A tunnel grouting diffusion form prediction system and method
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CHINA UNIV OF MINING & TECH
- Filing Date
- 2024-09-27
- Publication Date
- 2026-06-26
Smart Images
Figure CN119290676B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of tunnel grouting technology, and particularly relates to a morphological prediction system and method for tunnel grout diffusion. Background Technology
[0002] Tunnel engineering plays a vital role in urban construction and transportation, and grouting technology, as one of the important construction methods in underground engineering, is widely used in tunnel engineering to improve the stability and sealing of underground structures. However, research on the morphology of grout diffusion in tunnel fissures and porous media and its influencing factors still faces many challenges. Currently, research on the grout diffusion morphology in tunnel grouting projects mainly focuses on two aspects: first, quantitative analysis of the permeability and diffusion performance of grout during the grouting process; and second, assessment and prediction of the stability and sealing of underground structures after grouting.
[0003] Regarding the permeability and diffusion properties of grout, scholars both domestically and internationally have explored the diffusion patterns of grout in different media such as rocks and soils through laboratory simulation experiments and numerical simulations. However, existing studies often neglect the influence of geological structural complexity on grouting effectiveness and lack intuitive demonstrations and quantitative analysis of grout diffusion morphology in fractured and porous media.
[0004] Currently, the assessment and prediction of the stability and sealing performance of underground structures after grouting mainly rely on empirical formulas and numerical simulation methods. However, due to the complexity and uncertainty of underground engineering, existing methods often struggle to accurately predict the changes in grout diffusion patterns across different times and spaces, thus affecting the effectiveness and safety of grouting projects.
[0005] To address the problems in the existing technology, there is an urgent need to propose a morphological prediction system and method for tunnel grout diffusion. Summary of the Invention
[0006] To address the aforementioned technical problems, this invention proposes a morphological prediction system and method for tunnel grout diffusion, thereby resolving the issues present in the prior art.
[0007] To achieve the above objectives, the present invention provides a morphological prediction system for tunnel grout diffusion, comprising: a grout diffusion visualization system, a grout diffusion morphological quantification system, and a grout diffusion morphological prediction system;
[0008] The grouting diffusion visualization system includes a grouting simulation system and a tunnel simulation device;
[0009] The tunnel simulation device is used to prepare a tunnel model with fissures to simulate tunnel rock mass;
[0010] The grouting simulation system is used to grout the tunnel simulation device by adjusting the influencing factors of grout diffusion, so as to obtain the grout diffusion state under the influence of multiple factors.
[0011] The grout diffusion morphology quantification system is used to cut the grouting tunnel simulation device to obtain slices at several depth positions, and to divide the surface of each slice at a depth position into a grid, record the grout diffusion morphology, and obtain a grout diffusion morphology quantification dataset.
[0012] The grouting diffusion morphology prediction system is used to construct a multimodal spatiotemporal fusion depth prediction model, and inputs the grout diffusion morphology quantification dataset into the multimodal spatiotemporal fusion depth prediction model to predict the future diffusion morphology of the grout.
[0013] Optionally, the preparation process of the tunnel simulation device includes: mixing river sand, paraffin wax and pebbles in a preset ratio and casting them into a mold to obtain a tunnel model; cutting the tunnel model based on geological statistical fracture distribution parameters to obtain simulated tunnel fractures; mixing coal powder, mica fragments, perlite and foaming agent and applying them to the cut surface of the tunnel model to obtain a simulated tunnel fracture surface, thus completing the preparation of the tunnel simulation device.
[0014] Optionally, the tunnel model is horseshoe-shaped.
[0015] Optionally, the factors affecting grout diffusion include the tunnel model's own medium, fracture aperture, fracture length, fracture surface roughness, medium porosity, grout concentration, and grouting pressure.
[0016] Optionally, the grouting simulation system includes a grouting test platform, a grouting bucket, grout, and pigments, used to grout the tunnel model on the grouting test platform based on the grouting bucket, and to add different pigments to the grout for different fractures.
[0017] Optionally, the grouting diffusion morphology prediction system includes a model building module, a model training module, and a morphology prediction module;
[0018] The model building module is used to construct a multimodal spatiotemporal fusion deep prediction model based on convolutional neural networks, long short-term memory networks, graph neural networks, and spatiotemporal convolutional networks.
[0019] The model training module is used to train the multimodal spatiotemporal fusion deep prediction model by combining fluid dynamics and permeability constraints in physical modeling.
[0020] The morphology prediction module is used to obtain the grout diffusion morphology at each moment and location during the future tunnel grouting process based on the trained multimodal spatiotemporal fusion depth prediction model.
[0021] Optionally, the morphology prediction module includes a data input unit, a feature extraction unit, a feature fusion unit, and a morphology prediction unit;
[0022] The data input unit is used to input the quantitative dataset of slurry diffusion influencing factors and slurry diffusion morphology into the trained multimodal spatiotemporal fusion depth prediction model.
[0023] The feature extraction unit is used to extract the geometric features of the rock mass based on the convolutional neural network, extract the dynamic features of grout diffusion based on the long short-term memory network, and extract the fracture correlation based on the graph neural network.
[0024] The feature fusion unit is used to fuse the features output by the feature extraction unit based on a fusion mechanism to obtain fused features;
[0025] The morphology prediction unit is used to predict the morphology of slurry diffusion based on the fused features and the spatiotemporal features extracted by the spatiotemporal convolutional network.
[0026] This invention also provides a method for predicting the morphology of tunnel grout diffusion, and a morphology prediction system for tunnel grout diffusion, comprising the following steps:
[0027] A tunnel simulation device was prepared, and grouting was performed on the tunnel simulation device based on a grouting simulation system;
[0028] The grouting tunnel simulation device was cut to obtain slices at several depth positions;
[0029] The surface of the slice at each depth position is divided into grids, the slurry diffusion pattern is recorded, and a quantitative dataset of slurry diffusion pattern is obtained.
[0030] A multimodal spatiotemporal fusion deep prediction model is constructed, and the quantified dataset of slurry diffusion morphology is input into the multimodal spatiotemporal fusion deep prediction model to predict the future diffusion morphology of slurry.
[0031] Optionally, the process of dividing the slice surface at each depth location into a grid, recording the slurry diffusion pattern, and obtaining a quantitative dataset of slurry diffusion patterns includes:
[0032] The slurry diffusion in each grid is encoded with different values. By recording the grid code at each slice, the slurry diffusion data at each depth position is obtained. The slurry diffusion data at each depth position is then combined to form a complete slurry diffusion morphology quantification dataset.
[0033] Optionally, a multimodal spatiotemporal fusion depth prediction model is constructed based on the slurry diffusion morphology quantification dataset. The process of predicting the future diffusion morphology of slurry based on the multimodal spatiotemporal fusion depth prediction model includes:
[0034] A multimodal spatiotemporal fusion deep prediction model is constructed based on convolutional neural networks, long short-term memory networks, graph neural networks, and spatiotemporal convolutional networks.
[0035] The quantitative dataset of factors influencing slurry diffusion and slurry diffusion morphology is input into the trained multimodal spatiotemporal fusion depth prediction model. Based on the convolutional neural network, the geometric features of the rock mass are extracted; based on the long short-term memory network, the dynamic features of slurry diffusion are extracted; and based on the graph neural network, the fracture correlation is extracted.
[0036] The geometric features, dynamic features of slurry diffusion, and fracture correlation of the rock mass are fused based on the fusion mechanism to obtain fused features. Based on the fused features and combined with the spatiotemporal features extracted by the spatiotemporal convolutional network, the slurry diffusion morphology is predicted.
[0037] Compared with the prior art, the present invention has the following advantages and technical effects:
[0038] This invention proposes a visualization device for grout diffusion in tunnel fracture-porous media, comprising a tunnel simulation device and a grout simulation system. By simulating the preparation of rock mass and adjusting the grout simulation system, the influence of multiple factors such as rock mass properties, fracture characteristics, and grout injection parameters on the grout diffusion morphology can be considered simultaneously, thereby more realistically simulating the situation in actual engineering.
[0039] This invention proposes a grout diffusion morphology quantification system. Through steps such as cutting a tunnel simulation device, photographing and recording data, and meshing, it achieves precise quantification of the grout diffusion morphology. Compared with traditional methods, this system can more accurately record the diffusion state of the grout on the tunnel face and generate a digital diffusion morphology dataset, providing a more reliable foundation for subsequent data processing and analysis.
[0040] This invention proposes a grouting diffusion morphology prediction system, which realizes the spatiotemporal prediction of grouting diffusion morphology by establishing a prediction model and using a grout diffusion morphology quantification dataset as a spatiotemporal sequence dataset. Attached Figure Description
[0041] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:
[0042] Figure 1 This is a schematic diagram of the prediction process according to an embodiment of the present invention;
[0043] Figure 2 This is a schematic diagram of a horseshoe-shaped tunnel simulation device according to an embodiment of the present invention;
[0044] Figure 3 This is a schematic diagram of the grout diffusion morphology quantification system according to an embodiment of the present invention;
[0045] Figure 4 This is a schematic diagram illustrating the prediction of grouting diffusion morphology in an embodiment of the present invention. Detailed Implementation
[0046] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0047] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.
[0048] like Figure 1 As shown, this embodiment provides a morphological prediction system for tunnel grout diffusion, including: a grout diffusion visualization system, a grout diffusion morphological quantification system, and a grout diffusion morphological prediction system;
[0049] The grouting diffusion visualization system includes a grouting simulation system and a tunnel simulation device;
[0050] The tunnel simulation device is used to prepare a tunnel model with fissures to simulate tunnel rock mass;
[0051] The grouting simulation system is used to grout the tunnel simulation device by adjusting the influencing factors of grout diffusion, so as to obtain the grout diffusion state under the influence of multiple factors.
[0052] The grout diffusion morphology quantification system is used to cut the grouting tunnel simulation device to obtain slices at several depth positions, and to divide the surface of each slice at a depth position into a grid, record the grout diffusion morphology, and obtain a grout diffusion morphology quantification dataset.
[0053] The grouting diffusion morphology prediction system is used to construct a multimodal spatiotemporal fusion depth prediction model, and inputs the grout diffusion morphology quantification dataset into the multimodal spatiotemporal fusion depth prediction model to predict the future diffusion morphology of the grout.
[0054] As a specific embodiment, the tunnel simulation device, such as Figure 2The device uses a 100:10 mixture of river sand and paraffin wax to simulate tunnel rock mass, with appropriate pebbles added. After heating the mixed materials evenly, the mixture is poured into the test chamber multiple times. After the model cools, a horseshoe-shaped tunnel model is obtained. Based on geological statistics of fracture distribution parameters, the simulated rock mass is cut to simulate tunnel fractures by cutting the gaps. To more realistically simulate the surface morphology of tunnel fractures, auxiliary materials such as coal powder, mica fragments, perlite, and foaming agents are added and applied to the cut surface of the simulated rock mass to simulate the rough and irregular surface of joints and fractures. Adjusting the proportions of the above materials can simulate the surface roughness and width of the fracture surface.
[0055] As a specific embodiment, the grouting simulation system includes a grouting test operating platform, a grouting tank, grout, and pigments. The grout diffusivity is mainly affected by the following factors: the rock mass itself; fracture aperture, fracture surface roughness, and fracture length; porous media porosity; grout concentration; and grouting pressure. The permeability of the rock mass itself can be adjusted by regulating the ratio of river sand to paraffin wax and the number of pebbles. Based on geostatistical methods and referencing multiple tunnel engineering projects, fracture length and aperture are set. The fracture aperture and length can be adjusted by adjusting the model's cut voids. The fracture surface roughness is adjusted by changing the proportions of auxiliary materials such as coal powder, mica fragments, perlite, and foaming agents. Increasing the ratio of river sand to pebbles simulates porous media and adjusts its porosity. The grout concentration and grouting pressure are continuously adjusted. By adding different pigments to the grout in different fractures, the diffusion of grout in different fractures can be clearly distinguished. By changing the above parameters, the diffusion state of different grouts through tunnel fractures under the influence of multiple factors can be observed.
[0056] As a specific embodiment, the grout diffusion morphology quantification system cuts the horseshoe-shaped tunnel simulation device after grouting to simulate the process of tunnel face excavation and exposure, photographs and records the diffusion state of the grout (including color) on the tunnel face surface, and performs mesh division processing on the photographs. Figure 3 The appropriate grid size is selected based on the size of the cracks to facilitate subsequent data processing and analysis. Each grid represents a small area; a grid with one color of slurry diffusion is denoted as 1, a grid with another color of slurry diffusion as 2, and so on, with a grid without slurry diffusion as 0. The depth of the slice and the slurry diffusion pattern are recorded, i.e., the grid is digitally characterized to obtain a dense dataset of slurry diffusion patterns at multiple depth locations, thus establishing a quantitative dataset of slurry diffusion patterns.
[0057] As a specific embodiment, the grouting diffusion morphology prediction system includes a model building module, a model training module, and a morphology prediction module. The model building module is used to construct a multimodal spatiotemporal fusion deep prediction model based on convolutional neural networks, long short-term memory networks, graph neural networks, and spatiotemporal convolutional networks. The model training module is used to train the multimodal spatiotemporal fusion deep prediction model by combining fluid dynamics and permeability constraints in physical modeling. The morphology prediction module is used to obtain the grout diffusion morphology at each moment and location during the future tunnel grouting process based on the trained multimodal spatiotemporal fusion deep prediction model.
[0058] Furthermore, the morphology prediction module includes a data input unit, a feature extraction unit, a feature fusion unit, and a morphology prediction unit. The data input unit is used to input the quantified dataset of slurry diffusion influencing factors and slurry diffusion morphology into the trained multimodal spatiotemporal fusion depth prediction model. The feature extraction unit is used to extract the geometric features of the rock mass based on the convolutional neural network, extract the dynamic features of slurry diffusion based on the long short-term memory network, and extract the fracture correlation based on the graph neural network. The feature fusion unit is used to fuse the features output by the feature extraction unit based on the fusion mechanism to obtain fused features. The morphology prediction unit is used to predict the slurry diffusion morphology based on the fused features and combined with the spatiotemporal features extracted by the spatiotemporal convolutional network.
[0059] As a feasible approach, the grouting diffusion morphology prediction system predicts future grouting morphology by establishing a multimodal spatiotemporal fusion depth prediction model. The prediction model analogizes the horseshoe-shaped tunnel excavation depth to "time" in a spatiotemporal sequence, such as... Figure 4 As shown, the grout diffusion morphology at different depths of the tunnel face is presented as time-series data. A predictive model for the grout diffusion morphology is established based on this spatiotemporal data. By adjusting the model's parameters and input data, the model is repeatedly trained and optimized until it can accurately predict the grout diffusion morphology under different construction conditions. The input data includes multiple parameters affecting grout diffusion, such as fracture aperture, surface roughness, fracture length, porosity, grout concentration, and grouting pressure. The final output of the prediction system is the future grout diffusion morphology, which can provide a scientific basis for construction decisions in tunnel engineering.
[0060] Furthermore, the multimodal spatiotemporal fusion deep prediction model proposed in this embodiment integrates multimodal data and spatiotemporal information. It utilizes deep learning technologies such as convolutional neural networks (CNN), long short-term memory networks (LSTM), and graph neural networks (GNN), combined with physical modeling methods, to achieve accurate prediction of the grouting diffusion morphology in tunnels.
[0061] Multimodal data input includes: Geometric attribute input: including physical attribute data such as fracture length, aperture, surface roughness, and porosity. Environmental attribute input: including dynamic environmental parameters such as grouting pressure, grout concentration, and temperature during construction. Spatiotemporal attribute input: including spatiotemporal information such as tunnel excavation depth and grouting time.
[0062] Then, a convolutional neural network (CNN) was used to extract complex geometric features of the fractures and rock mass surface, such as the distribution of surface roughness and fracture width. A long short-term memory network (LSTM) was used to extract the dynamic features of grout diffusion during the grouting process, especially capturing long-term dependencies in the time series. A graph neural network (GNN) was used to analyze the spatial correlation and mutual influence between fractures, optimizing the path and morphology of grout diffusion in the fracture network.
[0063] The outputs from different feature extraction modules are integrated through a fusion mechanism. An attention mechanism can be used to dynamically adjust the weights of different features to enhance the model's ability to perceive key features. Finally, the spatiotemporal attributes are input into a spatiotemporal convolutional network, and combined with the previously extracted features, spatiotemporal prediction of grouting diffusion patterns is achieved.
[0064] Furthermore, during model training, fluid dynamics and permeability constraints from physical modeling are incorporated to ensure that the prediction results conform to physical laws. This module can be implemented by solving the Navier-Stokes equations or Darcy's law.
[0065] Furthermore, the model's final output is the grout diffusion pattern at every moment and location during the tunnel grouting process. By combining actual engineering data and prediction results, the model's parameters are continuously optimized to make the prediction results more consistent with actual engineering conditions.
[0066] This invention proposes a visualization device for grout diffusion in tunnel fracture-porous media, comprising a horseshoe-shaped tunnel simulation device and a grout simulation system. By simulating the preparation of the rock mass and adjusting the grout simulation system, the influence of multiple factors such as rock mass properties, fracture characteristics, and grout injection parameters on the grout diffusion morphology can be considered simultaneously, thereby more realistically simulating the situation in actual engineering.
[0067] This invention proposes a grout diffusion morphology quantification system. Through steps such as cutting a horseshoe-shaped tunnel simulation device, photographing and recording data, and meshing, it achieves precise quantification of the grout diffusion morphology. Compared with traditional methods, this system can more accurately record the diffusion state of the grout on the tunnel face and generate a digital diffusion morphology dataset, providing a more reliable foundation for subsequent data processing and analysis.
[0068] This invention proposes a grout diffusion morphology prediction system. By establishing a prediction model and utilizing spatiotemporal sequence datasets, it achieves the prediction of grout diffusion morphology. This system can predict the grout diffusion morphology at different depths of the tunnel face based on different grout influence parameters, providing an effective prediction tool for the design and practice of underground engineering. By combining data from multiple modalities such as geometric, dynamic, spatial, and spatiotemporal data, the system improves the model's predictive ability for grout diffusion morphology in complex scenarios. By combining spatiotemporal convolutional networks from deep learning with traditional physical constraints, the system ensures that the model not only has high prediction accuracy but also follows physical laws. Through an adaptive feedback mechanism, the model can continuously adjust and optimize itself to adapt to changes under different construction conditions.
[0069] On the other hand, based on the same inventive concept as the above embodiments, this embodiment also provides a method for predicting the morphology of tunnel grout diffusion. This prediction method is comparable in effect to the tunnel grout diffusion morphology prediction system provided in the above embodiments. The prediction method includes the following steps:
[0070] (1) Based on the actual project, a horseshoe-shaped tunnel simulation device was established, the ratio of river sand to paraffin wax was adjusted, a horseshoe-shaped simulated tunnel was established, and the surface roughness of the fracture was simulated by coal powder, mica fragments and perlite.
[0071] (2) Prepare grouting slurries of different viscosities and add pigments. The grouting simulation system grouts the horseshoe-shaped tunnel simulation device and can adjust the grouting pressure.
[0072] (3) The horseshoe-shaped tunnel simulation device is sliced and photographed;
[0073] (4) Use grid partitioning to quantify slurry diffusion morphology and establish a slurry diffusion morphology quantification dataset;
[0074] (5) The slurry diffusion morphology quantification dataset is used as the spatiotemporal sequence dataset of grouting diffusion morphology, and a neural network model is used to conduct prediction studies on it.
[0075] (6) Use engineering excavation data to perform transfer learning on the spatiotemporal prediction neural network model of grout diffusion morphology.
[0076] The feasible process of dividing the slice surface into meshes at each depth location, recording the slurry diffusion morphology, and obtaining a quantitative dataset of slurry diffusion morphology includes:
[0077] (1) Mesh division: During the cutting and exposure process of the horseshoe-shaped tunnel simulation device after grouting, the surface of the tunnel face is divided into meshes. The size of the mesh is selected according to the size of the crack to facilitate subsequent fine data processing.
[0078] (2) Color coding and numerical representation: The photographs are processed and the diffusion of slurry in each grid is coded with different values. For example, different colors of slurry are represented by 1, 2, etc., and areas without diffusion are recorded as 0.
[0079] (3) Slicing at depth locations: By slicing the model, the slurry diffusion patterns at multiple depth locations are obtained. Slurry diffusion data for each depth location is obtained through photography and mesh generation.
[0080] (4) Quantitative Dataset Establishment: By recording the grid encoding at each slice, a complete quantitative dataset of slurry diffusion morphology is finally established. This dataset can be used for subsequent analysis and training of prediction models.
[0081] The feasible process of predicting the future diffusion pattern of slurry based on a multimodal spatiotemporal fusion deep prediction model includes:
[0082] A multimodal spatiotemporal fusion deep prediction model is constructed based on convolutional neural networks, long short-term memory networks, graph neural networks, and spatiotemporal convolutional networks.
[0083] The quantitative dataset of factors influencing slurry diffusion and slurry diffusion morphology is input into the trained multimodal spatiotemporal fusion depth prediction model. Based on the convolutional neural network, the geometric features of the rock mass are extracted; based on the long short-term memory network, the dynamic features of slurry diffusion are extracted; and based on the graph neural network, the fracture correlation is extracted.
[0084] The geometric features, dynamic features of slurry diffusion, and fracture correlation of the rock mass are fused based on the fusion mechanism to obtain fused features. Based on the fused features and combined with the spatiotemporal features extracted by the spatiotemporal convolutional network, the slurry diffusion morphology is predicted.
[0085] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A morphological prediction system for tunnel grout diffusion, characterized in that, include: Grouting diffusion visualization system, grouting diffusion morphology quantification system, and grouting diffusion morphology prediction system; The grouting diffusion visualization system includes a grouting simulation system and a tunnel simulation device; The tunnel simulation device is used to prepare a tunnel model with fissures to simulate tunnel rock mass; The grouting simulation system is used to grout the tunnel simulation device by adjusting the influencing factors of grout diffusion, so as to obtain the grout diffusion state under the influence of multiple factors. The grout diffusion morphology quantification system is used to cut the grouting tunnel simulation device to obtain slices at several depth positions, and to divide the surface of each slice at a depth position into a grid, record the grout diffusion morphology, and obtain a grout diffusion morphology quantification dataset. The grouting diffusion morphology prediction system is used to construct a multimodal spatiotemporal fusion depth prediction model, and inputs the grout diffusion morphology quantification dataset into the multimodal spatiotemporal fusion depth prediction model to predict the future diffusion morphology of the grout. The grouting diffusion morphology prediction system includes a model building module, a model training module, and a morphology prediction module; The model building module is used to construct a multimodal spatiotemporal fusion deep prediction model based on convolutional neural networks, long short-term memory networks, graph neural networks, and spatiotemporal convolutional networks. The model training module is used to train the multimodal spatiotemporal fusion deep prediction model by combining fluid dynamics and permeability constraints in physical modeling. The morphology prediction module is used to obtain the grout diffusion morphology at each moment and location during the future tunnel grouting process based on the trained multimodal spatiotemporal fusion depth prediction model.
2. The tunnel grout diffusion morphology prediction system according to claim 1, characterized in that, The preparation process of the tunnel simulation device includes: mixing river sand, paraffin wax and pebbles in a preset ratio and casting them into a mold to obtain a tunnel model; cutting the tunnel model based on geological statistical fracture distribution parameters to obtain simulated tunnel fractures; mixing coal powder, mica fragments, perlite and foaming agent and coating them onto the cut surface of the tunnel model to obtain a simulated tunnel fracture surface, thus completing the preparation of the tunnel simulation device.
3. The tunnel grout diffusion morphology prediction system according to claim 2, characterized in that, The tunnel model is horseshoe-shaped.
4. The tunnel grout diffusion morphology prediction system according to claim 1, characterized in that, Factors affecting grout diffusion include the tunnel model's own medium, fracture aperture, fracture length, fracture surface roughness, medium porosity, grout concentration, and grouting pressure.
5. The tunnel grout diffusion morphology prediction system according to claim 1, characterized in that, The grouting simulation system includes a grouting test platform, a grouting bucket, grout, and pigments. It is used to grout the tunnel model on the grouting test platform based on the grouting bucket, and to add different pigments to the grout for different fissures.
6. The tunnel grout diffusion morphology prediction system according to claim 1, characterized in that, The morphology prediction module includes a data input unit, a feature extraction unit, a feature fusion unit, and a morphology prediction unit; The data input unit is used to input the quantitative dataset of slurry diffusion influencing factors and slurry diffusion morphology into the trained multimodal spatiotemporal fusion depth prediction model. The feature extraction unit is used to extract the geometric features of the rock mass based on the convolutional neural network, extract the dynamic features of grout diffusion based on the long short-term memory network, and extract the fracture correlation based on the graph neural network. The feature fusion unit is used to fuse the features output by the feature extraction unit based on a fusion mechanism to obtain fused features; The morphology prediction unit is used to predict the morphology of slurry diffusion based on the fused features and the spatiotemporal features extracted by the spatiotemporal convolutional network.
7. A method for predicting the morphology of grout diffusion in tunnels, characterized in that, The tunnel grouting diffusion morphology prediction system based on any one of claims 1-6 includes the following steps: A tunnel simulation device was prepared, and grouting was performed on the tunnel simulation device based on a grouting simulation system; The grouting tunnel simulation device was cut to obtain slices at several depth positions; The surface of the slice at each depth position is divided into grids, the slurry diffusion pattern is recorded, and a quantitative dataset of slurry diffusion pattern is obtained. A multimodal spatiotemporal fusion deep prediction model is constructed, and the quantified dataset of slurry diffusion morphology is input into the multimodal spatiotemporal fusion deep prediction model to predict the future diffusion morphology of slurry. The process of constructing a multimodal spatiotemporal fusion depth prediction model and predicting the future diffusion morphology of slurry based on the multimodal spatiotemporal fusion depth prediction model includes: A multimodal spatiotemporal fusion deep prediction model is constructed based on convolutional neural networks, long short-term memory networks, graph neural networks, and spatiotemporal convolutional networks. The quantitative dataset of factors influencing slurry diffusion and slurry diffusion morphology is input into the trained multimodal spatiotemporal fusion depth prediction model. Based on the convolutional neural network, the geometric features of the rock mass are extracted; based on the long short-term memory network, the dynamic features of slurry diffusion are extracted; and based on the graph neural network, the fracture correlation is extracted. The geometric features, dynamic features of slurry diffusion, and fracture correlation of the rock mass are fused based on the fusion mechanism to obtain fused features. Based on the fused features and combined with the spatiotemporal features extracted by the spatiotemporal convolutional network, the slurry diffusion morphology is predicted.
8. The method for predicting the morphology of tunnel grout diffusion according to claim 7, characterized in that, The process of dividing the slice surface at each depth location into a grid, recording the slurry diffusion pattern, and obtaining a quantitative dataset of slurry diffusion patterns includes: The slurry diffusion in each grid is encoded with different values. By recording the grid code at each slice, the slurry diffusion data at each depth position is obtained. The slurry diffusion data at each depth position is then combined to form a complete slurry diffusion morphology quantification dataset.