Shield tunnel ground surface settlement prediction method and system based on time sequence and physical information

Abstract

The present application relates to the technical field of tunnel and underground engineering, and particularly relates to a shield tunnel surface subsidence prediction method and system based on time sequence and physical information. The method comprises the following steps: acquiring multi-source heterogeneous data in the shield construction process, sequentially reconstructing to obtain space-time characteristics; inputting the space-time characteristics into a pre-trained prediction network to output a surface subsidence prediction value; when the prediction network is trained, a composite loss function including a data fitting loss term, a physical consistency loss term and an asymmetric physical boundary penalty term is used for training; the prediction network comprises a parallel backbone network and a physical perception branch network; the space-time characteristics are input into the backbone network to output the surface subsidence prediction value; and the space-time characteristics are input into the physical perception branch network to output a theoretical subsidence reference value. The present application can integrate the physical priori knowledge in shield tunneling into the training process of the deep learning model, and can improve the accuracy and physical reliability of shield tunnel surface subsidence prediction.

Description

Technical Field

[0001] This invention relates to the field of tunnel and underground engineering technology, specifically to a method and system for predicting surface settlement of shield tunnels based on time-series and physical information. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As underground space development continues to advance into deep and complex strata, ultra-large diameter shield tunnels have become the core construction method for cross-river and cross-sea tunnels. However, when traversing complex strata with uneven hardness, the intense excavation and unloading of ultra-large cross sections can easily disrupt the initial stress balance of the strata, inducing surface subsidence that is difficult to prevent.

[0004] Currently, deep learning methods are commonly used to predict surface settlement of shield tunnels. However, existing deep learning prediction methods lack the ability to deeply couple with and adaptively constrain the physical mechanisms of shield construction. This leads to the model blindly converging in the early stages of training, detached from the physical background. During the fitting process, the lack of flexible physical boundaries results in oscillations or overfitting that violate the laws of mechanics. At the same time, the fixed-weight multi-objective optimization method is difficult to coordinate the conflict between data and physics, thus limiting the accuracy, stability, and physical reliability of settlement prediction as a whole. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for predicting surface settlement of shield tunnels based on temporal and physical information. This invention can effectively and adaptively integrate prior physical knowledge from shield tunneling into the training process of a deep learning model, thereby improving the accuracy, stability, and physical reliability of surface settlement prediction for shield tunnels.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, a method for predicting surface settlement of shield tunnels based on temporal and physical information is provided, including: Acquire multi-source heterogeneous data during the tunnel boring machine construction process; sequentially reconstruct the multi-source heterogeneous data to obtain spatiotemporal features; The spatiotemporal features are input into a pre-trained prediction network to output predicted values ​​of land subsidence. When the prediction network is trained, a composite loss function including a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term is used for training. The physical consistency loss term is the difference between the predicted value of land subsidence and the theoretical subsidence reference value. The prediction network includes a parallel backbone network and a physical sensing branch network; the spatiotemporal features are input into the backbone network and the predicted surface subsidence value is output; the spatiotemporal features are input into the physical sensing branch network and the theoretical subsidence reference value is output.

[0007] Secondly, a surface settlement prediction system for shield tunnels based on temporal and physical information is provided, including: The data acquisition and feature extraction module is configured to: acquire multi-source heterogeneous data during the shield tunneling process; and sequentially reconstruct the multi-source heterogeneous data to obtain spatiotemporal features. The prediction model is configured to: input the spatiotemporal features into a pre-trained prediction network and output predicted land subsidence values; during training, the prediction network is trained using a composite loss function including a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term, wherein the physical consistency loss term is the difference between the predicted land subsidence value and the theoretical subsidence reference value; the prediction network includes a parallel backbone network and a physical sensing branch network; the spatiotemporal features are input into the backbone network and output predicted land subsidence values; the spatiotemporal features are input into the physical sensing branch network and output theoretical subsidence reference values.

[0008] Thirdly, an electronic device is also provided, comprising: Memory is used to store computer-readable instructions in a non-transitory manner. Processor, for executing the computer-readable instructions, When the computer-readable instructions are executed by the processor, they perform the method described in the first aspect above.

[0009] Fourthly, a computer-readable storage medium is provided having a program stored thereon that, when executed by a processor, implements the method described in the first aspect above.

[0010] Fifthly, a computer program product is provided, employing the following technical solution: A computer program product includes software code, wherein a program in the software code performs the steps of the method described in the first aspect of the present invention.

[0011] The above technical solution has the following advantages or beneficial effects: (1) Before inputting spatiotemporal features into a double-layer stacked LSTM network, this invention initializes the LSTM network state based on mechanical priors: extracting the latest spatial physical features within the spatiotemporal feature sliding window, inputting them into a mechanical prior state initializer, and transforming the spatial physical features into initial hidden states and cell states that match the dimensions of the LSTM network through fully connected mapping layers and layer normalization operations, and dynamically multiplying by a learnable soft-start gating factor. This mechanism breaks the limitation of the traditional LSTM defaulting to zero-state initialization, enabling the network to have physical prior states related to the geological geometry in the early stages of training, effectively avoiding early blind convergence, and significantly improving the convergence stability and physical logic consistency of model training.

[0012] (2) The composite loss function constructed in this invention includes a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term. Among them, the asymmetric physical boundary penalty term can force the inverted stratum loss parameter to not exceed the theoretical limit geometric gap, thereby forcing the prediction network to converge strictly within the solution space that conforms to the laws of rock and soil elasticity and geometric boundaries.

[0013] (3) The present invention sets learnable adaptive weights that are automatically updated with backpropagation, which can dynamically balance the high-frequency fine-tuning of data and the low-frequency support of physics, avoiding the conflict problem of multi-objective optimization under the traditional fixed weight method, and enabling the model to adaptively adjust the contribution ratio of data-driven and physics-driven at different training stages.

[0014] (4) When the predicted surface settlement value of this invention exceeds the control threshold or the boundary of the inverted physical parameters exceeds the limit, the Shapley additive interpretation (SHAP) is automatically triggered. SHAP can calculate the contribution of each dimension of construction parameters to the current abnormal settlement in real time. Based on the quantitative tracing results, dynamic adjustment instructions are issued to the front-end operators to guide them to adjust the synchronous grouting volume of the next stage or control the shield attitude, realizing the transformation from numerical fitting early warning to physical causal root cause investigation and intelligent management. Attached Figure Description

[0015] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0016] Figure 1 This is a flowchart of a shield tunnel surface settlement prediction method based on time-series and physical information in a specific embodiment of the present invention; Figure 2 This is a comparison diagram of the boundary verification of shield tail void filling efficiency and inverted formation loss physical parameters in a specific embodiment of the present invention; Figure 3This is a comparison chart showing the surface settlement prediction of a shield tunnel based on time and physical information in a specific embodiment of the present invention, and the actual surface settlement in the test set. Detailed Implementation

[0017] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0018] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the invention. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] In this embodiment of the invention, "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of this invention, "multiple" refers to two or more.

[0020] Furthermore, to facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0021] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0022] Example 1 This embodiment provides a method for predicting surface settlement of shield tunnels based on time-series and physical information, including: S1: Acquire multi-source heterogeneous data during the shield tunneling construction process; S2: Sequentially reconstruct multi-source heterogeneous data to obtain spatiotemporal features; S3: Input the spatiotemporal features into the pre-trained prediction network and output the predicted value of land subsidence; when training the prediction network, a composite loss function including a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term is used for training, and the physical consistency loss term is the difference between the predicted value of land subsidence and the theoretical subsidence reference value; the prediction network includes a parallel backbone network and a physical perception branch network; input the spatiotemporal features into the backbone network and output the predicted value of land subsidence; input the spatiotemporal features into the physical perception branch network and output the theoretical subsidence reference value.

[0023] In this embodiment, a super-large diameter water-balanced shield tunnel through yellow mud is used as the application scenario, and the specific steps implemented in this embodiment are given in detail. The excavation diameter is set to 16.8m.

[0024] In this embodiment, the method for predicting surface settlement of shield tunnels based on time and physical information is also known as the LSTM-PINN method.

[0025] Step S1 is as follows: Relying on the SCADA system mounted on the tunnel boring machine, real-time data on tunneling dynamics, spatial physics, and geological environment are collected. Tunneling dynamics data includes total thrust, cutterhead torque, tunneling speed, cutterhead water pressure, and synchronous grouting volume. Spatial physics data includes tunnel depth and the gap between the shield tail sections after assembly.

[0026] In this embodiment, the multi-source heterogeneous data includes tunneling dynamics data, spatial physics data, geological environment data, and current project design parameters. The design parameters include the predetermined tunnel excavation radius and Poisson's ratio of the soil.

[0027] In some embodiments, a Savitzky-Golay filter is also used to smooth and denoise the tunneling dynamics data, spatial physics data, and geological environment data.

[0028] Step S2 is as follows: Because shield settlement has a spatiotemporal lag effect, the sliding window technique is used to sequentially reconstruct the tunneling dynamics data, spatial physics data, and geological environment data in the form of discrete single-loop multi-source heterogeneous data. The movement of the spatial tunneling position is implicitly mapped to the evolution of the time series, and the dimension is increased to transform it into spatiotemporal features containing the current loop and historical continuous evolution information, so as to match the temporal memory perception characteristics of the LSTM network.

[0029] In this embodiment, the spatiotemporal characteristics specifically include tunneling dynamic characteristics, spatial physical characteristics, and geological environment characteristics.

[0030] In this embodiment, the sliding window sequence length is set to 3.

[0031] Step S3 is as follows: A prediction network is constructed, comprising a parallel backbone network and a physical sensing branch network. In this embodiment, a two-layer stacked LSTM network is used as the backbone network. The physical sensing branch network comprises a single-layer LSTM network and a differentiable physical layer.

[0032] In this embodiment, before inputting the spatiotemporal features into the stacked LSTM network, the LSTM network is initialized based on a mechanical prior. Specifically, the spatial physical features at the latest time within the spatiotemporal feature sliding window are extracted and input into the mechanical prior state initializer. Through fully connected mapping layers and layer normalization operations, and dynamically multiplied by a learnable soft-start gating factor, the spatial physical features are transformed into an initial hidden state matching the LSTM network dimension. and cell state ;combining the spatiotemporal characteristics with the aforementioned initial state that includes mechanical priors The high-dimensional hidden features are extracted from the common input double-layer stacked LSTM backbone network, and finally the predicted value of land subsidence at time t is output.

[0033] In this embodiment, the LSTM network is initialized in both the training and actual prediction phases. During the prediction phase of the actual tunnel boring machine (TBM) construction, a state initializer generates an initial state for the LSTM network with the current physical context before settlement time-series prediction is performed. If this step is not performed during prediction, the prediction network will lose its physical reference.

[0034] In this embodiment, the mechanical prior state initializer includes two parallel mapping branches, which are used to generate hidden states. and cell state The two parallel mapping branches have identical structures, including a fully connected mapping layer, a layer normalization layer, a soft-start gating regulator, and a nonlinear activation layer connected in sequence. The fully connected mapping layer receives the spatial physical features of the current time step and linearly maps them from a low-dimensional physical space to a high-dimensional feature space matching the dimensions of the LSTM network's hidden layers. The layer normalization layer normalizes the mapped high-dimensional features to prevent gradient explosion or network dead zones caused by differences in physical magnitudes during the initial feature mapping. The soft-start gating regulator dynamically multiplies the normalized features by a learnable soft-start gating factor to smoothly control the strength of the physical prior state's takeover of the LSTM's initial state. The nonlinear activation layer reconstructs the dimensions of the soft-started features and uses a hyperbolic tangent function to nonlinearly activate the cell state features, ultimately outputting an initial hidden state that conforms to the LSTM network's input format. and cell state .

[0035] In this embodiment, spatiotemporal features are input into a single-layer LSTM network within the physical perception branch network. The single-layer LSTM network implicitly inverts the equivalent ground loss parameters under the current working conditions. The inverted equivalent ground loss parameters, along with the tunnel excavation radius, tunnel depth, and soil Poisson's ratio, are substituted into a differentiable physical layer (i.e., the modified Loganathan & Poulos elastic foundation analytical solution). The differentiable physical layer includes a predefined modified equivalent ground loss rate formula. After calculation, the theoretical settlement reference value is output.

[0036] The predefined formula for the corrected equivalent formation loss rate is as follows: (1); in, This is a theoretical settlement reference value. For equivalent formation loss parameters, Poisson's ratio of soil The radius of the tunnel excavation. This refers to the tunnel's burial depth.

[0037] In this embodiment, when the tunnel excavation radius, tunnel depth, and Poisson's ratio of the soil are input into the differentiable physical layer in the physical sensing branch network, a physical dimension protection mechanism is executed, and no standardization operation is performed, forcibly retaining their original metric units.

[0038] Construct a composite loss function for the prediction network, which includes a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term, specifically: (2); in, For composite loss function, For data fitting loss, For learnable adaptive weights that are automatically updated with backpropagation, For physical consistency loss, Penalty for asymmetric physical boundaries; This is the physical consistency relaxation coefficient, with a value of 0.1. This is the physical boundary penalty coefficient, with a value of 10.0. Data fitting loss. This is the difference between the actual and predicted values ​​of surface subsidence, representing the loss of physical consistency. This represents the difference between the predicted surface subsidence and the theoretical reference subsidence. The formation loss parameter that can be forcibly inverted by the asymmetric physical boundary penalty term must not exceed the theoretical limit geometric void.

[0039] Learnable adaptive weights It will automatically update with backpropagation, dynamically balancing high-frequency fine-tuning of data with low-frequency physical support, thereby forcing the prediction network to converge strictly within the solution space that conforms to the laws of rock and soil elasticity and geometric boundaries.

[0040] Experimental verification like Figure 2 As shown, the shield tail void filling efficiency and the boundary of the inverted formation loss parameter are verified. Inversion is performed on a pre-defined validation set. The results show that the formation loss parameter g inverted by a single-layer LSTM in the physical sensing branch network is strictly limited by the composite loss function to the theoretical limit geometric void. Below the physical envelope. This proves that the asymmetric physical boundary penalty mechanism applied in this embodiment is effective, forcing the network to converge in the solution space that conforms to the laws of geotechnical mechanics, thus preventing blind fitting that deviates from physical reality.

[0041] like Figure 3 As shown, the method proposed in this embodiment is used to predict surface subsidence, and the results are compared with the actual surface subsidence values ​​in the test set. The method proposed in this embodiment can closely match the actual subsidence curve on the test set, effectively capturing the nonlinear fluctuation characteristics of subsidence, proving that the prediction network still has extremely high prediction accuracy under complex strata.

[0042] The aforementioned pre-defined validation set refers to the data collected during the historical tunnel boring machine (TBM) construction process. After obtaining spatiotemporal features through sequential reconstruction using a sliding window in step S2, the data is divided according to a pre-defined ratio (e.g., 80% used as the training set for network parameter updates, and the remaining 20% ​​used as an independent validation set). This portion of data does not participate in the gradient backpropagation of the composite loss function and is specifically used to evaluate and invert the physical parameter performance of the model under unknown working conditions.

[0043] In some embodiments, when the predicted surface settlement exceeds the control threshold or the inverted physical parameters exceed the boundary limits, Shapley Additive Explanations (SHAP) are automatically triggered. SHAP can calculate the contribution of each dimension of construction parameters to the current abnormal settlement in real time. Based on the quantitative tracing results, dynamic adjustment instructions are issued to the front-end operators, guiding them to adjust the synchronous grouting volume of the next stage or control the shield attitude, realizing the transformation from numerical fitting early warning to physical causal root cause investigation and intelligent management.

[0044] Example 2 This embodiment provides a shield tunnel surface settlement prediction system based on time-series and physical information, including: The data acquisition and feature extraction module is configured to: acquire multi-source heterogeneous data during the shield tunneling process; and sequentially reconstruct the multi-source heterogeneous data to obtain spatiotemporal features. The prediction model is configured to: input the spatiotemporal features into a pre-trained prediction network and output predicted land subsidence values; during training, the prediction network is trained using a composite loss function including a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term, wherein the physical consistency loss term is the difference between the predicted land subsidence value and the theoretical subsidence reference value; the prediction network includes a parallel backbone network and a physical sensing branch network; the spatiotemporal features are input into the backbone network and output predicted land subsidence values; the spatiotemporal features are input into the physical sensing branch network and output theoretical subsidence reference values.

[0045] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0046] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0047] Example 3 This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the method described in Embodiment 1.

[0048] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0049] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0050] In the implementation process, each step of the above method can be completed by the integrated logic circuits in the processor hardware or by software instructions.

[0051] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0052] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0053] Example 4 Embodiment 4 of the present invention provides a computer-readable storage medium.

[0054] A computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the steps of the method as described in Embodiment 1 of the present invention.

[0055] The detailed steps are the same as those provided in Example 1, and will not be repeated here.

[0056] Example 5 Embodiment 5 of the present invention provides a computer program product.

[0057] A computer program product includes software code, wherein the program in the software code performs the steps described in Embodiment 1 of the present invention.

[0058] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting surface settlement of shield tunnels based on temporal and physical information, characterized in that, include: Acquire multi-source heterogeneous data during the tunnel boring machine (TBM) construction process; The spatiotemporal features are obtained by sequentially reconstructing the multi-source heterogeneous data. The spatiotemporal features are input into a pre-trained prediction network to output predicted values ​​of land subsidence. When the prediction network is trained, a composite loss function including a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term is used for training. The physical consistency loss term is the difference between the predicted value of land subsidence and the theoretical subsidence reference value. The prediction network comprises a parallel backbone network and a physical sensing branch network; a double-layer stacked LSTM network is used as the backbone network, and the physical sensing branch network comprises a single-layer LSTM network and a differentiable physical layer; the spatiotemporal features are input into the backbone network and the predicted surface subsidence value is output; the spatiotemporal features are input into the physical sensing branch network and the theoretical subsidence reference value is output. The multi-source heterogeneous data includes tunneling dynamics data, spatial physics data, geological environment data, and current project design parameters; the spatial physics data includes tunnel depth and shield tail clearance after assembly; the current project design parameters include tunnel excavation radius and soil Poisson's ratio. A sliding window is used to sequentially reconstruct the tunneling dynamics data, spatial physical data, and geological environment data to obtain spatiotemporal features; the spatiotemporal features include tunneling dynamics features, spatial physical features, and geological environment features. The spatial physical features of the latest time within the sliding window are input into the mechanical prior state initializer, which outputs the initial hidden state and cell state. The initial hidden state, cell state, and spatiotemporal features are then input into a two-layer stacked LSTM network to output the predicted value of land subsidence.

2. The method for predicting surface settlement of shield tunnels based on time-series and physical information as described in claim 1, characterized in that, The physical sensing branch network includes a single-layer LSTM network and a differentiable physical layer; the spatiotemporal features are input into the single-layer LSTM network, and the equivalent stratum loss parameters are output; the equivalent stratum loss parameters, tunnel excavation radius, tunnel burial depth, and soil Poisson's ratio are substituted into the differentiable physical layer, and the theoretical settlement reference value is output.

3. The method for predicting surface settlement of shield tunnels based on time-series and physical information as described in claim 1, characterized in that, The composite loss function is specifically as follows: in, For composite loss function, For data fitting loss, For learnable adaptive weights that are automatically updated with backpropagation, For physical consistency loss, For asymmetric physical boundary penalties, The physical consistency relaxation coefficient; This is the physical boundary penalty coefficient.

4. A shield tunnel surface settlement prediction system based on time-series and physical information, employing the shield tunnel surface settlement prediction method based on time-series and physical information as described in any one of claims 1-3, characterized in that, include: The data acquisition and feature extraction module is configured to acquire multi-source heterogeneous data during the tunnel boring machine (TBM) construction process. The spatiotemporal features are obtained by sequentially reconstructing the multi-source heterogeneous data. The prediction model is configured to: input the spatiotemporal features into a pre-trained prediction network and output predicted land subsidence values; during training, the prediction network is trained using a composite loss function including a data fitting loss term, a physical consistency loss term, and an asymmetric physical boundary penalty term, wherein the physical consistency loss term is the difference between the predicted land subsidence value and the theoretical subsidence reference value; the prediction network includes a parallel backbone network and a physical sensing branch network; the spatiotemporal features are input into the backbone network and output predicted land subsidence values; the spatiotemporal features are input into the physical sensing branch network and output theoretical subsidence reference values.

5. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the shield tunnel surface settlement prediction method based on time and physical information as described in any one of claims 1-3.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the steps in the shield tunnel surface settlement prediction method based on time and physical information as described in any one of claims 1-3.

7. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps in the shield tunnel surface settlement prediction method based on time-series and physical information as described in any one of claims 1-3.

Images