Longitudinal deformation prediction method for lining structure of shield tunnel during construction period based on cascaded pinn

Abstract

The application relates to the technical field of construction safety control, and discloses a shield tunnel construction period lining structure longitudinal deformation prediction method based on a cascade PINN. The method comprises the following steps: a data set containing shield tunneling parameters and stratum parameters is constructed, a longitudinally continuous equivalent model and a physical control equation thereof are established, and a cascade physical information neural network model is innovatively adopted; through the series connection architecture of a load prediction model and a deformation prediction model, the directly unmeasurable load prediction is provided with physical constraints by using the monitorable deformation data, and the physical information invalidation problem caused by the load data loss in the traditional prediction model is effectively solved. The application realizes the full-process explicit modeling from shield tunneling to lining load to tunnel deformation, improves the prediction accuracy under the condition that the load true value is not needed, and enhances the interpretability of the model by outputting physical quantities such as equivalent load distribution.

Description

Technical Field

[0001] This invention relates to the field of construction safety control technology, and in particular to a method for predicting the longitudinal deformation of the lining structure during the construction of a shield tunnel based on cascaded PINN. Background Technology

[0002] During shield tunnel construction, inaccurate prediction of longitudinal deformation of the lining structure can easily lead to disasters such as segment misalignment, joint leakage, and structural damage, and exacerbate ground disturbance and surface subsidence, threatening project safety.

[0003] While traditional data-driven neural network methods can predict deformation, they neglect the physical and mechanical transmission process that "shield tunneling causes stress redistribution in the strata, which leads to changes in the load acting on the lining and thus causes longitudinal deformation of the tunnel structure." This results in poor model interpretability and difficulty in verifying prediction results.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a method for predicting the longitudinal deformation of the lining structure during the construction period of a shield tunnel based on cascaded PINN, aiming to solve the problems of poor interpretability and difficulty in verifying prediction results in the existing longitudinal deformation prediction model of the lining structure.

[0006] To achieve the above objectives, this invention provides a method for predicting the longitudinal deformation of the lining structure during the construction period of a shield tunnel based on cascaded PINN. The method includes the following steps: Obtain a training dataset, which includes shield tunneling parameters, stratum parameters, and corresponding longitudinal deformation monitoring values ​​of the shield tunnel in the constructed section. A cascaded physical information neural network model is constructed, comprising a first physical information neural network sub-model and a second physical information neural network sub-model. The first physical information neural network sub-model is configured to take the shield tunneling parameters and the stratum parameters as inputs and the equivalent longitudinal load applied to the tunnel lining as output. The second physical information neural network sub-model is configured to take the equivalent longitudinal load as input and the tunnel longitudinal deformation monitoring value as output, and embed physical control equations characterizing the longitudinal stress-deformation mechanism of the tunnel lining. The cascaded physical information neural network model is trained using the training dataset to obtain the target cascaded physical information neural network model. The shield tunneling parameters and stratum parameters of the section to be predicted are input into the target cascaded physical information neural network model to obtain the predicted value of the tunnel's longitudinal deformation.

[0007] Furthermore, the process of obtaining the training dataset also includes, prior to this: The hyperparameters in the cascaded physical information neural network model are globally searched and optimized using the Bayesian optimization algorithm. The hyperparameters include: bending stiffness reduction factor, shear stiffness reduction factor, load normalization parameter, and physical loss weight factor.

[0008] Furthermore, the step of training the cascaded physical information neural network model using the training dataset to obtain the target cascaded physical information neural network model specifically includes: The training dataset is input into the cascaded physical information neural network model for forward computation, and the longitudinal deformation prediction value of the tunnel is obtained by sequentially passing through the first physical information neural network sub-model and the second physical information neural network sub-model of the cascaded physical information neural network model. The error between the predicted value of the tunnel longitudinal deformation and the actual value of the tunnel longitudinal deformation is calculated, and the network parameters of the first physical information neural network sub-model and the second physical information neural network sub-model are updated according to the error using the backpropagation algorithm.

[0009] Furthermore, the total loss function of the cascaded physical information neural network model includes at least a data-driven loss term and a physical information loss term, wherein the data-driven loss term is calculated based on the difference between the tunnel longitudinal deformation monitoring value and the predicted value of the second physical information neural network sub-model, and the physical information loss term is calculated based on the residual of the physical control equation.

[0010] Furthermore, the total loss function is: ; in, For data-driven loss terms, For the loss term in the physical equation, n The total number of samples, This is the predicted value of the longitudinal deformation of the tunnel. This represents the true value of the tunnel's longitudinal deformation. For physical information weighting coefficients, This is the predicted load value. For the first i An estimated value for each tunneling mileage. GA For shear stiffness, EI For bending stiffness, k b and k s These are the reduction factors for bending and shear stiffness, respectively.

[0011] Furthermore, the tunneling parameters include one or more of the following: total thrust, torque, tunneling speed, soil chamber pressure, grouting pressure, and grouting volume.

[0012] Furthermore, the physical governing equations are as follows: ; in, w For longitudinal deformation of the tunnel, q For equivalent load, y For tunneling mileage, This represents the fourth derivative of tunnel deformation with respect to the tunneling mileage. The second derivative of the equivalent load with respect to the tunneling mileage is given. GA For shear stiffness, EI For bending stiffness, k b and k s These are the reduction factors for bending and shear stiffness, respectively.

[0013] Furthermore, the geological parameters include one or more of the deformation modulus, cohesion, and internal friction angle of the overlying soil and the base soil.

[0014] The beneficial effects of this invention are as follows: By cascading the first and second physical information neural network sub-models, the load prediction model can be indirectly constrained and trained using only monitorable tunnel deformation data, allowing physical laws to be reliably embedded even in the absence of some key data. Through this cascaded structure, the model not only outputs the final deformation prediction but also simultaneously outputs intermediate physical quantities (equivalent loads), making the prediction process physically transparent and significantly improving the model's interpretability. This allows engineers to understand and evaluate the physical rationality of the prediction results, enhancing its credibility and usability in practical engineering decisions. This invention enhances the model's generalization ability and reliability on unseen data, providing a solid technical foundation for advanced and precise deformation control during construction. This invention fully utilizes readily available tunneling parameters, geological survey data, and deformation monitoring data during construction to obtain a model with solidified knowledge. Subsequently, in subsequent construction, only planned or real-time tunneling parameters and geological information need to be input to directly output the deformation prediction of future sections. This represents a shift from "lagging monitoring" to "advanced prediction," providing real-time decision support for dynamically optimizing tunneling parameters and proactively controlling construction risks. Attached Figure Description

[0015] Picture 1 This is a flowchart of a preferred embodiment of the method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction of the present invention; Picture 2This is a schematic diagram of the core architecture and workflow of the cascaded physical information neural network model described in this invention. Detailed Implementation

[0016] This application provides a method for predicting the longitudinal deformation of the lining structure during the construction period of a shield tunnel based on a cascaded PINN (Physics-Informed Neural Network). To make the purpose, technical solution, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0017] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0018] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0019] The preferred embodiment of the present invention describes a method for predicting the longitudinal deformation of shield tunnel lining structures during construction based on cascaded PINN. Picture 1 As shown, the method for predicting the longitudinal deformation of the lining structure during the construction period of a shield tunnel based on cascaded PINN includes the following steps: S10. Obtain the training dataset, which includes shield tunneling parameters, stratum parameters, and corresponding longitudinal deformation monitoring values ​​of the shield tunnel in the constructed section.

[0020] The purpose of this step is to provide a real and reliable data foundation for the subsequent construction and training of the Cascaded Physics-Informed Neural Network (CPINN) model. The quality and representativeness of the dataset directly determine the model's prediction accuracy and generalization ability.

[0021] In this embodiment, the training dataset is derived from the on-site monitoring data of the "Xingye Expressway" ultra-large diameter shield tunnel project in City A. The tunnel has a diameter D=15.76m and a total length of 1739m. This study selects a 240-meter section (120 rings) as the research object. This section has a relatively shallow burial depth (13.1~14.3m) and traverses significantly varying strata, including complex geological conditions such as soft upper layers and hard lower layers, cutting into moderately weathered granite. It represents a typical working condition for studying longitudinal deformation prediction.

[0022] The constructed dataset mainly includes the following three types of data: (I) Shield tunneling data: This reflects the state of the tunnel boring machine during construction and is the direct cause of ground disturbance and tunnel deformation. The data is taken from sensor records during the construction of each ring, including: total thrust of the propulsion cylinders ( F b ), cutter head torque ( To ), total contact force ( Fc ), penetration ( P ), average propulsion speed ( v ), Total mud delivery ( V f ), Total amount of sludge discharged ( V D ), Excavated amount ( ΔV )wait.

[0023] (II) Stratigraphic Data: Characterizing the mechanical properties of the soil surrounding the tunnel, directly affecting load distribution and deformation response. Due to the heterogeneity of the strata and the uncertainty of the exploration data, this application has processed key stratigraphic parameters, including: the corrected deformation modulus of the overburden (…). Eo' ), modified cohesion ( co' Corrected internal friction angle ( φo' ), and the corresponding parameters of the underlying soil (foundation soil). E F , C F ,φ F These parameters are obtained from geological survey reports and interpolation methods.

[0024] (III) Deformation monitoring data: namely, the settlement monitoring values ​​of the tunnel lining arch, used to characterize the longitudinal deformation of the tunnel. wMonitoring is conducted after each ring segment has stabilized, serving as the target for model training and validation.

[0025] The specific variable definitions for the dataset are shown in Table 1: Table 1: PINN Model Variables

[0026] Furthermore, the shield tunneling parameters include one or more of the following: total thrust, torque, tunneling speed, soil chamber pressure, grouting pressure, and grouting volume; and the stratum parameters include one or more of the following: deformation modulus, cohesion, and internal friction angle of the overlying soil and the foundation soil.

[0027] In this embodiment, the shield tunneling parameters specifically include total thrust ( F b ), cutter head torque ( T o ), penetration ( P ), propulsion speed ( v ) etc.; specific stratigraphic parameters include the deformation modulus of the overlying and underlying soils ( E ), cohesion ( c ), internal friction angle ( φ These parameters together constitute the model input, comprehensively reflecting the construction and geological conditions.

[0028] Furthermore, the process of obtaining the training dataset also includes, prior to: S0. Using the Bayesian optimization algorithm, perform a global search optimization on the hyperparameters in the cascaded physical information neural network model. The hyperparameters include, but are not limited to: bending stiffness reduction factor, shear stiffness reduction factor, load normalization parameter, and physical loss weighting factor.

[0029] The purpose of this step is to automatically and efficiently determine the optimal values ​​of key physical hyperparameters in the CPINN model, which directly affect the accuracy of physical constraints and the model's final predictive performance.

[0030] In this embodiment, the hyperparameters that need to be optimized include: longitudinal bending stiffness reduction factor ( k b ), longitudinal shear stiffness reduction factor ( k s ), equivalent load q Standardized mean ( μ q ) and variance σ q ), and the weighting coefficients of the physical loss term ( λThe objective function (model loss) is probabilistically modeled and iteratively searched using the Bayesian optimization algorithm, ultimately yielding the hyperparameter combination that minimizes the loss function.

[0031] S20. Construct a cascaded physical information neural network model, which includes a first physical information neural network sub-model and a second physical information neural network sub-model. The first physical information neural network sub-model is configured to take the shield tunneling parameters and the stratum parameters as inputs and the equivalent longitudinal load applied to the tunnel lining as output. The second physical information neural network sub-model is configured to take the equivalent longitudinal load as input and the tunnel longitudinal deformation monitoring value as output, and embed physical control equations characterizing the longitudinal stress and deformation mechanism of the tunnel lining.

[0032] The purpose of this step is to construct a two-level cascaded neural network architecture to explicitly model the complete physical causal chain of "shield tunneling → lining load → tunnel deformation", thereby solving the problem of "physical information failure" caused by the lack of true load values ​​in traditional PINN.

[0033] In this embodiment, the core architecture of the Cascaded Physical Information Neural Network (CPINN) model is as follows: Picture 2 As shown, its workflow is as follows: 1. First physical information neural network sub-model (i.e., load prediction model, denoted as PINN-q): Its physical meaning is to quantify the process of "shield tunneling → lining load". This model takes shield tunneling parameters, stratum parameters and tunneling mileage (y) as inputs, and outputs the equivalent longitudinal distributed load (q) acting on the tunnel lining.

[0034] 2. Second physical information neural network sub-model (i.e., deformation prediction model, denoted as PINN-w): Its physical meaning is to quantify the process of "lining load → tunnel deformation". This model takes the equivalent load (q) predicted by PINN-q and the tunneling mileage (y) as input, and outputs the longitudinal deformation of the tunnel (w).

[0035] 3. Cascading and Embedding of Physical Information: The output (q) of PINN-q is used as the input of PINN-w, and the two are connected in series. Embedded in PINN-w is the physical governing equation derived from the longitudinal continuous equivalent model based on Timoshenko beam theory. This physical governing equation establishes the mechanical relationship between the load q and the deformation w, specifically: ; in, w For longitudinal deformation of the tunnel, q For equivalent load, y For tunneling mileage, This represents the fourth derivative of tunnel deformation with respect to the tunneling mileage. The second derivative of the equivalent load with respect to the tunneling mileage is given. GA For shear stiffness, EI For bending stiffness, k b and k s These are the reduction factors for bending and shear stiffness, respectively.

[0036] This equation treats the tunnel lining as a continuous beam considering shear deformation, thus achieving soft constraints based on the laws of physics and mechanics.

[0037] It should be noted that, in Picture 2 middle, Let be the acquisition function, representing the combination of hyperparameters to be optimized; where the expression of the acquisition function is: . This represents the number of internal iterations of PINN. This is the preset maximum number of internal iterations. Stop the internal iteration and output the current value. w and q The predicted value. Epoch is the global training epoch. max This is the preset maximum global training epoch, where Epoch ≥ Epoch max Stop global training at the specified time and output the final prediction results and hyperparameters.

[0038] Further, training the cascaded physical information neural network model using the training dataset to obtain the target cascaded physical information neural network model specifically includes: The training dataset is input into the cascaded physical information neural network model for forward computation, and the longitudinal deformation prediction value of the tunnel is obtained by sequentially passing through the first physical information neural network sub-model and the second physical information neural network sub-model of the cascaded physical information neural network model. The error between the predicted value of the tunnel longitudinal deformation and the actual value of the tunnel longitudinal deformation is calculated, and the network parameters of the first physical information neural network sub-model and the second physical information neural network sub-model are updated according to the error using the backpropagation algorithm.

[0039] The purpose of this step is to elaborate on the training mechanism of the CPINN model. The key is that, although the true value of the load q is unavailable, the error generated by monitoring the true value of tunnel deformation w can be backpropagated (e.g., through the concatenation of two subnetworks). Picture 2 (As shown) The network parameters of PINN-w and PINN-q are updated simultaneously. This allows PINN-q to learn and output a reasonable load prediction value q under the indirect constraint of the true value of w.

[0040] Furthermore, the total loss function of the cascaded physical information neural network model includes at least a data-driven loss term and a physical information loss term, wherein the data-driven loss term is calculated based on the difference between the tunnel longitudinal deformation monitoring value and the predicted value of the second physical information neural network sub-model, and the physical information loss term is calculated based on the residual of the physical control equation.

[0041] The purpose of this step is to define the optimization objective that drives model training. The loss function consists of two parts that work together to ensure that the model both fits the observed data and obeys physical laws.

[0042] Furthermore, the total loss function is: ; in, For data-driven loss terms, For the loss term in the physical equation, n The total number of samples, This is the predicted value of the longitudinal deformation of the tunnel. This represents the true value of the tunnel's longitudinal deformation. For physical information weighting coefficients, This is the predicted load value. For the first i An estimated value for each tunneling mileage. GA For shear stiffness, EI For bending stiffness, k b and k s These are the reduction factors for bending and shear stiffness, respectively.

[0043] In this embodiment, the model training uses the Adam optimizer to minimize the total loss function.

[0044] S30. Use the training dataset to train the cascaded physical information neural network model to obtain the target cascaded physical information neural network model.

[0045] The purpose of this step is to obtain a mature, parameter-optimized model that can be used for prediction through the actual training process.

[0046] In this embodiment, the dataset is divided into a training set (first 80%, rings 0-97) and a test set (last 20%, rings 98-120) according to the tunneling order to simulate a real-time advance prediction scenario. First, the optimal combination of hyperparameters is determined through Bayesian optimization as described in step S0. Then, the constructed CPINN model is trained using the training set data. After training, the model is validated on an independent test set.

[0047] S40. Input the shield tunneling parameters and stratum parameters of the section to be predicted into the target cascaded physical information neural network model to obtain the predicted value of the tunnel longitudinal deformation.

[0048] The purpose of this step is to use the trained model to make advanced and accurate predictions of the longitudinal deformation of tunnels in unconstructed sections, so as to guide construction decisions and risk control.

[0049] In this embodiment, during the subsequent construction of the shield tunnel, it is only necessary to obtain the shield tunneling parameters (such as preset thrust and torque) and geological survey data (stratum parameters) for the current and future planned sections, and input them into the trained CPINN model. The model can then output the predicted value of the longitudinal deformation of the tunnel in that section. This allows the construction unit to assess deformation risks in advance, optimize tunneling parameters, achieve proactive control, and ensure construction safety and structural stability.

[0050] The beneficial effects of this invention are as follows: By cascading the load prediction model (PINN-q) and the deformation prediction model (PINN-w) together, the load prediction model can be indirectly constrained and trained using only monitorable tunnel deformation data during training. This allows physical laws to be reliably embedded even in the absence of some key data. Through the cascaded structure, the model not only outputs the final deformation prediction but also simultaneously outputs intermediate physical quantities (equivalent loads), making the prediction process physically transparent and significantly improving the model's interpretability. This allows engineers to understand and evaluate the physical rationality of the prediction results, enhancing its credibility and usability in practical engineering decisions. By solving the problem of physical information failure, this invention greatly enhances the model's generalization ability and reliability on unseen data, providing a solid technical foundation for advanced and precise deformation control during construction. This invention fully utilizes readily available tunneling parameters, geological survey data, and deformation monitoring data during construction to obtain a model with solidified knowledge through training. Subsequently, in subsequent construction, only planned or real-time tunneling parameters and geological information need to be input to directly output the deformation prediction of future sections. This represents a shift from "lagging monitoring" to "advanced prediction," providing real-time decision support for dynamically optimizing tunneling parameters and proactively controlling construction risks. In summary, this invention solves the problem of embedding physical information when load data is missing through an innovative cascaded network structure, thereby achieving high-precision, interpretable, and advanced prediction of longitudinal deformation during shield tunnel construction, forming a complete solution that combines theoretical innovation with practical engineering value.

[0051] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0052] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0053] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for predicting longitudinal deformation of shield tunnel lining structure during construction based on cascaded PINN, characterized in that, The method for predicting the longitudinal deformation of the shield tunnel lining structure during construction based on cascaded PINN includes the following steps: Obtain a training dataset, which includes shield tunneling parameters, stratum parameters, and corresponding longitudinal deformation monitoring values ​​of the shield tunnel in the constructed section. A cascaded physical information neural network model is constructed, comprising a first physical information neural network sub-model and a second physical information neural network sub-model. The first physical information neural network sub-model is configured to take the shield tunneling parameters and the stratum parameters as inputs and the equivalent longitudinal load applied to the tunnel lining as output. The second physical information neural network sub-model is configured to take the equivalent longitudinal load as input and the tunnel longitudinal deformation monitoring value as output, and embed physical control equations characterizing the longitudinal stress-deformation mechanism of the tunnel lining. The cascaded physical information neural network model is trained using the training dataset to obtain the target cascaded physical information neural network model. The shield tunneling parameters and stratum parameters of the section to be predicted are input into the target cascaded physical information neural network model to obtain the predicted value of the tunnel's longitudinal deformation.

2. The method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction, as described in claim 1, is characterized in that... The process of obtaining the training dataset previously included: The hyperparameters in the cascaded physical information neural network model are globally searched and optimized using the Bayesian optimization algorithm. The hyperparameters include: bending stiffness reduction factor, shear stiffness reduction factor, load normalization parameter, and physical loss weight factor.

3. The method for predicting longitudinal deformation of shield tunnel lining structure during construction based on cascaded PINN as described in claim 1, characterized in that, The step of training the cascaded physical information neural network model using the training dataset to obtain the target cascaded physical information neural network model specifically includes: The training dataset is input into the cascaded physical information neural network model for forward computation, and the longitudinal deformation prediction value of the tunnel is obtained by sequentially passing through the first physical information neural network sub-model and the second physical information neural network sub-model of the cascaded physical information neural network model. The error between the predicted value of the tunnel longitudinal deformation and the actual value of the tunnel longitudinal deformation is calculated, and the network parameters of the first physical information neural network sub-model and the second physical information neural network sub-model are updated according to the error using the backpropagation algorithm.

4. The method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction, as described in claim 1, is characterized in that... The total loss function of the cascaded physical information neural network model includes at least a data-driven loss term and a physical information loss term. The data-driven loss term is calculated based on the difference between the longitudinal deformation monitoring value of the tunnel and the predicted value of the second physical information neural network sub-model, and the physical information loss term is calculated based on the residual of the physical control equation.

5. The method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction, as described in claim 4, is characterized in that... The total loss function is: ； in, For data-driven loss terms, For the loss term in the physical equation, n The total number of samples, This is the predicted value of the longitudinal deformation of the tunnel. This represents the true value of the tunnel's longitudinal deformation. For physical information weighting coefficients, This is the predicted load value. For the first i An estimated value for each tunneling mileage. GA For shear stiffness, EI For bending stiffness, k b and k s These are the reduction factors for bending and shear stiffness, respectively.

6. The method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction, as described in claim 1, is characterized in that... The tunnel boring machine (TBM) parameters include one or more of the following: total thrust, torque, tunneling speed, soil chamber pressure, grouting pressure, and grouting volume.

7. The method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction, as described in claim 1, is characterized in that... The physical control equation is: ； in, w For longitudinal deformation of the tunnel, q For equivalent load, y For tunneling mileage, This represents the fourth derivative of tunnel deformation with respect to the tunneling mileage. The second derivative of the equivalent load with respect to the tunneling mileage is given. GA For shear stiffness, EI For bending stiffness, k b and k s These are the reduction factors for bending and shear stiffness, respectively.

8. The method for predicting longitudinal deformation of shield tunnel lining structure based on cascaded PINN during construction, as described in claim 1, is characterized in that... The geological parameters include one or more of the following: deformation modulus of the overlying soil and the base soil, cohesion, and internal friction angle.

Images