Tunnel grouting simulation method and system based on conditional generative adversarial network

Abstract

The present invention discloses a tunnel grouting simulation method and system based on a conditional generative adversarial network, belonging to the field of geotechnical engineering technology. The method comprises the following steps: obtaining working condition information of the tunnel area to be grouted, using the working condition information as conditional variables of a conditional generative adversarial network, inputting random noise and the conditional variables into a generator, generating grouting simulation data suitable for the complexity of the current working condition, and constructing a grouting simulation dataset; gridding the area to be grouted, and constructing a physical mathematical model of the entire grouting process based on discretized continuity equations and momentum equations based on the grouting simulation dataset, and setting boundary conditions and initial conditions for each grid area; based on the constructed physical data model, predicting the grouting pressure, grouting velocity, and slurry diffusion morphology at the next time based on the current working condition information, and continuously updating the model to complete the simulation of the tunnel grouting process. The present invention can achieve accurate simulation of the tunnel grouting process.

Description

Technical Field

[0001] The present invention belongs to the technical field of geotechnical engineering, and in particular relates to a tunnel grouting simulation method and system based on a conditional generative adversarial network. Background Art

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

[0003] The construction and implementation of tunnels need to consider the impact of complex geological conditions, groundwater level changes, and the surrounding environment on project safety. Tunnel construction often encounters sudden water disasters, especially in complex water-rich sand and gravel formations. The timely identification and effective response to sudden water is more difficult, which brings huge safety hazards and risks to subsequent construction. Grouting is an effective means of reinforcing strata and controlling sudden water. Grouting numerical simulation refers to the use of basic principles such as fluid mechanics, solid mechanics, and porous media theory to treat rock and soil as porous media and slurry as fluid or a mixture of fluid and solid, and simulate the grouting process through computer calculations. However, due to the hidden nature of underground projects, traditional grouting simulation methods have problems such as large errors between simulation and actual projects, long simulation calculation time, and complex on-site parameters. There are great difficulties in actual calculations, making it difficult to effectively and accurately simulate the grouting process.

[0004] Existing grouting simulation methods offer certain simulation convenience and high solution efficiency. However, in actual projects, the grouting site often faces problems such as underground undercurrents, weak and easily broken surrounding rock, discrepancies with actual engineering plans, and changes in groundwater pressure and temperature. This results in differences between the simulated conditions and the actual conditions during slurry injection. Furthermore, due to the complex characteristics of fluid flow in geological media, such as nonlinearity, heterogeneity, and uncertainty, there is an urgent need for high-precision simulation data. However, in actual on-site projects, due to the complex working conditions, the only condition that is easiest to fully obtain is often pressure. Factors such as slurry flow rate and slurry diffusion shape are often difficult to accurately and fully obtain due to turbulent flow and complex geological conditions. The amount of data obtained is scarce and the data format is single, which limits the model's ability to capture complex physical phenomena, thereby affecting the accuracy of the model's calculation results and generalization performance. Summary of the Invention

[0005] In order to address the deficiencies of the above-mentioned prior art, the present invention provides a tunnel grouting simulation method and system based on a conditional generative adversarial network. According to the actual working conditions of tunnel construction, the conditional generative adversarial network is used to generate grouting simulation data close to the actual working conditions. While solving the problems of insufficient data volume and single data format, the generated simulation data is highly consistent with the actual working conditions. The generated grouting simulation data is then used to construct a physical and mathematical model of the entire grouting process, thereby achieving a comprehensive and accurate simulation of the entire tunnel grouting process.

[0006] In a first aspect, the present invention provides a tunnel grouting simulation method based on a conditional generative adversarial network.

[0007] A tunnel grouting simulation method based on conditional generative adversarial networks, comprising:

[0008] Obtain the working condition information of the tunnel area to be grouting, use this working condition information as the conditional variables of the generative adversarial network, input random noise and the conditional variables into the generator, and generate grouting simulation data suitable for the complexity of the current working condition, thereby constructing a grouting simulation dataset;

[0009] The grouting area is divided into grids. For each grid area, a physical mathematical model of the entire grouting process is constructed based on the discretized continuity equation and momentum equation according to the grouting simulation data set, and the boundary conditions and initial conditions of each grid area are set.

[0010] Based on the physical mathematical model of the entire grouting process in each grid area, as well as the boundary conditions and initial conditions, and according to the working conditions at the current moment, the grouting pressure, grouting speed, and slurry diffusion form at the next time are predicted. After continuous updating, the simulation of the tunnel grouting process is completed.

[0011] According to a further technical solution, the working condition information includes actual geological conditions and grouting measured data, and the grouting measured data includes grouting pressure, slurry composition, and slurry viscosity;

[0012] The grouting simulation data includes time series data of grouting pressure, grouting speed, grouting slurry composition, and grouting slurry viscosity during the grouting process, as well as effect diagrams of slurry diffusion path and flow field distribution after grouting.

[0013] In a second aspect, the present invention provides a tunnel grouting simulation system based on a conditional generative adversarial network.

[0014] A tunnel grouting simulation system based on conditional generative adversarial networks, comprising:

[0015] The simulation data generation module is used to obtain the working condition information of the tunnel area to be grouting, use the working condition information as the condition to generate the conditional variables of the adversarial network, input random noise and conditional variables into the generator, and generate grouting simulation data suitable for the complexity of the current working condition, thereby constructing a grouting simulation dataset;

[0016] The simulation model construction module is used to divide the grouting area into grids. For each grid area after division, according to the grouting simulation data set, a physical mathematical model of the entire grouting process is constructed based on the discretized continuity equation and momentum equation, and the boundary conditions and initial conditions of each grid area are set;

[0017] The grouting simulation module is used to predict the grouting pressure, grouting speed, and slurry diffusion form at the next time based on the physical and mathematical model of the entire grouting process, as well as the boundary conditions and initial conditions of each grid area, according to the current working conditions information. After continuous updating, the simulation of the tunnel grouting process is completed.

[0018] In a third aspect, the present invention also provides an electronic device, comprising: a memory for storing executable instructions; and a processor for implementing the above-mentioned tunnel grouting simulation method based on conditional generative adversarial network when executing the executable instructions stored in the memory.

[0019] In a fourth aspect, the present invention further provides a computer-readable storage medium storing executable instructions for causing a processor to execute the executable instructions to implement the above-mentioned tunnel grouting simulation method based on conditional generative adversarial network.

[0020] In a fifth aspect, the present invention also provides a computer program product, which includes executable instructions, and the executable instructions are stored in a computer-readable storage medium; wherein, when the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the above-mentioned tunnel grouting simulation method based on the conditional generative adversarial network is implemented.

[0021] One or more of the above technical solutions have the following beneficial effects:

[0022] 1. The present invention provides a tunnel grouting simulation method and system based on a conditional generative adversarial network. The conditional generative adversarial network is introduced. When all field data are not fully known, the actual working conditions of tunnel construction are used as conditional variables, and the conditional generative adversarial network is used to generate high-quality simulation data, effectively expanding the grouting simulation data set. While solving the problems of data scarcity and single data form, the generated simulation data is highly consistent with the actual working conditions. Through the generation of diversified high-quality data, a physical and mathematical model of the entire grouting process is constructed, so that the model can show stronger adaptability and generalization ability when facing different geological conditions and complex working conditions, thereby reducing the negative impact caused by data scarcity or imbalance. Finally, based on the accurately modeled physical and mathematical model of the entire grouting process, combined with known engineering information, the grouting pressure and grouting speed of each time step are accurately calculated, and these calculation results are used to realize accurate simulation of the tunnel grouting process.

[0023] 2. The present invention adopts a progressive model training strategy, taking into account the complexity of different working conditions, and gradually introduces field data, image data and video data, so as to enhance the model's ability to recognize spatial features and dynamic changes, enable the model to capture subtle changes and dynamic features in the grouting process in real time, and be able to analyze the grouting process more comprehensively, thereby improving the accuracy and reliability of the overall prediction, improving the performance of the model under complex working conditions, and providing important data support for real-time monitoring and decision-making of the project.

[0024] 3. The present invention uses a multi-layer convolutional neural network structure to conduct in-depth analysis of one-dimensional time series data and two-dimensional image data in the grouting process, which can effectively capture the time series characteristics, spatial characteristics and complex dependencies in the data. This can significantly improve the prediction accuracy of subsequent multi-dimensional data: time series data such as pressure and velocity, slurry diffusion path and flow field data, thereby providing a more reliable basis for construction decision-making.

[0025] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0027] Figure 1 This is a flowchart of the tunnel grouting simulation method based on conditional generative adversarial network described in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] It should be noted that the following detailed descriptions are exemplary only and are intended to describe specific embodiments and provide further explanation of the present invention, and are not intended to limit the exemplary embodiments according to the present invention. Unless otherwise indicated, all technical and scientific terms used herein have the same meanings as those commonly understood by those of ordinary skill in the art to which the present invention belongs. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0029] Example 1

[0030] This embodiment provides a tunnel grouting simulation method based on conditional generative adversarial networks. Figure 1 As shown, the following steps are included:

[0031] Step S1: Obtain the working condition information of the tunnel area to be grouting, use the working condition information as the condition to generate the conditional variables of the adversarial network, input random noise and conditional variables into the generator, generate grouting simulation data suitable for the complexity of the current working condition, and thus construct a grouting simulation data set.

[0032] In this embodiment, a conditional generative adversarial network (GAN) is used to train the generator and discriminator in the network, and the trained generator is used to generate high-quality grouting simulation data to expand the data set. In fact, the generated grouting simulation data is usually generated through simulation or preliminary modeling without fully knowing all the field data. These simulation data generated by the generator can provide complete preliminary working condition data for subsequent grouting simulation calculations. This simulation data set serves as the input of the subsequent grouting simulation model (i.e., the physical model built subsequently), providing initial conditions, boundary conditions, etc. for subsequent simulations, thereby optimizing the performance of the model; at the same time, it can also be used as verification data for the verification and validation of the subsequent grouting simulation model to ensure the rationality of the physical model.

[0033] Specifically, in the conditional generative adversarial network, the generator receives a low-dimensional random noise vector, which is used as the input of the generator and is gradually converted through the multi-layer neural network in the generator. In order to generate grouting simulation data under specific conditions, this embodiment also introduces conditional variables as the basis of the conditional generative adversarial network. The conditional variables are known working condition information obtained in actual engineering, including actual geological conditions and grouting measured data (such as grouting pressure, partial grouting speed, slurry composition, etc.). By substituting the grouting pressure, partial grouting speed, etc. in the actual engineering into the actual geological conditions, and using them as conditional variables, after training, the simulated grouting speed, grouting pressure, etc. are generated by the simulated conditions, that is, the grouting simulation data is obtained. Therefore, the random noise vector and the conditional variable are input into the generator, and the multi-layer fully connected network in the generator is used to extract and map features through the weights and activation functions of each layer of the network, and the low-dimensional noise vector and the conditional variable are converted into high-dimensional feature representations to capture the potential spatial and temporal characteristics of the grouting process. In order to accelerate training and improve stability, batch normalization is applied to each layer of the generator.

[0034] In this embodiment, the generator used includes the following components:

[0035] (1) Input layer: used to input a. random noise vector: a low-dimensional random vector and b. conditional variables: a set of known data under specific working conditions, including grouting pressure, grouting speed, geological conditions, etc.

[0036] (2) Concatenation layer: A concatenation layer is set at the input layer to concatenate the random noise vector and the conditional variable into a longer concatenated vector. For example, if the random noise vector is 2-dimensional and the conditional variable is 3-dimensional, the final concatenated vector will have a dimension of 2+3=5.

[0037] (3) Multi-layer fully connected network: Set up in sequence: a. Fully connected layer: The network converts the spliced ​​vector into a higher-dimensional feature representation through multiple layers of fully connected layers. Each layer of the network learns how to map the input data to a more complex feature space; b. Activation function: The output of each layer introduces nonlinear transformation through a nonlinear activation function; c. Batch normalization: Improve the training stability of the network and accelerate convergence.

[0038] Through the generator constructed above, the random noise vector and the conditional variable are input as input data to the generator, wherein the random noise vector is the random residual from the Gaussian distribution, which serves as the random source of the generation process, and the conditional variable is the known information in the specific working conditions, which can provide the background or restriction of the generated data, including the grouting pressure (unit MPa), grouting speed (unit m / s) and geological conditions (encoded as discrete values) in actual engineering; after the splicing layer, the input random noise vector and the conditional variable are spliced, and then input into the first fully connected layer of the multi-layer fully connected network. In this layer of the network, it learns how to map the original random noise vector and the conditional variable into a high-dimensional feature space; each layer of the network in the generator is sequentially processed. Features are extracted in one step, and the input data is transformed nonlinearly. The output of each layer is activated by the activation function, and batch normalization is performed after the output of each layer to accelerate training and stabilize the gradient. After the calculation of each layer, the feature dimension is gradually improved, and the network extracts more complex spatial and temporal features from the initial input data, especially integrating the different characteristics of noise and conditional variables into a common representation, and finally outputs simulation data related to the working condition information under the current actual working conditions, that is, outputs more comprehensive grouting simulation data that is similar to the real grouting data structure, including images (such as the effect diagram of the slurry diffusion path and flow field distribution after grouting) and time series data (such as the changes in grouting pressure, grouting speed, and grouting slurry composition during the grouting process).

[0039] Specifically, in the conditional generative adversarial network, the discriminator is used to evaluate the authenticity of the grouting simulation data generated by the generator. As the adversarial part of the conditional GAN, it takes data from two sources as input: one is the grouting simulation data generated by the generator, and the other is the real grouting measured data, including the time series data of grouting pressure, grouting speed, grouting concentration collected on site, or spatial images of slurry diffusion, etc. Among them, if the discriminator receives the time series data of the grouting process (such as grouting speed, grouting pressure, etc.), the discriminator will analyze and discriminate the time series characteristics of the input data; if it receives image data (such as images of slurry diffusion path, flow field distribution, etc.), the discriminator extracts the spatial features of the image through convolution operations, discriminates the authenticity of the image distribution and texture information, and thus completes the discrimination.

[0040] In this embodiment, the discriminator used includes the following components:

[0041] (1) Input layer: Input one-dimensional time series data or two-dimensional image data. For one-dimensional time series data, time series data usually represents a feature that changes over time (such as grouting pressure, grouting speed, etc.). The one-dimensional time series data is processed so that its input format is T×N, where T represents the number of time steps and N represents the number of features in each time step. For example, if the time series data contains T=100 time steps and each time step has N=3 features, then the input format is T×N=100×3; for two-dimensional image data, two-dimensional image data is usually represented as a two-dimensional pixel matrix, and the input format is H×W×C, where H and W represent the height and width of the image respectively, and C represents the number of color channels.

[0042] (2) Feature extraction layer: A convolutional neural network structure is used to extract features. Considering the difference in input form between one-dimensional time series data and image data, the core structure of the discriminator remains the same, but the design of the convolution layer is different. For time series data, a one-dimensional convolution layer is used to extract local features in the time series. The convolution kernel slides along the time axis to learn local patterns in the time series and extract the changing trend in the time series, such as the change in grouting pressure within a certain time period. For two-dimensional image data, a two-dimensional convolution layer is used to extract spatial features. The convolution operation is performed on the height and width of the image to identify local spatial patterns. For image data, a two-dimensional convolution layer is used to extract spatial features. During the image feature extraction process, a pooling layer (such as a maximum pooling layer) is also added after the convolution layer to reduce the spatial dimension of the image and retain key information.

[0043] (3) Activation layer: Set a nonlinear activation function and perform nonlinear mapping on the features output by the convolutional layer through the nonlinear activation function.

[0044] (4) Fully connected layer: The output of the feature extraction layer and the pooling layer is flattened and input into the fully connected layer. The fully connected layer learns from the extracted features and forms the final classification judgment, outputting the probability of whether it is real data.

[0045] (5) Output layer: The output layer uses the Sigmoid activation function to map the output of the fully connected layer to a single numerical value, generating a probability value between 0 and 1, which indicates the possibility that the input data is real data. If the value is close to 1, the discriminator considers the data to be real; if it is close to 0, it is judged to be data generated by the generator.

[0046] The output layer and fully connected layer of the above discriminator are universal. Regardless of whether the input is one-dimensional time series data or two-dimensional image data, after passing through the feature extraction layer and activation layer, the data is flattened through the fully connected layer for final classification.

[0047] Furthermore, during the training process of the aforementioned conditional generative adversarial network, the generator and discriminator are jointly optimized through adversarial training. Specifically, the working condition information is used as the conditional information for the conditional generative adversarial network, and the random noise vector and conditional information are input into the generator to generate grouting simulation data. The generator is then evaluated and trained using the discriminator in combination with the actual grouting data. The goal of the generator is to generate data that the discriminator cannot judge, causing it to mistakenly judge the generated data as real data; the goal of the discriminator is to judge the authenticity of the data as accurately as possible. As training progresses, the discriminator's recognition effect improves, and it is able to distinguish between generated data and real data. The generator is also continuously optimized, and by continuously generating more realistic data to train the discriminator, the quality of the generated data is continuously improved.

[0048] Through adversarial training with the discriminator, the generator receives feedback from the discriminator on the generated data. This feedback helps the generator continuously adjust its simulated data, reducing the discrepancy between the generated data and the real data. The gradual convergence of adversarial training ensures the accuracy of the generated data. Through adversarial learning with the discriminator, the generator gradually improves its ability to generate reliable simulated data under different grouting conditions and geological conditions.

[0049] As an implementation method, the complexity of the working conditions is judged according to the working condition information, and based on the different levels of complexity of the working conditions, a multi-dimensional data progressive model training strategy is adopted. The working condition information (including geological conditions and grouting measured data: grouting pressure, grouting speed, etc.) is used to train the conditional generative adversarial network, and the trained generator is used to generate field data, image data or video data suitable for the current working conditions.

[0050] Specifically, based on the complexity of the working conditions, different dimensions of grouting measured data are used to train the conditional generative adversarial network. The higher the complexity of the working conditions, the higher the dimension of the grouting measured data used for training, including:

[0051] (1) For working conditions where the conditions during grouting are relatively simple, the model is relatively intuitive and easy to simulate, that is, the geological conditions, grouting operations and slurry properties are relatively simple, the data fluctuations are small, and the data changes are easier to predict, the specific conditions are:

[0052] Based on engineering information, a generator generates one-dimensional simulation data such as grouting pressure, grouting speed, slurry viscosity and grouting density for tunnel grouting. The initialization of the segment data is set through segment processing, and the one-dimensional simulation data is converted into directly usable segment data. The segment data refers to the data form of one-dimensional time series data that can be directly used. Subsequently, a discriminator is used to evaluate the authenticity of the generated data in combination with actual measurement data. Based on the discrete simulation training of the finite element method, a preliminary training simulation data set is generated.

[0053] In fact, the simulation data sets generated by the generator are not directly from real data collected on-site, but virtual data obtained through the simulation process, which are used to simulate different working conditions. These simulation data sets will be used for the subsequent multi-dimensional training of the grouting simulation model to improve the model's recognition ability when facing complex working conditions.

[0054] (2) Under working conditions where the interaction of multiple factors such as geological conditions, slurry properties, and grouting operations during the grouting process causes the system behavior to become complex, nonlinear, and difficult to predict, specifically:

[0055] Based on engineering information, the complexity of the engineering situation is judged. When the complexity of the working conditions increases, image data is gradually introduced. Combined with multi-dimensional field engineering data, on the basis of the field data obtained by training under the above simple working conditions, the generator generates data such as the slurry diffusion morphology, slurry viscosity, constant volume specific heat capacity and thermal conductivity of tunnel grouting. The image data setting is initialized through image processing, and the two-dimensional image data is converted into directly usable image data, thereby enhancing the model's ability to recognize spatial features.

[0056] (3) For the most complex working conditions, specifically:

[0057] Based on the engineering data, the complexity of the engineering situation is judged. When the working conditions are the most complex, dynamic analysis is performed in combination with video data, so that the model can capture the dynamic characteristics of the working conditions during the grouting process, thereby further improving the prediction accuracy and generalization ability of the model under complex working conditions. Specifically:

[0058] According to the engineering data, on the basis of the field data and image data under the working conditions obtained by the above training, the dynamic real-time characteristics of the two-dimensional continuous frame images (i.e., video data) are extracted and analyzed through the generator to generate data such as the boundary conditions, temperature field and initial flow field of dynamic water of the tunnel grouting, and accurate image data setting is performed through data processing.

[0059] Step S2: Divide the grouting area into grids. For each divided grid area, construct a physical mathematical model of the entire grouting process based on the discretized continuity equation and momentum equation according to the grouting simulation data set, and set the boundary conditions and initial conditions of each grid area.

[0060] In this embodiment, a physical model of the entire grouting process is established based on the grouting simulation data set generated above. The model includes typical fluid equations such as Navier-Stokes, which are used to describe the flow characteristics of the grout in the tunnel.

[0061] Specifically, first, the area to be grouted is spatially discretized, or meshed, based on engineering information. By dividing the physical space into small discrete elements, the continuous equations can be effectively transformed into discrete equations. Sufficient mesh density should be ensured in critical areas for slurry diffusion and flow to improve simulation accuracy and reliability.

[0062] Secondly, based on the grouting simulation data set, a physical model of the entire grouting process was constructed. This model includes a discretized continuity equation and a momentum equation. The momentum equation is constructed and discretized based on the grouting velocity, grouting pressure, phase fraction, and slurry viscosity. The predicted grouting velocity can be obtained by solving the momentum equation. The momentum equation is:

[0063]

[0064] Where ρ represents the density of the slurry, u is the grouting velocity, p is the grouting pressure, μ is the viscosity of the slurry, f represents the body force term (such as gravity), and t represents time; Represents the gradient, that is, the vector in the direction of maximum increment of the constant function.

[0065] Afterwards, the continuity equation is discretized using numerical methods such as the finite difference method or the finite element method. The spatial derivative is approximated using the central difference method, while the time derivative is discretized using the backward difference method. The discretized continuity equation is used to describe the mass conservation of the slurry, and its basic form is:

[0066]

[0067] Where ρ represents the density of the slurry, u is the grouting speed, and t represents time.

[0068] Finally, the discretized continuity equation and momentum equation are combined to form a coupled system, namely the physical model of the entire grouting process. Given the current grouting velocity, the discretized continuity equation is used to calculate the mass change of the slurry to ensure that the principle of mass conservation is met during the simulation. The discretized momentum equation is then used to predict the grouting velocity at the next moment.

[0069] Furthermore, after determining the model and meshing, appropriate boundary conditions and initial conditions are set based on the grouting simulation data set to simulate the actual working conditions. In this embodiment, the boundary condition setting includes applying a fixed flow rate and pressure at the grouting hole, which can be expressed as:

[0070] Q = C;

[0071] P=P inlet ;

[0072] In the above formula, Q represents the injection flow rate, the unit is volume / time (m 3 / s); C represents a constant; P represents the injection pressure, P inlet Indicates the pressure injected at the grouting hole;

[0073] At the same time, a no-flow boundary condition is set on the discrete unit boundary to simulate the actual permeability characteristics. The no-flow boundary condition is:

[0074]

[0075] In the above formula, q represents the velocity vector of the fluid, is the normal vector of the boundary surface, representing the flux of fluid through the boundary;

[0076] In addition, the initial conditions are set as the concentration distribution of the slurry before injection and the initial water content of the soil. The concentration of the slurry before injection and the initial water content of the soil can be expressed as:

[0077] C(x,y,z,t=0)=C0(x,y,z);

[0078] w(x,y,z,t=0)=w0(x,y,z);

[0079] In the above formula, C(x,y,z) represents the concentration at the spatial position (x,y,z), C0(x,y,z) is the initial concentration distribution before the start of grouting (t=0); w(x,y,z) is the soil moisture content at the position (x,y,z), and w0(x,y,z) represents the initial soil moisture content.

[0080] Step S3: Based on the physical mathematical model of the entire grouting process of each grid area, as well as the boundary conditions and initial conditions, and according to the working condition information at the current moment, the grouting pressure, grouting speed, and slurry diffusion form at the next time are predicted, and the simulation of the tunnel grouting process is completed through continuous updating.

[0081] Specifically, through the above step S1, in the case that all field data are not fully known, the conditional generative adversarial network is used to generate simulation data, so as to provide initial conditions, boundary conditions or verification data for subsequent simulations through the data generated by such simulation or preliminary modeling, thereby ensuring the rationality of the subsequent construction of the physical model; based on the generated simulation data, the physical model is constructed through the above step S2; and then step S3 is performed, on the basis of the constructed physical model, through iterative calculation, that is, through physical equations and numerical methods, based on known engineering information, the grouting pressure and grouting speed of each time step are accurately calculated, and these calculation results are used to realize accurate simulation of the grouting process.

[0082] Therefore, first, according to the grouting simulation data set, the initial grouting rate and grouting pressure are set, which can usually be set to zero or reasonable initial values ​​based on existing data.

[0083] Secondly, grouting pressure and grouting velocity are not only independent physical quantities, but also influence each other. For example, grouting velocity is affected by grouting pressure, which is determined by the flow state (velocity) of the slurry. By repeatedly iterating and updating these two parameters, the flow and diffusion process of the slurry in the soil can be better simulated. Specifically, the momentum equation established above is used, combined with parameters such as the grouting pressure and slurry viscosity at the current time step, to calculate the predicted grouting velocity. The predicted grouting velocity is input into the discretized continuity equation to solve the grouting pressure for the next time step. Based on the calculated grouting pressure, it is again substituted into the momentum equation to update the grouting velocity. Continuous iteration is performed until the preset number of iterations is reached or the convergence condition is met (such as the change in grouting velocity and grouting pressure is less than a set threshold). Each iteration obtains the grouting pressure and grouting velocity for the current time step, facilitating the calculation of subsequent time steps. After the iterative calculation is completed, the final grouting pressure and grouting velocity are recorded as the result of the current time step.

[0084] Preferably, the grouting simulation data generated based on different working condition information can be used as a verification benchmark for the grouting simulation model. For example, the prediction results of the physical model under the current working condition information can be verified based on the generated grouting simulation data to ensure the accuracy of the final simulation prediction.

[0085] Furthermore, the flow field is subjected to fluid calculations, that is, according to the established mathematical model and the discretized equations, a suitable numerical solver is selected for calculation, and the numerical values ​​of the flow field and pressure field are continuously updated through iterative calculation methods until the model converges. The discrete continuity equation is combined with the momentum equation, and the grouting pressure and grouting speed of the current time step are predicted in each stage based on the grouting speed, grouting pressure, phase fraction and slurry viscosity of the previous time step; then, the grouting speed is simulated according to the current time calculation, and the discretized phase fraction equation is constructed. The phase fraction of the current time step is obtained by solving the phase fraction equation to characterize the slurry diffusion morphology. Preferably, the data is graphically displayed using visualization tools to generate contour maps, streamline maps, slurry diffusion maps, etc., so as to intuitively present the movement trajectory of the slurry in the soil.

[0086] Example 2

[0087] This embodiment provides a tunnel grouting simulation system based on a conditional generative adversarial network, specifically including:

[0088] The simulation data generation module is used to obtain the working condition information of the tunnel area to be grouting, use the working condition information as the condition to generate the conditional variables of the adversarial network, input random noise and conditional variables into the generator, and generate grouting simulation data suitable for the complexity of the current working condition, thereby constructing a grouting simulation dataset;

[0089] The simulation model construction module is used to divide the grouting area into grids. For each grid area after division, according to the grouting simulation data set, a physical mathematical model of the entire grouting process is constructed based on the discretized continuity equation and momentum equation, and the boundary conditions and initial conditions of each grid area are set;

[0090] The grouting simulation module is used to predict the grouting pressure, grouting speed, and slurry diffusion form at the next time based on the physical and mathematical model of the entire grouting process, as well as the boundary conditions and initial conditions of each grid area, according to the current working conditions information. After continuous updating, the simulation of the tunnel grouting process is completed.

[0091] Example 3

[0092] This embodiment provides an electronic device, including: a memory for storing executable instructions; and a processor for implementing the above method provided in this embodiment when executing the executable instructions stored in the memory.

[0093] Example 4

[0094] This embodiment further provides a computer-readable storage medium storing executable instructions. When the executable instructions are executed by a processor, the processor will be caused to execute the above method provided in this embodiment.

[0095] Example 5

[0096] This embodiment provides a computer program product including executable instructions, which are computer instructions stored in a computer-readable storage medium. When a processor of an electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the electronic device performs the method provided in this embodiment.

[0097] The steps involved in the above embodiments 2 to 5 correspond to those in embodiment 1. For detailed implementation, please refer to the relevant description of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media that includes one or more instruction sets; it should also be understood to include any medium that can store, encode, or carry an instruction set for execution by a processor and cause the processor to perform any method of the present invention.

[0098] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0099] The above description is only a preferred embodiment of the present invention. Although the specific implementation of the present invention is described in conjunction with the accompanying drawings, it does not limit the scope of protection of the present invention. Those skilled in the art should understand that on the basis of the technical solution of the present invention, various modifications or variations that can be made by those skilled in the art without creative work are still within the scope of protection of the present invention.

Claims

1. A tunnel grouting simulation method based on conditional generative adversarial network, characterized in that: include: Obtain the working condition information of the tunnel area to be grouting, use this working condition information as the conditional variables of the generative adversarial network, input random noise and the conditional variables into the generator, and generate grouting simulation data suitable for the complexity of the current working condition, thereby constructing a grouting simulation dataset; The training process of the conditional generative adversarial network is as follows: The working condition information is used as the conditional information of the adversarial network. Random noise and conditional information are input into the generator to generate grouting simulation data. The discriminator is then used to evaluate and train the grouting simulation data generated by the generator in combination with the actual grouting data. Among them, according to the complexity of the working conditions, a multi-dimensional data progressive model training strategy is adopted. That is, the working condition information is combined with grouting measured data of different dimensions to train the conditional generative adversarial network. The higher the complexity of the working condition, the higher the dimension of the grouting measured data used for training, including: Under simple working conditions, a conditional generative adversarial network is trained using text data; the text data is in the form of one-dimensional time series data that can be directly used; Under medium working conditions, the text data + image data training conditional generative adversarial network is used; When working under complex conditions, the adversarial network is generated by training text data + image data + video data; The grouting area is divided into grids. For each grid area, a physical mathematical model of the entire grouting process is constructed based on the discretized continuity equation and momentum equation according to the grouting simulation data set, and the boundary conditions and initial conditions of each grid area are set. Among them, the momentum equation is constructed and discretized according to the grouting velocity, grouting pressure, phase fraction and slurry viscosity. The discretized momentum equation is: ； The discretized continuity equation is used to describe the mass conservation of the slurry, and its basic form is: ； In the above formula, ρ represents the density of the slurry, u is the grouting speed, p is the grouting pressure, μ is the viscosity of the slurry, f represents the body force term, t Indicates time; represents the gradient, that is, the vector in the direction of the maximum increment of the constant function; The boundary condition setting includes applying a fixed flow rate and pressure at the grouting hole, which are expressed as: ； ； In the above formula, Q Indicates the injection flow rate, the unit is volume / time ( ); C represents a constant; P Indicates the injection pressure, Indicates the pressure injected at the grouting hole; At the same time, a no-flow boundary condition is set on the discrete unit boundary to simulate the actual permeability characteristics. The no-flow boundary condition is: ； In the above formula, q represents the velocity vector of the fluid, is the normal vector of the boundary surface, representing the flux of fluid through the boundary; The initial conditions are set as the concentration distribution of the slurry before injection and the initial water content of the soil. The concentration before slurry injection and the initial water content of the soil are expressed as: ； ； In the above formula, Indicates the spatial position ( x,y,z ), Before grouting begins ( t =0) initial concentration distribution; is in position ( x,y,z ) at the soil moisture content, represents the initial soil moisture content; Based on the physical mathematical model of the entire grouting process in each grid area, as well as the boundary conditions and initial conditions, and according to the working conditions at the current moment, the grouting pressure, grouting speed, and slurry diffusion form at the next time are predicted. After continuous updating, the simulation of the tunnel grouting process is completed.

2. A tunnel grouting simulation method based on conditional generative adversarial network according to claim 1, characterized in that: The working condition information includes actual geological conditions and grouting measured data, and the grouting measured data includes grouting pressure, slurry composition, and slurry viscosity; The grouting simulation data includes time series data of grouting pressure, grouting speed, grouting slurry composition, and grouting slurry viscosity during the grouting process, as well as effect diagrams of slurry diffusion path and flow field distribution after grouting.

3. The tunnel grouting simulation method based on conditional generative adversarial network according to claim 1, characterized in that: Based on the physical mathematical model of the entire grouting process, the grouting pressure, grouting speed, and slurry diffusion form at the next time are predicted, including: Based on the discretized momentum equation, the grouting velocity of the next time step is predicted by combining the grouting velocity, grouting pressure, phase fraction and slurry viscosity of the current time step. Based on the discretized continuity equation, the grouting pressure at the next time step is predicted according to the predicted grouting velocity; The grouting speed and grouting pressure are updated in a continuous loop until the preset number of iterations is reached or the preset convergence condition is met. After the iterative calculation is completed, the final predicted grouting pressure and grouting speed of the next time step are obtained; the convergence condition is that the change values ​​of the grouting speed and grouting pressure are both less than the set threshold According to the predicted grouting pressure and grouting velocity, combined with the discrete phase fraction equation, the phase fraction of the next time step is solved to characterize the slurry diffusion morphology.

4. A tunnel grouting simulation system based on conditional generative adversarial network, characterized in that: include: The simulation data generation module is used to obtain the working condition information of the tunnel area to be grouting, use the working condition information as the condition to generate the conditional variables of the adversarial network, input random noise and conditional variables into the generator, and generate grouting simulation data suitable for the complexity of the current working condition, thereby constructing a grouting simulation dataset; The training process of the conditional generative adversarial network is as follows: The working condition information is used as the conditional information of the adversarial network. Random noise and conditional information are input into the generator to generate grouting simulation data. The discriminator is then used to evaluate and train the grouting simulation data generated by the generator in combination with the actual grouting data. Among them, according to the complexity of the working conditions, a multi-dimensional data progressive model training strategy is adopted. That is, the working condition information is combined with grouting measured data of different dimensions to train the conditional generative adversarial network. The higher the complexity of the working condition, the higher the dimension of the grouting measured data used for training, including: Under simple working conditions, a conditional generative adversarial network is trained using text data; the text data is in the form of one-dimensional time series data that can be directly used; Under medium working conditions, the text data + image data training conditional generative adversarial network is used; When working under complex conditions, the adversarial network is generated by training text data + image data + video data; The simulation model construction module is used to divide the grouting area into grids. For each grid area after division, according to the grouting simulation data set, a physical mathematical model of the entire grouting process is constructed based on the discretized continuity equation and momentum equation, and the boundary conditions and initial conditions of each grid area are set; Among them, the momentum equation is constructed and discretized according to the grouting velocity, grouting pressure, phase fraction and slurry viscosity. The discretized momentum equation is: ； The discretized continuity equation is used to describe the mass conservation of the slurry, and its basic form is: ； In the above formula, ρ represents the density of the slurry, u is the grouting speed, p is the grouting pressure, μ is the viscosity of the slurry, f represents the body force term, t Indicates time; represents the gradient, that is, the vector in the direction of the maximum increment of the constant function; The boundary condition setting includes applying a fixed flow rate and pressure at the grouting hole, which are expressed as: ； ； In the above formula, Q Indicates the injection flow rate, the unit is volume / time ( ); C represents a constant; P Indicates the injection pressure, Indicates the pressure injected at the grouting hole; At the same time, a no-flow boundary condition is set on the discrete unit boundary to simulate the actual permeability characteristics. The no-flow boundary condition is: ； In the above formula, q represents the velocity vector of the fluid, is the normal vector of the boundary surface, representing the flux of fluid through the boundary; The initial conditions are set as the concentration distribution of the slurry before injection and the initial water content of the soil. The concentration before slurry injection and the initial water content of the soil are expressed as: ； ； In the above formula, Indicates the spatial position ( x,y,z ), Before grouting begins ( t =0) initial concentration distribution; is in position ( x,y,z ) at the soil moisture content, represents the initial soil moisture content; The grouting simulation module is used to predict the grouting pressure, grouting speed, and slurry diffusion form at the next time based on the physical and mathematical model of the entire grouting process, as well as the boundary conditions and initial conditions of each grid area, according to the current working conditions information. After continuous updating, the simulation of the tunnel grouting process is completed.

5. An electronic device, characterized in that: include: a memory for storing executable instructions; The processor is configured to implement the tunnel grouting simulation method based on a conditional generative adversarial network as described in any one of claims 1 to 3 when executing the executable instructions stored in the memory.

6. A computer-readable storage medium, characterized in that Executable instructions are stored, which are used to cause a processor to execute the executable instructions to implement the tunnel grouting simulation method based on conditional generative adversarial network as described in any one of claims 1 to 3.

7. A computer program product, characterized in that The computer program product includes executable instructions stored in a computer-readable storage medium; When the processor of the electronic device reads the executable instructions from the computer-readable storage medium and executes the executable instructions, the tunnel grouting simulation method based on the conditional generative adversarial network described in any one of claims 1 to 3 is implemented.

Images