Tunnel grouting process multi-physics field real-time assimilation simulation and regulation method and system

Abstract

The present application belongs to the technical field of tunnel engineering, in order to solve the problem of lacking real-time dynamic simulation and automatic regulation and control in the existing tunnel construction, a tunnel grouting process multi-physical field real-time assimilation simulation and regulation and control method and system are proposed, using ensemble Kalman filtering, combining real-time monitoring data in the tunnel grouting process, dynamically correcting the parameters of the multi-physical model; considering the time correlation in the slurry setting process, integrating the time-varying setting model depicting the change of physical properties of slurry with time evolution, embedding the time-varying setting model as an external function in the multi-physical field model in the correction, so that the numerical simulation self-adapts the slurry flow state; according to the dynamic simulation results, the control parameters of tunnel grouting are generated, and the closed-loop regulation and control of tunnel grouting is realized. The present application realizes the synchronous linkage of numerical simulation and field working condition.

Description

Technical Field

[0001] This invention belongs to the technical field of tunnel engineering, and in particular relates to a method and system for real-time assimilation and control of multi-physics fields in the tunnel grouting process. Background Technology

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

[0003] Tunnel grouting technology is an important means of controlling stratum permeability, stabilizing surrounding rock structure, and preventing water and mud inrushes, and is widely used in complex environments such as water-rich and loose strata. However, at present, grouting construction generally adopts the method of "pre-modeling + construction monitoring + experience adjustment", which relies on the static model established before construction to simulate grout diffusion. However, this mode is difficult to adapt to the real-time changes in geological parameters during construction, and there are problems such as lag in simulation results and insufficient accuracy.

[0004] During grouting, the formation structure, water pressure, stress field, and grout rheological properties often change drastically over time. Traditional static simulation cannot dynamically correct model parameters based on real-time monitoring results, leading to a disconnect between simulation and reality. In addition, although a large number of sensors are installed on site to collect data such as pressure, flow rate, and deformation in real time, most of this data is not effectively fed back to the model system, lacking a closed-loop control capability of "monitoring-simulation-control".

[0005] Meanwhile, on-site grouting operations often rely on manual experience for adjustments, lacking unified quantitative standards and failing to systematically identify problems such as grout leakage, deviation, or insufficient diffusion. Especially in water-rich sand and gravel layers or fractured zones, the grout diffusion path is poorly controlled, easily leading to grouting failure or the risk of water inrush. Traditional simulation platforms have low computational efficiency and long response cycles, making it difficult to meet the minute-level decision-making and control needs in complex environments.

[0006] Faced with sudden and abnormal working conditions such as water inrush and surrounding rock slippage, current systems generally lack the ability to respond quickly based on historical cases, often requiring work stoppages, delaying opportunities, and increasing risks. With the advancement of digital construction, there is an urgent need for a new type of grouting system that integrates IoT sensing, data assimilation, high-performance simulation, and intelligent control to achieve data-driven real-time dynamic simulation and automated regulation, thereby improving the level of intelligent construction. Summary of the Invention

[0007] To overcome the shortcomings of the existing technologies, this invention provides a method and system for real-time multi-physics field assimilation simulation and control of tunnel grouting process. It constructs a closed-loop control of "monitoring-simulation-control" to realize an automated process from perception to decision-making and then to execution, reducing human intervention and improving the level of intelligent construction.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] In a first aspect, the present invention provides a method for real-time multi-physics field assimilation simulation and control of the tunnel grouting process, including:

[0010] Based on the initial physical parameters of the tunnel, a multiphysics model coupling groundwater, stratum stress, and slurry diffusion is constructed.

[0011] By using ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting, the parameters of the multi-physics model are dynamically corrected.

[0012] Considering the time correlation in the slurry coagulation process, a time-varying coagulation model that characterizes the changes in the physical properties of the slurry over time is integrated. The time-varying coagulation model is then embedded as an external function in the time step into the modified multiphysics model, enabling the numerical simulation to adaptively adjust the slurry flow state.

[0013] Control parameters for tunnel grouting are generated based on dynamic simulation results, enabling closed-loop control of tunnel grouting.

[0014] Secondly, the present invention provides a multi-physics field real-time assimilation simulation and control system for tunnel grouting process, comprising:

[0015] The module is used to construct a multiphysics model of groundwater-stratum stress-grout diffusion coupling based on the initial physical parameters of the tunnel.

[0016] The correction module is used to dynamically correct the parameters of the multi-physics model by using ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting.

[0017] The simulation module is used to consider the time correlation in the slurry coagulation process, integrates a time-varying coagulation model that describes the changes in the physical properties of the slurry over time, and embeds the time-varying coagulation model as an external function in the time step into the modified multiphysics model, so that the numerical simulation can adaptively adjust the slurry flow state.

[0018] The control module is used to generate control parameters for tunnel grouting based on dynamic simulation results, thereby achieving closed-loop control of tunnel grouting.

[0019] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0020] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0021] The above one or more technical solutions have the following beneficial effects:

[0022] In this invention, by using an ensemble Kalman filter algorithm, the model can be continuously updated based on real-time monitoring data, solving the problem of lag in traditional simulations; achieving "synchronous linkage" between numerical simulation and on-site working conditions, significantly improving the simulation's realism and guiding value, and achieving second-level synchronization between numerical simulation and actual construction; embedding a time-varying solidification model that characterizes the changes in the physical properties of grout over time into a multiphysics model can better reflect the actual grouting hardening process.

[0023] This invention constructs a closed-loop control system of "monitoring-simulation-regulation" to realize an automated process from perception to decision-making and then to execution, reducing human intervention and improving the level of intelligent construction.

[0024] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0025] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0026] Figure 1 This is a flowchart of the dynamic simulation and control method for tunnel grouting process in Embodiment 1 of the present invention. Detailed Implementation

[0027] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0029] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0030] Example 1

[0031] like Figure 1 As shown, this embodiment discloses a method for real-time multiphysics field assimilation simulation and control of the tunnel grouting process, including:

[0032] Based on the initial physical parameters of the tunnel, a multiphysics model coupling groundwater, stratum stress, and slurry diffusion is constructed.

[0033] By using ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting, the parameters of the multi-physics model are dynamically corrected.

[0034] Considering the time correlation in the slurry coagulation process, a time-varying coagulation model that characterizes the changes in the physical properties of the slurry over time is integrated. The time-varying coagulation model is embedded as an external function in the time step into the modified multiphysics model, so that the numerical simulation can adaptively adjust the slurry flow state.

[0035] Control parameters for tunnel grouting are generated based on dynamic simulation results, enabling closed-loop control of tunnel grouting.

[0036] This embodiment constructs a synchronous integrated "monitoring-simulation-decision" system, breaking through the bottlenecks of traditional grouting technology such as static simulation, feedback lag, and slow response. It integrates IoT sensing, data assimilation algorithms, high-performance simulation, and intelligent decision control to achieve closed-loop management of the grouting process through "real-time sensing-intelligent simulation-automatic regulation," thereby significantly improving construction efficiency, safety, and resource utilization under complex geological conditions.

[0037] The following is a detailed description of the multiphysics field real-time assimilation simulation and control method for the tunnel grouting process proposed in this embodiment:

[0038] To achieve comprehensive perception of the tunnel grouting process, the deployment of the monitoring network in this embodiment will be explained first. The sensors should be deployed according to the principle of "zonal deployment and density classification", giving priority to covering water-rich strata, fracture zones and areas with active deformation, forming a three-dimensional monitoring network that integrates vertical, horizontal and depth dimensions.

[0039] Distributed fiber optic sensors and microseismic sensors are deployed to acquire information such as stratum distribution, layer thickness, and fracture development characteristics. Specifically, distributed fiber optic sensors are installed or pre-embedded along the tunnel axis, support structure, or boreholes, with one set every 10–20 meters. A single set can cover multiple observation points to monitor stratum strain, temperature changes, and fracture propagation. Microseismic sensors are deployed on the surrounding rock surface or inside the borehole, forming an array of three or more nodes, with one set every 50 meters. They are used to sense microseismic events, fracture initiation and propagation processes, and to locate the fracture source.

[0040] The grouting pressure and flow sensors are installed at the grouting pump outlet, pipeline nodes, and each grouting hole. Each grouting system is equipped with one pressure sensor and one flow sensor. If there are multiple branches, each branch line needs to be laid out separately.

[0041] Pore ​​water pressure gauges and displacement gauges are optional sensors. If there are high-risk areas, such as water inrush faults, 1-2 sets should be installed every 50 meters, buried in the surrounding rock or boreholes, to monitor groundwater pressure and stratum displacement, and to support the improvement of model inversion accuracy.

[0042] Each sensor is connected to the edge computing node via an industrial Ethernet network. The sampling frequency of key sensing points can reach 10 Hz, and that of ordinary areas is no less than 1 Hz. After preliminary processing at the edge node, the data is uploaded to the central control system in real time.

[0043] The following is a detailed description of the dynamic simulation and control method for tunnel grouting process proposed in this embodiment:

[0044] Based on detailed geological exploration data, drilling and sampling reports, ground-penetrating radar scan results, historical grouting construction data, and laboratory parameter test results, an initial formation model is constructed. The initial physical parameters that need to be determined include formation permeability, elastic modulus and Poisson's ratio, grout viscosity and diffusion radius, and fracture model. Among them, formation permeability is obtained through inversion of seepage tests and field water injection experiments; elastic modulus and Poisson's ratio are calculated based on in-situ pressure gauge tests, standard penetration tests, and triaxial compression test data of rock samples; grout viscosity and diffusion radius are obtained through laboratory rheological property tests and previous grouting hole diffusion data; and a joint and fracture network is established using GPR, CT scan, or three-dimensional laser scan data as discrete element input parameters to build the fracture model.

[0045] Specifically, fracture modeling includes: fracture parameter extraction and processing, fracture network modeling, fracture-multiphysics coupling simulation, and model verification and correction.

[0046] 1. Crack parameter extraction and processing: Image recognition and machine learning algorithms are used to extract the number, length, dip angle, aperture, and filling status of cracks from the scan data. Based on the strike-dip clustering DFN construction method, the spatial distribution of cracks is restored by combining the structural surface inversion algorithm. The crack attribute parameters are standardized and used as input for subsequent modeling.

[0047] 2. Discrete Fracture Network (DFN) Modeling: Construct a DFN model to simulate the spatial distribution, geometric structure, and mechanical properties of natural fractures. Fracture network model parameters such as fracture density, continuity, and equivalent porosity are verified using field data.

[0048] 3. Fracture-Multiphysics Coupled Simulation: A fracture network is embedded in a finite element-discrete element coupled model to simulate the preferential flow and distribution of slurry within the fractures. Local mesh refinement technology is employed in the fracture region to enhance the accuracy of numerical calculations. Deformation responses during fracture closure and opening are considered to achieve dynamic correction of permeability.

[0049] 4. Validation and Correction of the Fracturing Network Model: The fracture parameters were iteratively corrected by comparing the results with those from on-site grout diffusion tests and grout coverage measurements. Ensemble Kalman filtering (EnKF) was used to perform real-time correction of properties such as fracture permeability during the simulation process to ensure synchronization between the simulation and the field.

[0050] Geotechnical mechanics simulation software was used for modeling, and a finite element method (FEM)-discrete element method (DEM) coupled algorithm was employed to simulate the diffusion behavior of grout in complex fractured media. A multiphysics model was constructed, incorporating the coupling of three fields: groundwater, formation stress, and grout diffusion. The model was deployed to a high-performance computing platform supporting GPU acceleration to improve computational efficiency and achieve second-level updates. Specifically, the finite element method was used to simulate the deformation response of continuous formations, while the discrete element method was used to simulate the seepage and diffusion of grout in fractured media.

[0051] All multiphysics model structure files, parameter libraries, and boundary conditions are managed using a modular modeling approach, which facilitates subsequent iterations and updates.

[0052] Constructing and updating a multiphysics model that couples the three fields of groundwater, formation stress, and slurry diffusion, specifically as follows:

[0053] Step 1: Model Domain Definition and Mesh Generation: Based on the geological survey results and tunnel design cross-section, a three-dimensional region model is established, encompassing the surrounding rock ahead of the tunnel face, the construction area, and the strata at the far end. The three-dimensional region model employs an unstructured mesh generation method, with local mesh refinement implemented in areas ahead of the tunnel face, fault zones, and areas with dense fractures. The mesh element type is hexahedral or tetrahedral, supporting spatial interpolation of multiphysics variables at nodes.

[0054] Step 2: Physics module configuration.

[0055] (1) Groundwater seepage field.

[0056] Establish the seepage control equation based on Darcy's law:

[0057]

[0058] in, k The permeability coefficient is indirectly controlled by the stress field. Pore ​​water pressure; Hydrodynamic viscosity; q is the specific water storage coefficient; q is the grout inflow term, with boundary conditions provided by the grout field; t is time. This is the gradient operator. Coupling terms: permeability k is updated as formation stress changes; q comes from grout injection.

[0059] The groundwater flow boundary conditions are set as fixed head, no permeability boundary, or flow control boundary; the groundwater model parameters are obtained based on water injection experiments, pore pressure tests, and permeability inversion.

[0060] (2) Formation stress field.

[0061] The governing equations of the mechanical field are established based on the elastic-plastic constitutive model:

[0062]

[0063] Where σ is the stress tensor and ρ is the density. For effective stress tensor; The pore pressure transmission coefficient; From groundwater field; It is a unit tensor.

[0064] Elastic modulus, Poisson's ratio, lithological zoning, and fault-plane contact characteristics are set for the strata.

[0065] (3) Slurry diffusion field.

[0066] The flow process of slurry in cracks / pores is based on the convection-diffusion-reaction equation:

[0067]

[0068] Where C is the slurry concentration; v = -(k / μ(t))▽Pg: flow velocity, controlled by permeability and viscosity; D(t) is the time-varying diffusion coefficient; Pg is the slurry pressure boundary; and μ is the slurry viscosity.

[0069] The initial conditions for slurry flow are derived from the grouting pump discharge rate and the orifice flow velocity. The instantaneous solidification model considers the changes in viscosity and diffusion coefficient over time caused by the slurry hardening process.

[0070] Step 3: Implementation of multi-field coupling.

[0071] The coupling steps are as follows:

[0072] Step 31: Initial model construction and parameter initialization.

[0073] Based on geological exploration data and borehole information, a three-dimensional geological model was established to delineate the stratigraphic structure, faults, joints, and fractures.

[0074] The initial parameters for each physical field are set as follows:

[0075] Groundwater field: permeability coefficient k, porosity n, specific storage coefficient ;

[0076] Stress field: elastic modulus E, Poisson's ratio ν, initial ground stress σ0, pore pressure transfer coefficient α;

[0077] Grout field: initial grout viscosity μ0, diffusion coefficient D0, grouting pressure boundary Pg.

[0078] Step 32: Establishment of the three-field control equations.

[0079] Based on Darcy's law, the governing equations for the groundwater field are established:

[0080]

[0081] in, k The permeability coefficient is indirectly controlled by the stress field. Pore ​​water pressure; Hydrodynamic viscosity; is the specific water storage coefficient; q is the grout inflow term, provided by the grout field as boundary conditions; t is time. Coupled terms: permeability k is updated with formation stress; q comes from grout injection.

[0082] Based on the elastic-plastic constitutive model, the governing equations of the geostress field are established:

[0083]

[0084] Where σ is the stress tensor and ρ is the density. For effective stress tensor; The pore pressure transmission coefficient; From groundwater field; It is a unit tensor.

[0085] Coupling term: Output of groundwater field The total stress is modified to the effective stress; the effective stress, in turn, is used to update the element deformation, thereby modifying... k .

[0086] Slurry diffusion field governing equation:

[0087]

[0088] Where C is the slurry concentration; v = -(k / μ(t))▽Pg: flow velocity, controlled by permeability and viscosity; D(t) is the time-varying diffusion coefficient; and Pg is the slurry pressure boundary.

[0089] Coupling terms: k is the permeability coefficient, derived from stress field feedback, i.e., stress → fracture → permeability; μ(t) comes from the time-varying condensation model; Pg and Shared.

[0090] The above equations are coupled with time steps through a unified grid system for computation.

[0091] Step 33: Defining coupling relationships and variable passing mechanisms.

[0092] To achieve high-fidelity coupled evolution of the three-field model, the system uses a unified spatial grid and time step to synchronously update the control variables of each field, and introduces the following coupling terms into the control equations:

[0093] Water-stress field coupling: expressing the effective stress σ′=σ-α through explicit introduction α is the pore water stress transfer coefficient, which directly embeds the groundwater pressure into the stress control equation, while stress feedback controls the permeability k(σ), which is used for the groundwater and slurry field.

[0094] Grout-stress field coupling: Grout injection causes changes in the local pressure field. After solving the stress field, the fracture aperture and deformation are updated, the permeability tensor is dynamically corrected, and the grout diffusion path is affected.

[0095] Slurry-water field coupling: The two are coupled through sharing pore space, and the slurry concentration field and pressure field affect the pore water pressure through governing equations. The reaction affects the slurry propulsion velocity and diffusion front, and the diffusion coefficient D(t) and the permeability-pressure field dynamically evolve together.

[0096] Step 34: Numerical solution strategy.

[0097] The following joint solution strategy is used for three-field coupling:

[0098] 1. Time step separation – field alternation method: Each global time step tn is divided into multiple sub-stages to solve the three physical fields in sequence.

[0099] The sub-steps are as follows:

[0100] Step 1: Fix the stress field, solve the seepage equation, and update the pore pressure distribution;

[0101] Step 2: Fix the pore pressure field, solve the consolidation equation, and update the deformation field and fracture state;

[0102] Step 3: Fix the stress and pore pressure field, and solve the equations for the evolution of slurry diffusion and coagulation;

[0103] Return to step one and repeat the iteration until the inter-field variables converge.

[0104] 2. Strongly Coupled Field Iteration: The three field variables are unified into a single solution system, forming a block matrix structure. Parallel processing is achieved using a linear solver, which is suitable for scenarios requiring short time steps and high precision.

[0105] Step 35: Coupling result feedback and dynamic update:

[0106] After each calculation, the following intermediate quantities are output: effective stress change field, porosity change field, and slurry concentration field.

[0107] The simulation results are compared with real-time monitoring data, such as sensor pressure, pore pressure, and diffusion boundary, and then ensemble Kalman filtering (EnKF) correction is performed.

[0108] Update model parameters, such as permeability, cracking index, and condensation parameters, and proceed to the next calculation cycle;

[0109] When anomalies such as water inrush or crack penetration are detected, a local mesh refinement and high-frequency iteration mechanism is triggered to improve simulation accuracy.

[0110] Step 4: Boundary and initial condition settings.

[0111] Groundwater boundary: The upstream and downstream water heads are set based on the water level measured on site;

[0112] Stress boundaries: The upper and lower boundaries of the formation are defined as vertical force boundaries, while the horizontal plane adopts symmetrical or fixed boundaries;

[0113] Grout boundary: The grouting orifice is a constant pressure or constant flow velocity boundary, and the far field is a zero flux boundary.

[0114] Step 5: Model validation and dynamic updates.

[0115] After the initial model calculations were completed, the model's prediction accuracy was verified by comparing measured grouting pressure, grout diffusion range, and pore pressure response data. During subsequent operation, sensor data was assimilated in real time using a Kalman filter algorithm to dynamically update permeability, fracture channel permeability, and grout diffusion parameters in the multi-physics model.

[0116] In this embodiment, the Ensemble Kalman Filter (EnKF) algorithm is used as the core method for data assimilation. Combined with real-time monitoring data, the parameters of the multi-physics model are dynamically corrected. The three-dimensional grouting diffusion simulation is updated every 30 seconds to simulate the grout penetration, diffusion and solidification process, ensuring that the calculation results are consistent with the actual working conditions.

[0117] Real-time monitoring data includes geological environmental parameters, grouting material characteristics, and dynamic data of the construction process. Among them, geological environmental parameters include information such as stratum distribution, layer thickness, and fracture development characteristics; dynamic parameters of the real-time state of the soil and rock environment such as pore water pressure, permeability coefficient, and stratum deformation; and geological environmental influencing factors such as groundwater level changes, surrounding rock stress evolution, and soil physical and mechanical properties.

[0118] The characteristics of grouting materials include the density, viscosity, setting time, and rheological properties of different grouting materials such as cement-based grout and chemical grout; the curing rate of the grout under different temperature and humidity conditions, which is used to analyze its influence on diffusion behavior; and the permeability of the grout in formations with different pore sizes, which is used to evaluate the characteristics of the water-stop curtain formed after the grout has cured.

[0119] Dynamic data during construction includes grouting pressure, flow rate, grout diffusion range, and other construction process data; as well as information such as formation deformation and pore water pressure changes collected per second.

[0120] The collected sensor data, such as grouting pressure, flow rate, and surrounding rock displacement, are synchronized according to timestamps, and lagging and missing values ​​are processed and time-aligned. Sliding window Z-score test and interquartile range (IQR) analysis are used to remove abnormal fluctuation points and anomalies. Data of different dimensions are normalized, and linear interpolation or spline interpolation is used to fill data gaps to ensure data integrity and usability.

[0121] The system state is represented as a vector form of a set of parameters to be estimated and physical variables. The parameters to be estimated include formation permeability, slurry diffusion coefficient, and surrounding rock modulus, while the physical variables include pore pressure field, concentration field, and deformation field. Each component of the system state is represented by multiple ensemble members (100-300 sets), whose initial values ​​are derived from historical data or expert experience. Here, each ensemble member refers to a sample set of possible values ​​for the system parameters (i.e., the parameters to be estimated) and the variables (i.e., the state variables) under uncertain conditions. Each member corresponds to a set of initial simulation values, such as permeability, modulus, and diffusion coefficient, and participates synchronously in the numerical simulation and parameter correction process over time.

[0122] During the prediction phase, each ensemble member is input into the current numerical model, i.e., the multiphysics model, and run for one step, such as a 30-second simulation, to obtain the predicted state of the multiphysics model. Then, observation mapping is performed, projecting field monitoring data, such as sensor-measured slurry front positions and pore pressure changes, into the model space. Error calculation is then performed between the predicted and actual data. After error calculation, the system constructs a Kalman gain matrix based on the residual vector and the state covariance matrix, and performs Bayesian estimation correction on the state vector of each ensemble member. The updated parameters include key physical properties such as formation permeability, slurry diffusion coefficient, and elastic modulus, while state variables include pore pressure field, diffusion boundary, and formation deformation response. All corrected ensemble states are then used as initial conditions for the next simulation input into the multiphysics model, forming a complete data assimilation loop.

[0123] The Ensemble Kalman Filter (EnKF) method updates the system state vector in real time. For the i-th ensemble member, the update of its state vector is achieved by the following formula:

[0124] x i a =x i f +K ( y i -Hx i f )

[0125] in, x i fTo predict the state, x i a To update the status, y i For the measured observations, H is the observation mapping matrix, and K is the Kalman gain.

[0126] The Kalman gain matrix K is calculated by the following formula:

[0127] K=P f H T ( HP f H T +R ) -1

[0128] in, P f To predict the covariance matrix, R is the observation error covariance matrix, and the superscript T indicates transpose.

[0129] Through the aforementioned Bayesian update mechanism, the system completes iterative correction of the state vector at each time step, ensuring that the simulation results closely match the actual working conditions at the construction site.

[0130] The multiphysics model parameters are updated synchronously every second, especially key parameters such as formation permeability, slurry diffusion rate, and formation deformation modulus; the simulation results are dynamically corrected to ensure that they reflect the current working conditions.

[0131] In this embodiment, to further improve the adaptability and prediction accuracy of the multiphysics model under complex geological conditions, parameter optimization and self-learning are performed based on data assimilation, specifically including:

[0132] Historical data archiving and tagging: Construction monitoring data of completed grouting sections in multiple projects, such as grouting pressure, flow rate, diffusion boundary, pore pressure change, and surrounding rock deformation, are classified and archived with corresponding formation parameters such as lithology, permeability, fracture rate, and saturation, as well as control effects such as optimized pressure, flow rate, and grouting saturation. A three-element tag structure of "input parameters - grouting response - control effect" is constructed for subsequent machine learning model training.

[0133] This embodiment assists in parameter optimization in two stages: First, in the modeling stage of a new engineering project, based on geological survey data, the trained MK-SVM or random forest regression model recommends initial physical parameters such as permeability and diffusion coefficient as basic initial values; Second, in the EnKF prediction-correction loop, the machine learning model is called to generate "prior parameter correction" based on real-time monitoring features, dynamically adjusting the prediction input, improving model stability and accelerating convergence speed.

[0134] Parameter Initialization and Model Building: An initial model is proposed → Permeability, diffusion coefficient, etc., are recommended using MK-SVM or Random Forest Regression models. Specifically, in the initial modeling stage of a new engineering project, similar geological types are selected from the historical database based on its exploration data. Initial physical parameters such as permeability and diffusion coefficient are recommended and preset using a multi-kernel support vector machine (MK-SVM) or random forest regression model. Machine learning-generated "prior parameter corrections" can be inserted between the EnKF prediction-correction loops to improve the stability of the assimilated initial values ​​and reduce convergence time. The input to the MK-SVM or random forest regression model is geological exploration data (lithology, fracture rate, etc.); the output is a set of initial parameters that may more closely reflect the actual situation; replacing traditional empirical settings, improving the quality of the model's starting point; providing the basis for the "first prediction," and enhancing the stability of the initial values.

[0135] Data assimilation and Real-Time Control Mechanism: Introducing EnKF → Using machine learning to generate prior parameters before prediction. Specifically, the mapping relationship between input parameters and control effects is trained in the machine learning model. Preliminary control parameters are generated based on the trained machine learning model. These preliminary control parameters are inserted as "prior parameter corrections" in the EnKF prediction-correction loop to dynamically adjust the initial conditions of the numerical model, improving model stability and convergence speed. Before each round of EnKF prediction, the prior distribution of the parameters to be estimated is dynamically adjusted using the machine learning model, shortening the convergence time of EnKF and reducing parameter oscillations.

[0136] All historical data undergoes unified preprocessing, including time-series normalization, interpolation completion, unit conversion, and anomaly removal, to ensure data structure consistency.

[0137] Model self-learning training mechanism: After each construction phase, the corrected model parameters and actual observation data are added to the historical case library. Using deep neural networks (DNN) or LSTM temporal neural networks, the model's parameter responses under different formation / grout combinations are trained. When modeling new projects or new sections, the most similar historical model is automatically called and fine-tuned based on formation information and grout type, quickly completing the construction and correction of the initial model.

[0138] Enhanced generalization mechanism: The system incorporates a geological type labeling system, such as sand and gravel layers, silt with gravel, and weathered rock, enabling the machine learning model to recognize cross-project conditions. For novel grouting solutions or construction strategies, an incremental learning mechanism is used to dynamically expand the model's parameter space.

[0139] In this embodiment, considering the time correlation in the slurry coagulation process, a time-varying coagulation model that characterizes the changes in the physical properties of the slurry over time is integrated. The time-varying coagulation model is embedded as an external function in the time step into the modified multiphysics model, so that the numerical simulation can adaptively adjust the slurry flow state.

[0140] Specifically, considering the time-dependent nature of the slurry setting process, an integrated time-varying setting model is used to characterize the changes in the physical properties of the slurry over time, particularly the evolution of its rheological properties during the diffusion-hardening-retention stage. Slurry setting is considered a nonlinear evolution process related to time, temperature, and environmental hydration conditions.

[0141] This embodiment adopts an empirical-constitutive coupling model, combining experimental data and numerical functions to establish the following basic model:

[0142]

[0143] Where μ(t) represents the viscosity of the slurry at time t, μ0 is the initial viscosity, and α and β are empirical fitting parameters, representing the coagulation growth rate and acceleration factor; t c This is the critical time for condensation.

[0144] The time-varying coagulation model is embedded as an external function in the time step into the finite element-discrete element coupled simulation module of the multiphysics model. In each simulation step, the system updates the current slurry viscosity and diffusion control coefficient according to the time evolution relationship, so that the coupled calculation unit can adaptively adjust the flow state of the slurry in the heterogeneous medium. This ensures that the diffusion front is not only controlled by the slurry injection pressure and formation permeability, but also affected by the slurry's own coagulation process, thus more closely matching the actual grouting hardening process.

[0145] The updated slurry viscosity is used to control the drag term in the Darcy flow model:

[0146]

[0147] Where μ(t) is the slurry viscosity, v(t) is the slurry seepage velocity vector, k is the formation permeability coefficient, and ▽P is the pressure gradient.

[0148] The formula for the evolution of viscosity over time:

[0149]

[0150] Where μ0 is the initial viscosity and b is the solidification growth coefficient, which affects the flow rate and on / off state of the slurry in the fractured medium.

[0151] In slurry diffusion simulation, viscosity is a key factor controlling the diffusion coefficient D:

[0152]

[0153] in, This is the slurry diffusion capacity coefficient, which is related to the pore structure.

[0154] The updated diffusion coefficient is used in the diffusion term of the convection-diffusion-reaction equation to determine the advance rate and diffusion range of the slurry concentration front.

[0155] This embodiment can also calculate key indicators such as diffusion range, grouting saturation, and potential grout leakage area in real time, and predict the diffusion path in the next 5 minutes.

[0156] Specifically, 1. Real-time calculation of diffusion range: In three-dimensional space, based on the slurry concentration C of each unit... i (t) Determine whether the effective value C for grouting has been reached. thresh Cells that meet the conditions are recorded in the diffusion region:

[0157]

[0158] Where cell represents a cell element, Ω represents a diffusion region, and i represents the i-th cell element.

[0159] 2. Grouting saturation calculation: Grouting saturation S(t) is defined as the ratio of the diffused volume to the target grouting volume.

[0160]

[0161] Among them, V spread (t) represents the volume of the diffused region at the current time; V design To design the grouting volume.

[0162] 3. Identification of potential slurry escape areas.

[0163] Calculation method: Calculate the normal velocity at the diffusion boundary: v n = v·n ; v n This indicates the normal flow velocity of the fluid along a surface or boundary. v It is the velocity vector of the fluid. n It is the unit vector in the normal direction.

[0164] If v n >v max v max If the threshold is set, it is marked as a "potential slurry escape region".

[0165] 4-5 minute diffusion path prediction method.

[0166] Input conditions: current assimilated state vector, current diffusion boundary, flow rate change trend and diffusion range in the last 1-3 minutes, grouting saturation, potential grout escape area and other key indicators, grout time-varying viscosity prediction based on time-varying coagulation model, and fracture structure.

[0167] Prediction method: Using an updated multiphysics model, the simulation is rapidly advanced 300 steps (5 minutes) at short time steps (e.g., per second) with fixed parameters; GPU-accelerated parallel solution technology is used, and viscosity is updated in each iteration;

[0168] The diffusion path sequence for the next 5 minutes is obtained:

[0169] {Ω spread (t+1),Ω spread (t+2),...,Ω spread (t+300)}.

[0170] This embodiment utilizes a BIM+GIS integrated platform to dynamically display diffusion cloud maps, risk heat maps, equipment status, and geological profiles. If the grout front deviates from the design path, the programmable logic controller automatically generates optimized pressure and flow control parameters and sends them to the grouting equipment, achieving real-time closed-loop control; it also displays the current and predicted grouting full areas, substandard areas, and recommended construction parameters.

[0171] Specifically, when the system identifies areas where the grout front deviates from the design path or where grout saturation is insufficient, the deviation radius is extracted: δ r =R design -R current And establish the following multi-objective optimization function:

[0172]

[0173] Among them, R sim R represents the predicted diffusion radius from the simulation. target For the design diffusion range, Q is the grouting flow rate, P is the grouting pressure, Psafe is the formation safety pressure limit, λ1 and λ2 are the weighting coefficients for balancing control efficiency and safety, and δ r R is the deviation radius, representing the distance difference between the designed grouting front line and the actual grouting front line. design It is the design diffusion radius, which is the ideal grouting front position predicted based on the design scheme and geological conditions, R. current This is the current diffusion radius, i.e., the current actual grouting front position obtained through real-time monitoring. This optimization problem is solved using fast iterative algorithms such as particle swarm optimization or genetic algorithms to find the optimal combination of grouting parameters, outputting the optimal grouting pressure P. ∗ Optimal grouting flow rate Q ∗As the setpoint for the grouting equipment, the PLC control system sends the data to the intelligent grouting equipment in real time to complete the automatic parameter adjustment.

[0174] This embodiment continuously analyzes real-time data streams to identify abnormal conditions such as sudden water inrush, overpressure, and unexpected deformation. It then initiates localized mesh refinement calculations to detail the model resolution and improve simulation accuracy. By accessing a historical case knowledge base, it matches similar situations within 45 seconds and outputs emergency response suggestions. Combined with a formation stress evolution model, it predicts potential construction risks and provides early warnings.

[0175] This embodiment can also accurately identify uncovered areas and over-injected areas by tracking the grouting diffusion boundary in real time; automatically generate pressure-flow optimization schemes to avoid grout waste, improve grouting fullness and seepage control effect, and improve grouting accuracy and material utilization efficiency.

[0176] Specifically, the slurry concentration field C(x,t) generated based on real-time simulation and the designed grouting area Ω target Based on spatial relationships, two types of key regions are identified using threshold determination and region classification methods:

[0177] 1. Uncovered area: defined as an area within the target region where the current grouting concentration has not reached the effective grouting threshold C. thresh All positions, the expression is:

[0178]

[0179] Among them, Ω ungrouted For the grouting blind zone, x represents the spatial coordinate, Ω target To design the grouting area.

[0180] 2. Over-injection zone is defined as the location outside the target area where the slurry has spread and exceeded the threshold, i.e.:

[0181]

[0182] Among them, Ω overgrouted This is an over-diffusion zone.

[0183] In addition, the system incorporates the normal flow velocity at the diffusion boundary: v n = v·n The system determines whether there is a continuous outward expansion trend of slurry overflow. All identified areas are highlighted in the BIM+GIS platform, with uncovered areas marked in red to guide supplementary slurry design, and over-slurry areas marked in yellow to assist in pressure reduction or flow adjustment. The above area identification is automatically updated every 30 seconds, providing spatial data support for the control module to generate parameter optimization suggestions.

[0184] This implementation introduces dynamic assimilation algorithms such as Ensemble Kalman Filter (EnKF) to enable the numerical model to be continuously updated based on real-time monitoring data, thus solving the problem of lag in traditional simulations. It achieves "synchronous linkage" between numerical simulation and on-site working conditions, significantly improving the simulation's relevance to reality and its guiding value, and realizing second-level synchronization between numerical simulation and actual construction.

[0185] This implementation constructs a closed-loop system of "monitoring-simulation-control", which integrates sensor acquisition, model updating, result visualization and equipment control, opens up the data chain and forms a complete closed-loop control; it realizes the automated process from perception to decision-making and then to execution, reduces human intervention and improves the level of intelligent construction.

[0186] This implementation accurately identifies uncovered and over-injected areas by tracking the grouting diffusion boundary in real time; the system automatically generates pressure-flow optimization schemes to avoid grout waste, improve grouting fullness and seepage control effect, and enhance grouting accuracy and material utilization efficiency.

[0187] This implementation addresses abnormal situations such as water inrush and abnormal deformation of surrounding rock. The system can trigger local mesh refinement and case matching, providing emergency countermeasures within 45 seconds; effectively shortening response time, improving engineering safety, and enhancing the ability to respond quickly to emergencies.

[0188] Example 2

[0189] The purpose of this embodiment is to provide a multi-physics field real-time assimilation simulation and control system for the tunnel grouting process, including:

[0190] The module is used to construct a multiphysics model of groundwater-stratum stress-grout diffusion coupling based on the initial physical parameters of the tunnel.

[0191] The correction module is used to dynamically correct the parameters of the multi-physics model by using ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting.

[0192] The simulation module is used to consider the time correlation in the slurry coagulation process, integrates a time-varying coagulation model that describes the changes in the physical properties of the slurry over time, and embeds the time-varying coagulation model as an external function in the time step into the modified multiphysics model, so that the numerical simulation can adaptively adjust the slurry flow state.

[0193] The control module is used to generate control parameters for tunnel grouting based on dynamic simulation results, thereby achieving closed-loop control of tunnel grouting.

[0194] In further embodiments, the following is also provided:

[0195] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor. When executed by the processor, the computer instructions perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0196] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0197] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0198] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0199] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0200] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0201] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for real-time multi-physics field assimilation simulation and control of tunnel grouting process, characterized in that, include: Based on the initial physical parameters of the tunnel, a multiphysics model coupling groundwater, stratum stress, and slurry diffusion is constructed. By using ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting, the parameters of the multi-physics model are dynamically corrected. Considering the time correlation in the slurry coagulation process, a time-varying coagulation model that characterizes the changes in the physical properties of the slurry over time is integrated. The time-varying coagulation model is embedded as an external function in the time step into the finite element-discrete element coupled simulation module in the multiphysics model. In each simulation step, the current slurry viscosity and diffusion control coefficient are updated according to the time evolution relationship, so that the coupled calculation unit can adaptively adjust the flow state of the slurry in the heterogeneous medium. The basic model is established using an empirical-constitutive coupling model: Where μ(t) represents the viscosity of the slurry at time t, μ0 is the initial viscosity, and α and β are empirical fitting parameters, representing the condensation growth rate and acceleration factor; t c This is the critical time for condensation. The updated slurry viscosity is used to control the drag term in the Darcy flow model: Where μ(t) is the slurry viscosity, v(t) is the slurry seepage velocity vector, and k is the formation permeability coefficient. P is the pressure gradient; The formula for the evolution of viscosity over time: Where μ0 is the initial viscosity; b is the solidification growth coefficient, which affects the flow rate and on / off state of the slurry in the fractured medium. In slurry diffusion simulation, viscosity is a key factor controlling the diffusion coefficient D: in, This is the slurry diffusion capacity coefficient, which is related to the pore structure; The updated diffusion coefficient is used in the diffusion term of the convection-diffusion-reaction equation to determine the advance rate and diffusion range of the slurry concentration front. Control parameters for tunnel grouting are generated based on dynamic simulation results, enabling closed-loop control of tunnel grouting.

2. The method for real-time multiphysics field assimilation simulation and control of tunnel grouting process as described in claim 1, characterized in that, Based on the initial physical parameters of the tunnel, a finite element-discrete element coupled algorithm is used to simulate the diffusion behavior of grout in complex fractured media, and a multiphysics model coupling groundwater-formation stress-grout diffusion is constructed, specifically as follows: The three-dimensional region model, which includes the surrounding rock in front of the tunnel face, the construction area, and the far-field layer, is meshed. Based on the divided three-dimensional region model, the groundwater seepage field, the formation stress field and the slurry diffusion field are coupled by multiple physics and numerically solved. Based on the solution results, an initial multiphysics model coupling groundwater-formation stress-slurry diffusion is constructed.

3. The method for real-time multiphysics field assimilation simulation and control of tunnel grouting process as described in claim 1, characterized in that, By utilizing ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting, the parameters of the multi-physics model are dynamically corrected, specifically as follows: Each set member of the system state is input into the current multiphysics model for simulation to obtain the predicted state; Real-time monitoring data during tunnel grouting is projected into a multiphysics model space, and the error is calculated by comparing it with the prediction results. Based on the residual vector and the state covariance matrix, a Kalman gain matrix is ​​constructed, and Bayesian estimation correction is performed on the state vector of each set member. The corrected set state will be used as the initial condition for the next simulation and input into the multiphysics model to dynamically correct the parameters of the multiphysics model, forming a complete data assimilation loop.

4. The method for real-time multi-physics field assimilation simulation and control of tunnel grouting process as described in claim 1, characterized in that, Also includes: Based on historical construction data, the multiphysics model was optimized, specifically as follows: Based on the construction monitoring data of the completed grouting section and the corresponding stratum parameters, a three-element label of "input parameters – grouting response – control effect" is constructed. Based on the constructed ternary labels, the machine learning model is trained, and the trained machine learning model is used to predict the prior control parameters of the multiphysics model. The prior distribution of the parameters to be estimated in the ensemble Kalman filter is dynamically adjusted using prior control parameters.

5. The method for real-time multi-physics field assimilation simulation and control of tunnel grouting process as described in claim 1, characterized in that, By considering slurry coagulation as a nonlinear evolution process related to time, temperature, and environmental hydration conditions, a time-varying coagulation model is constructed by combining empirical fitting parameters and numerical functions.

6. The method for real-time multiphysics field assimilation simulation and control of tunnel grouting process as described in claim 1, characterized in that, Also includes: Using the updated multiphysics model, future diffusion paths are predicted, specifically: The grout concentration of each unit is used to determine whether the effective grouting value has been reached, and the units that meet the conditions are recorded in the diffusion zone. Based on the current state vector, current diffusion boundary, flow rate change trend, slurry time-varying viscosity prediction results, and fracture structure, the updated multiphysics model is used to predict future diffusion paths.

7. The method for real-time multi-physics field assimilation simulation and control of tunnel grouting process as described in claim 6, characterized in that, Also includes: By tracking the grouting diffusion boundary in real time, uncovered areas and over-grouting areas are identified, specifically: Based on the spatial relationship between the slurry concentration field generated by real-time simulation and the designed grouting area, two types of key areas are identified by threshold determination and area classification methods.

8. A multi-physics field real-time assimilation simulation and control system for tunnel grouting process, characterized in that, include: The module is used to construct a multiphysics model of groundwater-stratum stress-grout diffusion coupling based on the initial physical parameters of the tunnel. The correction module is used to dynamically correct the parameters of the multi-physics model by using ensemble Kalman filtering and combining real-time monitoring data during tunnel grouting. The simulation module is used to consider the time correlation in the slurry coagulation process. It integrates a time-varying coagulation model that describes the changes in the physical properties of the slurry over time. The time-varying coagulation model is embedded as an external function in the time step into the finite element-discrete element coupled simulation module in the multiphysics model. In each simulation step, the current slurry viscosity and diffusion control coefficient are updated according to the time evolution relationship, so that the coupled calculation unit can adaptively adjust the flow state of the slurry in the heterogeneous medium. The basic model is established using an empirical-constitutive coupling model: Where μ(t) represents the viscosity of the slurry at time t, μ0 is the initial viscosity, and α and β are empirical fitting parameters, representing the condensation growth rate and acceleration factor; t c This is the critical time for condensation. The updated slurry viscosity is used to control the drag term in the Darcy flow model: Where μ(t) is the slurry viscosity, v(t) is the slurry seepage velocity vector, and k is the formation permeability coefficient. P is the pressure gradient; The formula for the evolution of viscosity over time: Where μ0 is the initial viscosity; b is the solidification growth coefficient, which affects the flow rate and on / off state of the slurry in the fractured medium. In slurry diffusion simulation, viscosity is a key factor controlling the diffusion coefficient D: in, This is the slurry diffusion capacity coefficient, which is related to the pore structure; The updated diffusion coefficient is used in the diffusion term of the convection-diffusion-reaction equation to determine the advance rate and diffusion range of the slurry concentration front. The control module is used to generate control parameters for tunnel grouting based on dynamic simulation results, thereby achieving closed-loop control of tunnel grouting.

9. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-7.

Images