end of the grouting pipe 7 located outside pipe gallery 2 6 and the other end connected to the conical groove 8. After the jacking connection is completed, water is injected into the conical groove 8 through the grouting pipe 7 to test the sealing performance between pipe gallery 1 5 and pipe gallery 2 6. Figure 16 .

[0087] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for waterproofing precast pipe gallery joints used in integrated construction of U-shaped shield structures, characterized in that, Includes the following steps: Install the sealing gasket in the sealing groove of the first pipe rack segment; Calculate the target thrust of each cylinder of the U-shaped shield; Each cylinder of the U-shaped shield performs the target jacking force to push the second pipe gallery segment toward the first pipe gallery segment, compressing the sealing gasket to achieve waterproof treatment of the prefabricated pipe gallery joints. The calculation of the target thrust of each cylinder of the U-shaped shield includes: The observation of thrust, attitude parameters, soil stress parameters, and sealing gasket mechanical parameters under the real-time jacking action of the U-shaped shield were collected. Based on the observed thrust and soil stress parameters under the current jacking operation of the U-shaped shield and several historical jacking operations, a multi-dimensional time series feature matrix is ​​constructed. The multi-dimensional time series feature matrix is ​​input into the PPO network to output the estimated sidewall friction coefficient and the estimated bottom friction coefficient. The theoretical thrust is calculated by combining the estimated sidewall friction coefficient, the estimated bottom friction coefficient, and the soil stress parameters under the current jacking action of the U-shaped shield. The target bonus value is calculated based on the deviation between the theoretical thrust and the observed thrust under the current jacking action of the U-shaped shield. Calculate the intensity of geological environment change under the current jacking action of the U-shaped shield, use the intensity of geological environment change to perform gain compensation on the base learning rate of the PPO network to obtain the target learning rate; update the PPO network by combining the target reward value and the target learning rate, and output the optimal sidewall friction coefficient and the optimal bottom friction coefficient. For any cylinder of the U-shaped shield, the basic friction force of the cylinder is calculated based on the optimal sidewall friction coefficient, the optimal bottom friction coefficient, and the soil force parameters under the current jacking operation of the U-shaped shield; the adaptive off-center load compensation amount of the cylinder is calculated based on the attitude parameters; and the target jacking force of the cylinder is calculated by combining the adaptive off-center load compensation amount, the basic friction force, and the mechanical parameters of the sealing gasket under the current jacking operation of the U-shaped shield.

2. The method according to claim 1, characterized in that: The attitude parameter is the attitude deflection angle of the U-shaped shield during its current jacking action; The soil stress parameters include the soil resistance in front of the shield, the lateral active earth pressure, and the current contact length between the U-shaped shield and the lateral soil. The mechanical parameters of the sealing gasket include its nonlinear elastic modulus, bearing area, equivalent stiffness coefficient, initial thickness, damping coefficient, compressive strain, real-time compression amount, compression rate, measured contact stress, and normalized transverse contact stress difference.

3. The method according to claim 2, characterized in that: A multidimensional time-series feature matrix is ​​constructed based on the observed thrust and soil stress parameters under the current jacking action of the U-shaped shield and several historical jacking actions. This multidimensional time-series feature matrix is ​​input into the PPO network, which outputs estimated sidewall friction coefficients and estimated bottom friction coefficients, including: Based on the observed thrust and soil stress parameters under the current jacking operation of the U-shaped shield and several historical jacking operations, a multi-dimensional time series feature matrix is ​​constructed. The multidimensional temporal feature matrix is ​​input into the sequence feature extraction layer of the PPO network, and the extracted temporal feature vector is mapped to the probability distribution parameters of the continuous action space through the fully connected layer of the PPO network. The physical boundary constraint layer of the PPO network generates estimated sidewall friction coefficients and estimated bottom friction coefficients based on probability distribution parameters.

4. The method according to claim 3, characterized in that, The theoretical thrust is calculated by combining the estimated sidewall friction coefficient, the estimated bottom friction coefficient, and the soil stress parameters under the current jacking action of the U-shaped shield. The target bonus value is then calculated based on the deviation between the theoretical thrust and the observed thrust under the current jacking action of the U-shaped shield, including: The estimated sidewall friction force is obtained by calculating the product of the lateral active earth pressure, the current contact length between the U-shaped shield and the lateral soil, the estimated sidewall friction coefficient, and the number of sidewalls. The estimated bottom friction force is obtained by calculating the product of the bottom friction coefficient and the weight of the U-shaped shield. The theoretical thrust is obtained by summing the soil resistance in front of the shield, the estimated sidewall friction, and the estimated bottom friction. Calculate the absolute value of the difference between the observed thrust and the theoretical thrust under the current jacking action of the U-shaped shield, divide this absolute value by the observed thrust under the current jacking action of the U-shaped shield, and take the negative of the quotient to obtain the target reward value.

5. The method according to claim 4, characterized in that: The calculation of the intensity of geological environment abrupt change under the current jacking action of the U-shaped shield includes: Calculate the standard deviation of the observed thrust under the current jacking action of the U-shaped shield and several historical jacking actions, and use it as the thrust fluctuation rate; The absolute value of the difference between the observed thrust and the theoretical thrust under the current jacking action of the U-shaped shield is calculated as the thrust deviation. Calculate the ratio of thrust deviation to thrust fluctuation rate, and take the negative of this ratio as the exponential term. Calculate the exponential function value of the exponential term with the natural constant as the base to obtain the abrupt decay term. The difference between a constant 1 and the abrupt decay term is used as the intensity of geological environmental abrupt changes.

6. The method according to claim 5, characterized in that: The base learning rate of the PPO network is obtained by gain compensation based on the intensity of geological environmental changes; The PPO network is updated by combining the target reward value and the target learning rate, outputting the optimal sidewall friction coefficient and the optimal bottom friction coefficient, including: Calculate the target learning rate using the following formula: ； In the formula, This represents the target learning rate for the current top-level push. Indicates the base learning rate; This indicates the severity of sudden changes in the geological environment under the current top-down action; Indicates the gain coefficient; The PPO network uses the target learning rate as the step size and combines the target reward value to perform gradient descent and PPO network weight updates. After convergence and stabilization, it outputs the optimal sidewall friction coefficient and the optimal bottom friction coefficient.

7. The method according to claim 6, characterized in that: For any cylinder of the U-shaped shield, calculate the basic friction force of the cylinder based on the optimal sidewall friction coefficient, the optimal bottom friction coefficient, and the soil stress parameters; including: Extract the effective control area of ​​any cylinder of the U-shaped shield frame corresponding to the side wall group or bottom group and divide it by the total area of ​​the corresponding group to obtain the force weight of the side wall and bottom of the cylinder respectively. The average frontal resistance is calculated by dividing the soil resistance in front of the shield by the total number of hydraulic cylinders on the U-shaped shield. The optimal sidewall friction coefficient, lateral active earth pressure, current contact length between the U-shaped shield and the lateral soil, and the product of the number of sidewalls are calculated as the sidewall friction force, and multiplied by the sidewall force weight as the distributed sidewall friction force. The product of the optimal bottom friction coefficient and the self-weight of the U-shaped shield is calculated as the bottom friction force, and multiplied by the bottom force weight as the distributed bottom friction force. The basic friction force of the hydraulic cylinder is obtained by summing the average frontal resistance, the distributed sidewall friction force, and the distributed bottom friction force.

8. The method according to claim 7, characterized in that: The expression for the adaptive off-center load compensation is: ； In the formula, Indicates the current top-level push action. The adaptive off-center load compensation amount for each hydraulic cylinder; This represents the normalized lateral contact stress difference of the current jacking action of the U-shaped shield. This indicates the attitude deflection angle of the U-shaped shield during its current jacking action; This indicates the current jacking action of the U-shaped shield. The lateral eccentricity ratio of each cylinder; This represents the thrust conversion factor for eccentric load compensation. Represents a symbolic function; Represents the natural logarithm function; Represents the sine function; This indicates taking the absolute value.

9. The method according to claim 8, characterized in that: The expression for the target thrust is: ； In the formula, This indicates the current jacking action of the U-shaped shield. The target thrust of each hydraulic cylinder; Represents the nonlinear elastic modulus; Indicates the pressure-bearing area; This indicates the current jacking action of the U-shaped shield. The compressive strain of each hydraulic cylinder; Indicates the equivalent stiffness coefficient; This indicates the current jacking action of the U-shaped shield. Real-time compression of the sealing gasket corresponding to each hydraulic cylinder; Indicates the damping coefficient; This indicates the current jacking action of the U-shaped shield. The compression rate corresponding to each hydraulic cylinder; This indicates the current jacking action of the U-shaped shield. The basic friction force of each oil cylinder.

10. The method according to claim 9, characterized in that: The hydraulic cylinders of the U-shaped shield execute the target jacking force to advance the second tunnel segment toward the first tunnel segment, including: Set the target contact stress for the sealing gasket; The deviation between the target contact stress and the measured contact stress, as well as the rate of change of this deviation, are calculated and input into the fuzzy controller. The proportional coefficient, integral coefficient, and derivative coefficient of the PID controller are then tuned. The product of the deviation between the target contact stress and the measured contact stress and the proportionality coefficient is calculated as the proportional adjustment term; The integral of the deviation between the target contact stress and the measured contact stress over time, multiplied by the integral coefficient, is used as the integral adjustment term. The product of the derivative of the deviation between the target contact stress and the measured contact stress with respect to time and the derivative coefficient is used as the differential adjustment term; The target thrust of the hydraulic cylinder is summed with the proportional, integral, and derivative adjustment terms to obtain the target thrust of each hydraulic cylinder of the U-shaped shield for the next thrust action.

Images