Shield tunnel parameter prediction method and system based on deep learning

Abstract

The invention discloses a shield tunnel parameter prediction method and system based on deep learning, and relates to the technical field of geological engineering data analysis, and the method comprises the steps: collecting the dynamic working parameters of a shield tunneling machine, and generating a point cloud matrix; performing position coding and high-frequency feature mapping on the three-dimensional coordinates and the multi-dimensional attribute information in the point cloud matrix to generate a standardized geologic feature vector set; constructing a physical simulation deviation model, inputting the ground surface settlement prediction value into the physical simulation deviation model, outputting shield tunnel parameters, and generating a deviation analysis report based on the shield tunnel parameters; the adjusted model parameters are applied to a physical simulation deviation model, and updated shield tunnel parameters are generated through iteration processing. According to the invention, through fusion of deep learning and physical simulation modeling, high-precision prediction and dynamic optimization of shield tunnel parameters are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of geological engineering data analysis, and in particular to a shield tunnel parameter prediction method and system based on deep learning. BACKGROUND

[0002] Shield tunnel parameter prediction usually adopts a method combining historical data statistical analysis and physical mechanism model to realize the guidance of excavation parameter optimization and ground settlement prediction. Such a method relies on multi-source sensor data acquisition and space-time alignment technology, and completes geological feature extraction and settlement trend deduction through traditional machine learning or numerical simulation means to provide decision basis for intelligent tunnel construction.

[0003] However, the shield tunnel parameter prediction method in the prior art has limitations in capturing high-frequency details of various composite geological features, and the spatio-temporal correlation of multi-source heterogeneous data is not fully analyzed, resulting in difficulty in further improving the prediction accuracy and generalization ability. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides a shield tunnel parameter prediction method based on deep learning to solve the problems of insufficient representation of high-frequency geological features and insufficient fusion of multi-source data.

[0006] To solve the above technical problems, the present application provides the following technical solutions: In a first aspect, the present application provides a shield tunnel parameter prediction method based on deep learning, which comprises: collecting shield machine dynamic working parameters and generating a point cloud matrix; performing position encoding and mapping processing on three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix to generate a standardized geological feature vector set; fusing the standardized geological feature vector set and the shield machine dynamic working parameters to generate a multi-dimensional spatio-temporal feature tensor, and performing correlation analysis of geological settlement trend and equipment parameters based on the multi-dimensional spatio-temporal feature tensor to generate a ground settlement prediction value; constructing a physical simulation bias model, inputting the ground settlement prediction value into the physical simulation bias model, outputting a shield tunnel parameter, and generating a bias analysis report based on the shield tunnel parameter; taking the bias value of the shield tunnel parameter as an optimization target, adjusting the model parameters of the physical simulation bias model to generate adjusted model parameters; applying the adjusted model parameters to the physical simulation bias model to generate updated shield tunnel parameters through iterative processing.

[0007] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, wherein: the generation of the point cloud matrix is specifically as follows: the tunneling parameters in the shield machine dynamic working parameters are executed time alignment to obtain a tunneling parameter sequence; the tunneling parameter sequence is executed coordinate system with the geological correlation parameters, position and attitude parameters, running state parameters, and environment and auxiliary parameters in the shield machine dynamic working parameters to generate tunneling state data; the tunneling state data is executed spatial grid reorganization to generate the point cloud matrix. As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, wherein: the position encoding and mapping processing of the three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix is specifically as follows: based on the spatial distribution of the point cloud matrix, a mapping relationship framework of coordinates to geological features is established, and a continuous mapping function of spatial points to geological attribute values is established; the high-frequency response range of the position encoding is expanded according to the continuous mapping function, and a parameterized model for continuously representing geological features is constructed; the position encoding of each three-dimensional point in the point cloud matrix is executed based on the parameterized model to obtain an encoded point cloud; the high-frequency mapping processing of the three-dimensional coordinates in the encoded point cloud is executed to generate high-dimensional position features.

[0008] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, wherein: the fusion of the standardized geological feature vector set and the shield machine dynamic working parameters generates a multi-dimensional space-time feature tensor, and the specific steps are as follows: the time sequence alignment of the standardized geological feature vector set and the shield machine dynamic working parameters is executed to obtain a joint feature sequence; the spatial reorganization processing of the joint feature sequence is executed to generate the multi-dimensional space-time feature tensor.

[0009] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, wherein: the correlation analysis of the geological settlement trend and the equipment parameters based on the multi-dimensional space-time feature tensor generates a ground settlement prediction value, and the specific steps are as follows: the feature decomposition of the multi-dimensional space-time feature tensor obtains a long-term trend tensor of geological settlement and an equipment parameter tensor; the ground settlement prediction value is obtained based on the correlation calculation of the long-term trend tensor and the equipment parameter tensor.

[0010] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, the physical simulation deviation model is constructed by integrating the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters based on the interaction between the shield tunnel axis geometric parameters and the stratum, and obtaining a physical simulation basic framework.

[0011] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, the physical simulation deviation model is constructed by integrating the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters based on the interaction between the shield tunnel axis geometric parameters and the stratum, and obtaining a physical simulation basic framework.

[0012] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, the physical simulation deviation model is constructed by integrating the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters based on the interaction between the shield tunnel axis geometric parameters and the stratum, and obtaining a physical simulation basic framework.

[0013] As a preferred scheme of the shield tunnel parameter prediction method based on deep learning, the physical simulation deviation model is constructed by integrating the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters based on the interaction between the shield tunnel axis geometric parameters and the stratum, and obtaining a physical simulation basic framework. The adjusted model parameters are updated to the physical simulation deviation model to obtain an updated physical simulation deviation model. Based on the updated physical simulation deviation model, deviation calculation is performed on the shield tunnel parameters to obtain a deviation value. Based on the deviation value, gradient adjustment is performed on the shield tunnel parameters to generate a target function of the shield tunnel parameters, and physical simulation deviation calculation is performed through finite difference method to obtain the gradient direction of the shield tunnel parameters. Based on the gradient direction of the shield tunnel parameters, iterative updating is performed on the shield tunnel parameters to generate the updated shield tunnel parameters.

[0014] In a second aspect, the present application provides a deep learning-based shield tunnel parameter prediction system, comprising: a parameter processing module, configured to obtain shield machine dynamic working parameters and generate a point cloud matrix; a feature encoding module, configured to perform position encoding and mapping processing on three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix to generate a standardized geological feature vector set; a prediction analysis module, configured to fuse the standardized geological feature vector set and the shield machine dynamic working parameters to generate a multi-dimensional spatio-temporal feature tensor, perform correlation analysis of geological settlement trend and equipment parameters based on the multi-dimensional spatio-temporal feature tensor, and generate a ground settlement prediction value; a model analysis module, configured to build a physical simulation bias model, input the ground settlement prediction value into the physical simulation bias model, generate shield tunnel parameters, and generate a bias analysis report based on the shield tunnel parameters; a parameter calibration module, configured to take the bias value of the shield tunnel parameters as an optimization target, adjust the model parameters of the physical simulation bias model, and generate adjusted model parameters; and a parameter updating module, configured to apply the adjusted model parameters to the physical simulation bias model, and generate updated shield tunnel parameters through iterative processing.

[0015] The present application has the following advantages: the present application generates a multi-dimensional spatio-temporal feature tensor by fusing a standardized geological feature vector set and shield machine dynamic working parameters to realize high-dimensional spatio-temporal correlation representation of geological attributes and equipment states, and improve the accuracy and reliability of settlement trend prediction. In addition, the present application corrects and controls shield tunnel parameters by building and iteratively optimizing a physical simulation bias model, thereby enhancing the adaptability and decision support capability of the system under complex geological conditions. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0017] Fig. 1 Flowchart of the deep learning-based shield tunnel parameter prediction method.

[0018] Fig. 2 Schematic diagram of the generation of the point cloud matrix.

[0019] Fig. 3 Flowchart of the generation of feature encoding and multi-dimensional spatio-temporal feature tensor.

[0020] Fig. 4 Flowchart of the construction and update optimization of the physical simulation bias model. DETAILED DESCRIPTION

[0021] In order to make the above objectives, features and advantages of the present application more apparent, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0022] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without the specific details given herein. In other instances, well-known methods have not been described in detail in order to avoid unnecessarily obscuring the present application. Therefore, the specific embodiments given herein are not to be interpreted as limiting the scope of the application.

[0023] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments.

[0024] Reference Figs. 1-4 For one embodiment of the present application, the embodiment provides a shield tunnel parameter prediction method based on deep learning, comprising the following steps: S1: collecting shield machine dynamic working parameters and generating a point cloud matrix.

[0025] In the present application, the generation of the point cloud matrix in step S1 includes the following steps A1 to A3: Step A1: performing time alignment on the tunneling parameters in the shield machine dynamic working parameters to obtain a tunneling parameter sequence.

[0026] Specifically, through time alignment processing, the data points corresponding to the tunneling parameters in the shield machine dynamic working parameters are arranged and adjusted in time sequence, ensuring that all data points have consistent time references, thereby obtaining the tunneling parameter sequence.

[0027] It should be noted that the time alignment processing is to synchronize the tunneling parameters with the shield machine dynamic working parameters in the time dimension by adjusting the time stamp, so that the data points have consistent time references.

[0028] The shield machine dynamic working parameters refer to the assembly of various sensors on the shield machine to collect different types of parameters in real time during shield tunnel construction, specifically the tunneling parameter sequence, including time sequence data such as thrust force, cutter torque, and thrust speed; geological correlation parameters, including earth pressure, mud pressure, and stratum density; position and attitude parameters, including GPS coordinates, mileage, pitch angle, and roll angle; operating state parameters, including cutter rotation speed, screw conveyor rotation speed, and equipment temperature; environmental and auxiliary parameters, including grouting pressure, grouting volume, and environmental temperature.

[0029] It should be pointed out that the data point refers to each independent record in the tunneling state data, which contains the tunneling parameters of the shield machine dynamic working, the geological correlation parameters, the position and attitude parameters, the system running state parameters and the environment and auxiliary parameters. Step A2: The tunneling parameter sequence in the shield machine dynamic working parameter is coordinated with the geological correlation parameter, the position and attitude parameter, the running state parameter, and the environment and auxiliary parameter in the shield machine dynamic working parameter to generate the tunneling state data.

[0030] Specifically, the tunneling parameter sequence and the geological correlation parameter, the position and attitude parameter, the running state parameter and the environment and auxiliary parameter in the shield machine dynamic working parameter are unified into the same coordinate system through time synchronization and space coordinate conversion processing, so as to ensure that all data points are aligned in time and space, thereby generating the tunneling state data.

[0031] Step A3: The spatial grid reorganization is performed on the tunneling state data to generate a point cloud matrix.

[0032] Specifically, the three-dimensional coordinates (x, y, z) of each data point are calculated based on the position and attitude parameters in the tunneling state data, and the three-dimensional coordinates and the corresponding original shield machine dynamic working parameters (including the tunneling parameters, the geological correlation parameters, the position and attitude parameters, the running state parameters, and the environment and auxiliary parameters collected in step S1) are arranged in a grid according to the spatial position, forming a two-dimensional matrix, each row of the two-dimensional matrix corresponding to a spatial point, and the column corresponding to the three-dimensional coordinate value and the multi-dimensional attribute value, thereby generating the point cloud matrix.

[0033] It should be pointed out that in the present application, the spatial point refers to the position representation of the data point in the three-dimensional space after coordinate conversion, each spatial point not only has three-dimensional coordinates (x, y, z), but also carries all multi-dimensional attribute information of the corresponding data point. Therefore, the spatial point can be understood as the mapping form of the data point in the unified spatial coordinate system. The three-dimensional coordinates of each spatial point are calculated by coordinate conversion of the position and attitude parameters in the tunneling state data, representing the position values of the spatial point in the X-axis, Y-axis and Z-axis directions under the unified spatial coordinate system. The multi-dimensional attribute information represents all shield machine dynamic working parameter values associated with each spatial point, specifically including the original values of the tunneling parameters, the geological correlation parameters, the position and attitude parameters, the running state parameters, and the specific values of the environment and auxiliary parameters. S2: The three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix are position encoded and mapped to generate a standardized geological feature vector set.

[0034] In the present application, the specific operation of generating the standardized geological feature vector set in step S2 includes the following steps B1 to B4: Step B1: based on the spatial distribution of the point cloud matrix, a mapping relationship framework of coordinates to geological features is established, and a continuous mapping function of spatial points to geological attribute values is established.

[0035] Further, the point cloud matrix has a spatial range, which refers to a three-dimensional space region defined by the maximum and minimum values of the three-dimensional coordinates of all data points in the point cloud matrix in the X, Y and Z directions, specifically the cuboid space of [X min ,X max ] in the X direction, [Y min ,Y max ] in the Y direction, and [Z min ,Z max ] in the Z direction.

[0036] Further, the three-dimensional coordinates are mapped to the geological attribute values of the corresponding grid cells, and a mapping relationship framework of coordinates to geological features is established, and the specific operation is as follows: According to the spatial range of the point cloud matrix, a regular grid is established; Using the bilinear interpolation method, for any target position, the geological attribute values of the data points in the four adjacent grid cells around the target position are weighted and averaged, and the weight is determined according to the Euclidean distance from the target position to each adjacent data point. The Euclidean distance refers to the straight line distance between the three-dimensional coordinates (x, y, z) of the target position and the three-dimensional coordinates of the adjacent data points. The weight is inversely proportional to the Euclidean distance, that is, the data points with closer distance have greater weight, and the data points with farther distance have smaller weight. Then the continuous geological attribute value of the target position is obtained, thereby establishing a continuous mapping function of spatial points to geological attribute values.

[0037] It should be noted that the present application establishes a corresponding relationship between three-dimensional space coordinates and geological attribute values through a continuous mapping function. Among them, the input of the continuous mapping function is the three-dimensional coordinates (x, y, z) of any position in the unified spatial coordinate system, and the output is the corresponding geological attribute value of this arbitrary position. By constructing the continuous mapping function through the bilinear interpolation method, the geological attribute values originally existing only at discrete spatial point positions are extended to any position within the spatial range of the point cloud matrix, thereby realizing the conversion from discrete sampling to continuous field.

[0038] The working principle of the continuous mapping function is that for any target position in the unified spatial coordinate system, the continuous mapping function identifies a plurality of spatial points closest to the target position, and calculates the geological attribute value of the target position according to the geological attribute values of the plurality of spatial points and the distance relationship between the spatial points and the target position. The closer the spatial point is to the target position, the greater the influence of the spatial point on the target position, and the farther the spatial point is from the target position, the smaller the influence of the spatial point on the target position. Therefore, the continuous mapping function can reasonably estimate the corresponding geological attribute value of any target position (x, y, z) in the spatial range of the point cloud matrix by distance weighting, even if the target position is not the position of the original spatial point.

[0039] Preferably, the conversion from the discrete spatial points to the continuous spatial field in step B1 enables the position encoding processing in step B2 and the high-frequency mapping processing in step B4 to be performed at any target position in the spatial range of the point cloud matrix, rather than being limited to the positions of the original spatial points, thereby improving the accuracy and continuity of the geological feature representation.

[0040] It should be noted that the bilinear interpolation method in the present application calculates the spatial data of the geological attribute value by performing linear interpolation in two directions respectively. The present application uses the bilinear interpolation method in the spatial gridding and reorganization stage of the point cloud matrix to interpolate the discrete geological attribute values to generate a continuous mapping function. Specifically, the geological attribute value of any point is obtained by weighted average calculation of the geological attribute values of adjacent grid points in the horizontal direction and the vertical direction. The geological attribute value in the present application refers to a continuous variable used to represent the physical and mechanical properties of the stratum during shield tunneling, and is collected in real time through field exploration data, geological radar detection or sensors, reflecting the stability, drillability and influence on the surrounding environment of the stratum at different spatial positions, and serving as a basic component of the multi-dimensional attribute information of the geological feature in the point cloud matrix together with the three-dimensional coordinates.

[0041] Step B2: extending the high-frequency response range of the position encoding according to the continuous mapping function, and constructing a parameterized model for continuously representing the geological feature.

[0042] Specifically, according to the mapping relationship between the spatial points and the geological attribute values defined by the continuous mapping function, the high-frequency response range is extended by enhancing the density and range of the frequency components in the position encoding, thereby constructing a parameterized model for continuously representing the geological feature based on learnable parameters.

[0043] It should be noted that the high-frequency response range refers to the range of the ability of the position encoding function to represent rapid changes or high-frequency details in the spatial position. The parameterized model for continuously representing geological features in the present application is a mathematical model that converts three-dimensional coordinates into a high-dimensional encoding vector. The parameterized model uses a plurality of sine and cosine basis functions of different frequencies to map each three-dimensional coordinate (x, y, z) to a high-dimensional feature space. The construction process of the parameterized model includes: setting a plurality of frequency level basis functions, the frequency range is determined according to the scale of the geological feature variation revealed by the continuous mapping function; encoding the three-dimensional coordinates through the basis functions of different frequencies respectively; and combining the multi-frequency encoding results to form a high-dimensional position encoding vector. Therefore, the parameterized model in the present application can guarantee the capturing ability of the geological features at different scales from macro to micro.

[0044] Step B3: performing position encoding on each three-dimensional point in the point cloud matrix based on the parameterized model to obtain an encoded point cloud.

[0045] Specifically, the three-dimensional coordinates (x, y, z) of each spatial point in the point cloud matrix are input into the parameterized model for continuously representing geological features, and the parameterized model calculates a high-dimensional position encoding vector corresponding to each three-dimensional coordinate through multi-frequency basis functions.

[0046] The calculation process of the multi-frequency basis function is as follows: the parameterized model pre-sets N different frequency values, the frequency values increase exponentially from low to high, covering spatial scales from meters to millimeters; for each frequency value, the parameterized model multiplies the X, Y, and Z components of the three-dimensional coordinates by the frequency value to obtain modulated coordinate values; the modulated coordinate values are input into a sine function and a cosine function respectively to generate periodic encoding; and the sine and cosine encoding results at all frequencies are arranged in order to form a high-dimensional vector. For example, for a single coordinate component x and a frequency f, the generated encoding includes two values sin(fx) and cos(fx); when there are N frequencies, a single coordinate component generates 2N encoding values; three coordinate components generate a total of 6N encoding values, which constitute a high-dimensional position encoding vector.

[0047] The parameterized model combines the high-dimensional position encoding vector with the multi-dimensional attribute information of the original spatial point to form an enhanced feature representation and obtain an encoded point cloud.

[0048] It should be noted that each spatial point in the encoded point cloud includes two parts of information: one is the high-dimensional position encoding vector generated by the parameterized model, which captures the multi-scale features of the spatial position; and the other is the original multi-dimensional attribute information, which retains the actual measured values of the dynamic working parameters of the shield machine. Preferably, the position encoding enhances the representation ability of the spatial point to the geological detail changes, and provides sufficient feature basis for the high-frequency mapping processing of step B4.

[0049] Step B4: Perform high-frequency mapping processing on the three-dimensional coordinates in the encoded point cloud to generate high-dimensional position features.

[0050] Specifically, the high-frequency mapping processing is to extract the high-dimensional position encoding vector generated in step B3 for each spatial point in the encoded point cloud, and enhance the expression of high-frequency components through further nonlinear transformation. Specifically, the high-frequency mapping processing includes: inputting the high-dimensional position encoding vector into a multi-layer perception structure, each layer of the multi-layer perception structure containing a linear transformation and a nonlinear activation function; wherein the nonlinear activation function amplifies the high-frequency components in the encoding vector and suppresses the low-frequency components; through residual connection, the original encoding information is retained while the detailed features are enhanced; and the output is the high-dimensional position features with expanded dimensions.

[0051] It should be noted that the high-dimensional position features are spatial feature representations further extracted and enhanced through nonlinear transformation based on the high-dimensional position encoding vector in step B3. Compared with the original position encoding vector, the high-dimensional position features have stronger high-frequency detail expression capability and can more accurately depict the local changes and boundary information of the geological features. It should be noted that the generated high-dimensional position features will form a standardized geological feature vector set together with the multi-dimensional attribute information.

[0052] S3: Fuse the standardized geological feature vector set and the shield machine dynamic working parameters to generate a multi-dimensional spatio-temporal feature tensor, and perform correlation analysis of geological subsidence trend and equipment parameters based on the multi-dimensional spatio-temporal feature tensor to generate a surface subsidence prediction value.

[0053] In the present application, the step of generating a surface subsidence prediction value includes the following steps C1 to C4: Step C1: Perform time sequence alignment on the standardized geological feature vector set and the shield machine dynamic working parameters to obtain a joint feature sequence.

[0054] Specifically, by identifying the time stamps of the standardized geological feature vector set and the time stamps of the shield machine dynamic working parameters, the two are adjusted to a unified time grid based on time sequence matching, ensuring that the standardized geological feature vector and the shield machine dynamic working parameter value corresponding to each time point exist, so as to combine into a joint feature sequence in chronological order.

[0055] Step C2: Perform spatial reorganization processing on the joint feature sequence to generate a multi-dimensional spatio-temporal feature tensor; Specifically, the spatial reorganization processing is performed on the joint feature sequence, and by identifying the original spatial point positions corresponding to the standardized geological feature vectors in the joint feature sequence, the standardized geological feature vectors of each time step are distributed to the corresponding grid positions in the spatial range of the point cloud matrix, the feature values of different time steps at the same spatial position are aggregated in each grid element, and a four-dimensional tensor structure is organized along the time dimension to generate a multi-dimensional spatio-temporal feature tensor.

[0056] Step C3: performing eigen decomposition on the multi-dimensional space-time feature tensor to obtain a long-term trend tensor of geological settlement and a device parameter tensor.

[0057] Specifically, the eigen decomposition method is applied to the multi-dimensional space-time feature tensor, and the multi-dimensional space-time feature tensor is decomposed into a series of feature components, from which a component representing the slow change process of geological settlement is identified and separated as the long-term trend tensor of geological settlement, and a component reflecting the change mode of the dynamic working parameters of the shield machine is extracted as the device parameter tensor.

[0058] It should be noted that the eigen decomposition method is a mathematical technique for decomposing the multi-dimensional space-time feature tensor into eigenvectors and eigenvalues, and the long-term trend tensor of geological settlement and the device parameter tensor are extracted to facilitate correlation analysis. The component of the slow change process of geological settlement is the long-term trend tensor of geological settlement in the slow change process of geological settlement, which is extracted from the multi-dimensional space-time feature tensor by eigen decomposition, representing the long-term change trend of geological settlement.

[0059] Step C4: obtaining the surface settlement prediction value based on the correlation calculation of the long-term trend tensor and the device parameter tensor.

[0060] Specifically, by calculating the covariance matrix or similarity index between the long-term trend tensor and the device parameter tensor, the current value of the device parameter tensor is applied as a weight factor to the corresponding element of the long-term trend tensor through the linear dependence relationship between the geological settlement trend and the device operating state, and the long-term trend tensor and the device parameter tensor are combined linearly or multiplied by a scalar to derive the surface settlement prediction value.

[0061] It should be noted that in the present application, the linear dependence relationship refers to the linear relationship between two or more variables, in which the change of one variable can be predicted by the linear function of another variable, and the linear dependence relationship is reflected in the correlation calculation of the long-term trend tensor and the device parameter tensor, which is used to quantify the linear influence degree between the geological settlement and the device parameters. In the present application, linear combination or scalar multiplication is a linear relationship between the geological settlement trend and the device parameters.

[0062] S4, constructing a physical simulation deviation model, inputting the surface settlement prediction value into the physical simulation deviation model, outputting the shield tunnel parameters, and generating a deviation analysis report based on the shield tunnel parameters.

[0063] In the present application, the step of generating the deviation analysis report includes the following steps D1 to D5: Step D1: integrating the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters based on the interaction between the shield tunnel axis geometric parameters and the stratum to obtain a physical simulation basic framework.

[0064] Specifically, based on the interaction between the shield tunnel axis geometric parameters and the stratum, the coordinate system method is used to align the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters to the same spatial reference system, and through spatial meshing and reorganization to generate structured data, and the mapping relationship framework is used to establish the correlation between parameters to obtain the physical simulation basic framework.

[0065] It should be noted that the coordinate system method is to unify the tunneling parameter sequence and the geological correlation parameters, position and attitude parameters, running state parameters and environmental and auxiliary parameters in the shield machine dynamic working parameters into the same coordinate system through coordinate conversion or alignment operation, thereby generating consistent tunneling state data; the segment assembly parameters refer to the key parameters in the segment installation process in the shield tunnel construction, including the position coordinates, installation angle, assembly sequence and joint state of the segment; the soil mechanical parameters refer to the parameters describing the mechanical properties of the soil, including shear strength, compression modulus and permeability coefficient, etc., which are used to characterize the mechanical behavior of the stratum in the shield tunnel construction process and affect the interaction simulation between the stratum and the tunnel; in the same spatial reference system, all coordinates and data are based on the unified geometric coordinate system and measurement standard, ensuring the consistency and comparability of the shield machine dynamic working parameters, geological features and tunnel design parameters in spatial position, thereby accurately integrating and analyzing multi-dimensional information.

[0066] Step D2: Obtain historical construction deviation data, and perform correction on the segment stress and deformation parameters of the physical simulation basic framework according to the historical construction deviation data and the multi-working-condition mechanical response relationship to obtain a physical simulation deviation model.

[0067] Specifically, after obtaining the historical construction deviation data, the mechanical behavior under different construction conditions is analyzed by using the multi-working-condition mechanical response relationship, and by comparing the historical construction deviation data with the output results of the physical simulation basic framework, the segment stress and deformation parameters of the physical simulation basic framework are optimized and adjusted to make the simulation output match the historical construction deviation data, thereby obtaining the physical simulation deviation model.

[0068] It should be noted that in the present application, the multi-condition mechanical response relationship refers to the mechanical behavior correlation between the force and deformation of the shield tunnel segment under various construction conditions, different geological conditions, excavation parameters and environmental factors. The relationship is established through historical construction deviation data analysis and mechanical simulation, and is used to correct the parameters in the physical simulation model. In the present application, the process of comparing the historical construction deviation data with the physical simulation basic framework is to compare the actual deviation values at each spatial point with the simulation deviation values calculated by the physical simulation basic framework under the same input conditions, and to complete the comparison by calculating the distance between the actual deviation values and the simulation deviation values at each spatial point. Through the multi-condition mechanical response relationship, the force and deformation parameters of the shield tunnel segment in the physical simulation basic framework are mathematically optimized and numerically adjusted, so that the physical simulation deviation model can accurately reflect the mechanical response behavior under various actual construction conditions.

[0069] Step D3: inputting the ground settlement prediction value into the physical simulation deviation model, performing local deformation correction on the shield tunnel spatial coordinates based on the physical simulation deviation model, and obtaining the shield tunnel parameters.

[0070] Specifically, the ground settlement prediction value is input into the physical simulation deviation model, and based on the physical simulation deviation model, the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters are integrated, combined with the historical construction deviation data and the multi-condition mechanical response relationship, to perform local deformation correction on the shield tunnel spatial coordinates, and the influence of the ground settlement prediction value is reflected through coordinate adjustment, to obtain the shield tunnel parameters.

[0071] It should be noted that the multi-condition mechanical response relationship refers to the response relationship between the mechanical parameters such as the force and deformation of the shield tunnel segment and the geological conditions, equipment parameters and other factors in the shield tunnel construction under various construction conditions. The relationship is established based on the historical construction deviation data and is used to correct the force and deformation parameters of the physical simulation basic framework; the local deformation correction is a process of applying local deformation adjustment to the shield tunnel spatial coordinates by the physical simulation deviation model, and the small deformation caused by geological settlement or mechanical response is corrected, so as to output more accurate shield tunnel parameters.

[0072] Step D4: performing global scanning on the shield tunnel parameters to generate an original deviation data set.

[0073] Specifically, when performing global scanning on the shield tunnel parameters, each data point of the tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters contained in the shield tunnel parameters is respectively calculated with the corresponding tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters in the physical simulation basic framework to obtain the deviation value of each data point, and all the deviation values are collected to generate an original deviation data set.

[0074] It should be noted that the point-by-point difference calculation refers to calculating the difference between each spatial point or data point of the shield tunnel parameters and the tunnel design axis coordinate or expected parameter value during the generation of the original deviation data set, thereby quantifying the deviation value of each point.

[0075] Step D5: performing multi-dimensional correlation analysis on the original deviation data set, and generating a deviation analysis report based on the analysis results of the multi-dimensional correlation analysis.

[0076] Specifically, multi-dimensional correlation analysis is performed on the deviation values of the shield tunnel parameters in the original deviation data set, and the analysis content includes: identifying the correlation between the tunnel design axis coordinate deviation, the segment assembly parameter deviation, and the soil mechanical parameter deviation; grouping data points with similar deviation characteristics through clustering analysis, the clustering analysis adopts the K-means algorithm, and the data points are divided into several categories according to the Euclidean distance of the deviation values; each cluster center represents a typical deviation characteristic, and the clustering result reveals the internal distribution structure of the deviation data; the association rule mining method is used to find the association relationship between different deviation modes, the association rule mining analyzes the frequently occurring mode combinations in the deviation data set, identifies the probability of the occurrence of another deviation mode when a deviation mode occurs, for example, when the horizontal deviation of the tunnel axis exceeds a pre-set horizontal deviation threshold (e.g., 50 mm), the probability that the segment assembly angle deviation also exceeds a pre-set angle deviation threshold (e.g., 2 degrees) reaches a set confidence level (e.g., 85%), and an association rule is established between the two, thereby identifying the potential causes and propagation paths of the deviation; the above correlation identification results are used as the main deviation characteristic description, the internal distribution structure revealed by the clustering analysis is used as the deviation spatial distribution rule, the association relationship obtained by the association rule mining is used as the parameter correlation evaluation result, and the identified potential causes and propagation paths are used as the deviation cause analysis, which are integrated to form a deviation analysis report.

[0077] S5: Taking the deviation values of the shield tunnel parameters as the optimization target, adjusting the model parameters of the physical simulation deviation model, and generating adjusted model parameters.

[0078] Specifically, based on the shield tunnel parameters and the deviation analysis report output in step S4, the deviation values of each spatial point and the tunnel design axis coordinate in the shield tunnel parameters are extracted as the optimization target, and the segment force parameters and deformation parameters in the physical simulation deviation model are adjusted and optimized through step-by-step training. The training process adopts the gradient descent method, takes the minimization of the total deviation value as the objective function, iteratively calculates the optimal adjustment amount of each model parameter, updates the parameter value according to the deviation gradient and the set learning rate each time, and after multiple iterations until the deviation value converges to a pre-set threshold, the adjusted model parameters are finally output.

[0079] It should be noted that the step-by-step training refers to decomposing the model parameter adjustment process into multiple training steps, each step optimizing different categories of parameters. For example, the first step optimizes segment force parameters, including segment axial force coefficients, bending moment coefficients, and shear force coefficients; the second step optimizes deformation parameters, including segment radial deformation coefficients, longitudinal deformation coefficients, and joint rotation angle coefficients; and the third step optimizes soil reaction force parameters, including foundation reaction force coefficients and lateral soil pressure coefficients. Through step-by-step training, the present application first adjusts the segment force parameters that have a greater impact on deviation, then adjusts the deformation-related parameters, and finally fine-tunes the soil reaction force parameters, improving training efficiency and convergence speed. The adjusted model parameters include updated segment force coefficients, deformation modulus, soil reaction force coefficients, and other key physical simulation parameters.

[0080] S6: applying the adjusted model parameters to the physical simulation deviation model to generate updated shield tunnel parameters through iterative processing.

[0081] In the present application, step S6 specifically includes the following steps E1 to E4: Step E1: replacing the adjusted model parameters (including segment force coefficients and deformation parameters) generated in step S5 into the original parameter values in the physical simulation deviation model, completing the update of the model parameters, and forming an updated physical simulation deviation model.

[0082] Specifically, the main deviation characteristics and parameter correlation results obtained through multi-dimensional correlation analysis are used to adjust the segment force and deformation parameters in the physical simulation deviation model, and the gradient descent method is used to minimize the deviation value between the shield tunnel parameters and the tunnel design axis coordinates, thereby generating adjusted model parameters. These adjusted model parameters are updated to the physical simulation deviation model to form an updated physical simulation deviation model.

[0083] Step E2: based on the updated physical simulation deviation model, performing deviation calculation on the shield tunnel parameters to obtain a deviation value.

[0084] Specifically, according to the updated physical simulation deviation model, the corrected segment force parameters and deformation parameters are used to perform deviation verification calculation on the shield tunnel parameters generated in step S4. The deviation calculation process is as follows: extracting the three-dimensional coordinates of each spatial point in the shield tunnel parameters, including the actual position coordinates of the tunnel axis, the actual assembly position coordinates of the segments, and the actual coordinates after deformation due to geological settlement; comparing each actual coordinate with the theoretical target coordinate in the tunnel design axis coordinate, calculating the Euclidean distance between each spatial point's actual coordinate and the corresponding theoretical target coordinate, i.e. the straight-line distance between two points in three-dimensional space; summing up the distance values of all spatial points to obtain a deviation value representing the overall deviation degree. The larger the deviation value, the greater the deviation of the shield tunnel parameters from the design requirements, and further optimization and adjustment are needed.

[0085] Step E3: performing gradient adjustment on the shield tunnel parameters based on the deviation value, generating an objective function of the shield tunnel parameters, and performing physical simulation deviation calculation by finite difference method to obtain the gradient direction of the shield tunnel parameters.

[0086] Specifically, the objective function of the shield tunnel parameters is defined as minimizing the deviation value, then a small perturbation is performed on each component of the shield tunnel parameters using the finite difference method, and the perturbed shield tunnel parameters are input into the physical simulation deviation model to recalculate the deviation value. By comparing the change of the deviation value before and after the perturbation, the partial derivative of the objective function with respect to the shield tunnel parameters is estimated, and thus the gradient direction of the shield tunnel parameters is obtained.

[0087] It should be noted that in the present application, the objective function is a mathematical expression centered on minimizing the deviation value, and the gradient direction of the shield tunnel parameters with respect to the deviation value is calculated by the finite difference method to guide parameter updating to approximate the tunnel design axis coordinates. The finite difference method is a numerical method for approximating derivatives by calculating the difference in function values under a small change, which estimates the gradient direction by changing the input parameters by a small step and observing the output change.

[0088] Step E4: performing iterative updating of the shield tunnel parameters based on the gradient direction of the shield tunnel parameters to generate updated shield tunnel parameters.

[0089] Specifically, according to the gradient direction obtained in step E3, the shield tunnel parameters are updated according to a set step size (e.g. 0.01), and after each update, steps E2 and E3 are re-executed to calculate a new deviation value and gradient direction. The iterative process is repeated until the deviation value converges to an acceptable range (e.g. less than 10 mm) or the maximum number of iterations (e.g. 100 times) is reached, and finally the updated shield tunnel parameters are output.

[0090] The present embodiment also provides a shield tunnel parameter prediction system based on deep learning, comprising: a parameter processing module for obtaining dynamic working parameters of a shield machine and generating a point cloud matrix; a feature encoding module for performing position encoding and mapping processing on three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix to generate a standardized geological feature vector set; a prediction analysis module for fusing the standardized geological feature vector set and the dynamic working parameters of the shield machine, generating a multi-dimensional spatio-temporal feature tensor, and performing correlation analysis of geological settlement trend and equipment parameters based on the multi-dimensional spatio-temporal feature tensor to generate a surface settlement prediction value; The model analysis module is configured to construct a physical simulation deviation model, input the ground surface settlement prediction value into the physical simulation deviation model, generate a shield tunnel parameter, and generate a deviation analysis report based on the shield tunnel parameter. The parameter calibration module is configured to adjust model parameters of the physical simulation deviation model through step-by-step training processing according to the shield tunnel parameter and the deviation analysis report, and generate adjusted model parameters. The parameter updating module is configured to apply the adjusted model parameters to the physical simulation deviation model, and generate updated shield tunnel parameters through iterative processing.

[0091] The embodiment also provides a computer device suitable for the shield tunnel parameter prediction method based on deep learning, which comprises a memory and a processor.

[0092] The computer device can be a terminal, and the computer device comprises a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for running the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to perform wired or wireless communication with an external terminal. The wireless communication can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, a trackball or a touchpad arranged on the shell of the computer device. In addition, the input device can also be an external keyboard, a touchpad or a mouse, etc.

[0093] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the method for predicting parameters of a shield tunnel based on deep learning proposed in the above embodiment; and the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk.

[0094] To sum up, the method for predicting parameters of a shield tunnel based on deep learning has the following advantages: the fusion of the standardized geological feature vector set and the dynamic working parameters of the shield machine generates a multi-dimensional space-time feature tensor, thereby realizing high-dimensional space-time correlation representation of geological properties and equipment states, and improving the accuracy and reliability of the settlement trend prediction; the construction and iterative optimization of the physical simulation deviation model realize dynamic correction and deviation control of the parameters of the shield tunnel, and enhance the adaptability and decision support capability of the system under complex geological conditions.

[0095] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A deep learning-based method for predicting parameters of a shield tunnel, characterized in that: The application relates to a shield tunneling parameter optimization method and device. Collect shield machine dynamic working parameters and generate a point cloud matrix; Perform position coding and mapping processing on three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix to generate a standardized geological feature vector set; Fuse the standardized geological feature vector set and the shield machine dynamic working parameters to generate a multi-dimensional space-time feature tensor, and perform correlation analysis on geological subsidence trends and equipment parameters based on the multi-dimensional space-time feature tensor to generate a surface subsidence prediction value; Construct a physical simulation deviation model, input the surface subsidence prediction value into the physical simulation deviation model, output shield tunnel parameters, and generate a deviation analysis report based on the shield tunnel parameters; Take the deviation value of the shield tunnel parameters as an optimization target, adjust model parameters of the physical simulation deviation model, and generate adjusted model parameters; Apply the adjusted model parameters to the physical simulation deviation model, and generate updated shield tunnel parameters through iterative processing. 2.The deep learning-based shield tunnel parameter prediction method of claim 1, wherein: The point cloud matrix is generated through the following specific steps, Perform time alignment on tunneling parameters in the shield machine dynamic working parameters to obtain a tunneling parameter sequence; Perform coordinate system alignment on the tunneling parameter sequence, geological correlation parameters, position and attitude parameters, running state parameters, and environment and auxiliary parameters in the shield machine dynamic working parameters to generate tunneling state data; Perform spatial grid reorganization on the tunneling state data to generate the point cloud matrix. 3.The deep learning-based shield tunnel parameter prediction method of claim 2, wherein: The position coding and mapping processing on the three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix are performed through the following specific steps, Based on the spatial distribution of the point cloud matrix, a mapping relationship framework of coordinates to geological features is established, and a continuous mapping function of spatial points to geological attribute values is established; According to the continuous mapping function, the high-frequency response range of the position coding is expanded, and a parameterized model for continuously representing geological features is constructed; Based on the parameterized model, position coding is performed on each three-dimensional point in the point cloud matrix to obtain coded point clouds; High-frequency mapping processing is performed on the three-dimensional coordinates in the coded point clouds to generate high-dimensional position features.

4. The deep learning-based shield tunnel parameter prediction method of claim 3, wherein: The fusion of the standardized geological feature vector set and the shield machine dynamic working parameters to generate a multi-dimensional space-time feature tensor is performed through the following specific steps, Perform time sequence alignment on the standardized geological feature vector set and the shield machine dynamic working parameters to obtain a joint feature sequence; Perform spatial reorganization processing on the joint feature sequence to generate the multi-dimensional space-time feature tensor.

5. The deep learning-based shield tunnel parameter prediction method of claim 4, wherein: The correlation analysis on geological subsidence trends and equipment parameters based on the multi-dimensional space-time feature tensor to generate a surface subsidence prediction value is performed through the following specific steps, Perform feature decomposition on the multi-dimensional space-time feature tensor to obtain a long-term trend tensor of geological subsidence and an equipment parameter tensor; Based on the correlation calculation of the long-term trend tensor and the equipment parameter tensor, the surface subsidence prediction value is obtained. 6.The deep learning-based shield tunnel parameter prediction method of claim 5, wherein: The construction of the physical simulation deviation model includes the following steps, Integrate tunnel design axis coordinates, segment assembly parameters and soil mechanical parameters based on shield tunnel axis geometric parameters and stratum interaction to obtain a physical simulation basic framework; Obtain historical construction deviation data, and perform correction on pipe segment stress and deformation parameters of the physical simulation foundation framework according to the historical construction deviation data and multi-working-condition mechanical response relationship, to obtain the physical simulation deviation model.

7. The deep learning-based shield tunnel parameter prediction method of claim 6, wherein: Input the ground surface settlement prediction value into the physical simulation deviation model, output the shield tunnel parameters, and generate a deviation analysis report based on the shield tunnel parameters, the specific steps being as follows, Input the ground surface settlement prediction value into the physical simulation deviation model, perform local deformation correction on shield tunnel space coordinates based on the physical simulation deviation model, and obtain the shield tunnel parameters; Perform global scanning on the shield tunnel parameters, and generate an original deviation data set; Perform multi-dimensional correlation analysis on the original deviation data set, and generate the deviation analysis report based on analysis results of the multi-dimensional correlation analysis. 8.The deep learning-based shield tunnel parameter prediction method of claim 7, wherein: The analysis results include main deviation characteristics, deviation spatial distribution rules, parameter correlation evaluation, and deviation cause analysis. 9.The deep learning-based shield tunnel parameter prediction method of claim 8, wherein: The specific steps of applying the adjusted model parameters to the physical simulation deviation model and generating updated shield tunnel parameters through iterative processing are as follows: Update the adjusted model parameters to the physical simulation deviation model to obtain updated physical simulation deviation model; Perform deviation calculation on the shield tunnel parameters based on the updated physical simulation deviation model to obtain a deviation value; Perform gradient adjustment on the shield tunnel parameters based on the deviation value, generate a target function of the shield tunnel parameters, and perform physical simulation deviation calculation through finite difference method to obtain a gradient direction of the shield tunnel parameters; Perform iterative update on the shield tunnel parameters based on the gradient direction of the shield tunnel parameters to generate the updated shield tunnel parameters. 10.A shield tunnel parameter prediction system based on deep learning, based on the shield tunnel parameter prediction method based on deep learning of any one of claims 1-9, characterized in that: The specific steps of applying the adjusted model parameters to the physical simulation deviation model and generating updated shield tunnel parameters through iterative processing are as follows: The parameter processing module is configured to obtain shield machine dynamic working parameters and generate a point cloud matrix. The feature encoding module is configured to perform position encoding and mapping processing on three-dimensional coordinates and multi-dimensional attribute information in the point cloud matrix to generate a standardized geological feature vector set. The prediction analysis module is configured to fuse the standardized geological feature vector set and the shield machine dynamic working parameters, generate a multi-dimensional space-time feature tensor, and perform correlation analysis on geological settlement trend and equipment parameters based on the multi-dimensional space-time feature tensor to generate a ground surface settlement prediction value. The model analysis module is configured to construct a physical simulation deviation model, input the ground surface settlement prediction value into the physical simulation deviation model, generate shield tunnel parameters, and generate a deviation analysis report based on the shield tunnel parameters. The parameter calibration module is configured to adjust model parameters of the physical simulation deviation model with a deviation value of the shield tunnel parameters as an optimization target to generate adjusted model parameters. The parameter update module is configured to apply the adjusted model parameters to the physical simulation deviation model and generate updated shield tunnel parameters through iterative processing.

Images