A system and method for foundation pit curtain formation and groundwater regulation based on digital twin and risk feedback control

Abstract

This invention relates to the field of extra-large and extra-deep foundation pit construction technology, specifically a foundation pit curtain formation and groundwater control system and method based on digital twin and risk feedback control. The system includes a physical execution layer, a multi-source sensing layer, a digital twin and intelligent decision-making layer, and a closed-loop control layer; it constructs an intelligent closed-loop collaborative control system integrating sensing, simulation, decision-making, and execution. This achieves precise curtain formation, proactive risk control, and intelligent engineering decision-making, ultimately forming a high-quality water-tight curtain efficiently and economically under complex hydrogeological conditions while ensuring foundation pit safety.

Description

Technical Field

[0001] This invention relates to the field of construction technology for extra-large and extra-deep foundation pits, specifically to a foundation pit curtain formation and groundwater regulation system and method based on digital twin and risk feedback control. Background Technology

[0002] As urban underground space development continues to advance towards ultra-deep and ultra-large scales, the hydrogeological conditions faced by foundation pit engineering are becoming increasingly complex, placing higher demands on construction technology. Taking the foundation of a vertical continuous casting machine in the steel industry as an example, this ultra-large and ultra-deep concrete structure, typically buried at depths of -20 meters or more, is the foundation for ensuring the long-term stable and efficient operation of critical equipment; its construction quality directly determines the equipment's lifespan. Especially when conducting such construction in high-water-level areas such as coastal regions, abundant groundwater poses a severe threat to the structure's durability. Therefore, it is essential to achieve comprehensive breakthroughs in key technologies such as deep foundation pit excavation and support, and efficient and precise dewatering within confined spaces.

[0003] However, existing technologies suffer from a "monitoring-control separation" limitation: on the one hand, advanced curtain grouting methods, such as those induced by microorganisms to precipitate calcium carbonate, lack effective process quality control, making the grouting process a "black box" and resulting in uncontrollable uniformity, continuity, and final impermeability of the curtain; on the other hand, while existing intelligent monitoring systems can provide risk warnings based on sensor data, they cannot translate these warnings into direct, precise control commands for construction equipment, essentially "diagnosing but not diagnosing." Neither approach addresses the core issue of "how to dynamically optimize construction and control parameters based on real-time engineering feedback to achieve the best engineering results." Especially in deep soft soil and highly confined aquifer strata, the curtain formation process itself disturbs the groundwater field, which in turn affects grout diffusion and reaction efficiency. How to simultaneously construct a reliable impermeable curtain and precisely and actively regulate the groundwater environment within a dynamically coupled and complex system to ensure the safety of the foundation pit and its surrounding environment has become a critical technological bottleneck that urgently needs to be overcome. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the purpose of this invention is to provide an integrated system and method that can perceive the curtain formation status and foundation pit environmental risks in real time, and dynamically adjust curtain construction parameters and groundwater control strategies through intelligent algorithms. This enables precise curtain formation, proactive risk control, and intelligent engineering decision-making, ultimately achieving efficient and economical formation of a high-quality waterproof curtain under complex hydrogeological conditions and ensuring foundation pit safety.

[0005] The technical solution adopted by this invention to solve its technical problem is:

[0006] A foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control is proposed. The system constructs an intelligent closed-loop collaborative control system integrating perception, simulation, decision-making, and execution, specifically including the following levels:

[0007] The physical execution layer includes an array of microbial-induced calcium carbonate deposition grouting holes and a multi-functional control well array deployed in the strata surrounding the foundation pit. The grouting hole array is used to inject microbial liquid, calcium source solution and urea solution. The multi-functional control well array is connected to hydraulic control equipment and has the ability to switch between pumping mode and reinjection mode.

[0008] The multi-source sensing layer includes distributed temperature sensing optical fibers, inclinometer arrays and piezometer networks embedded in the soil of the target curtain formation area and the foundation pit influence area, which are used to collect temperature field data, deformation field data and pore water pressure field data in real time.

[0009] Digital twins and intelligent decision-making layers include:

[0010] A multi-physics coupled digital twin model is used to assimilate data collected by a multi-source sensing layer in real time to update the model state, and to predict the future state through a time-progression solution method.

[0011] The calcium carbonate formation inversion module, based on temperature field data monitored by distributed temperature sensing fiber optics, uses a physical calibration model or a convolutional neural network data-driven model to invert and calculate the rate and spatial distribution of calcium carbonate formation within the curtain.

[0012] The risk assessment module is used to calculate the curtain underdevelopment index based on the rate and spatial distribution of calcium carbonate formation within the curtain, and to calculate the pit seepage risk coefficient based on the future state predicted by the digital twin model.

[0013] The model predictive controller, based on the nonlinear model predictive control method, takes the multi-objective collaborative optimization of curtain thickness uniformity, seepage risk and construction cost as the criterion, and solves the optimal control instruction set for future moments through rolling optimization.

[0014] The closed-loop control layer is used to automatically send control command sets to the physical execution layer to dynamically adjust the grouting pressure, slurry ratio, and pumping or reinjection flow rate of each grouting hole and each multi-functional control well.

[0015] As a preferred embodiment, a further technical solution of the present invention is:

[0016] Preferably, the process of the calcium carbonate formation inversion module calculating the rate and spatial distribution of calcium carbonate formation within the curtain includes:

[0017] Step A1: Data Preprocessing

[0018] Raw temperature field data is obtained from distributed temperature sensing optical fibers. The raw temperature field data is then cleaned and calibrated to remove abnormal noise and unify timestamps and spatial coordinates.

[0019] Step A2: Calculate the spatiotemporal variation characteristics of the temperature field

[0020] From the preprocessed temperature field data, the temperature rise rate of each monitoring point is extracted as the key feature. The temperature rise rate is the amount of temperature change per unit time.

[0021] Step A3: Select and apply the inversion model

[0022] Input the temperature rise rate into the preset physical calibration model. The physical calibration model uses the following formula to calculate the mass of calcium carbonate formed per unit volume of soil:

[0023] ;

[0024] in, Indicates the mass of calcium carbonate produced. Indicates soil density. Indicates the specific heat capacity of soil. This indicates the temperature difference at the monitoring points before and after grouting. This represents the molar enthalpy change of the microbially induced calcium carbonate deposition reaction. Indicates soil porosity;

[0025] Step A4: Output the spatiotemporal distribution of calcium carbonate formation

[0026] The physical calibration model outputs the calcium carbonate formation rate and amount at different times at various spatial locations within the curtain, quantitatively describing the curtain formation process in a three-dimensional spatial and temporal form.

[0027] Preferably, the process by which the model predictive controller solves for the optimal control instruction set for future time periods through rolling optimization in each control cycle includes:

[0028] Step B1: Measuring Sensing

[0029] At time k, the actual output data of the system at the current time is obtained from the multi-source sensing layer. The actual output data should include at least the temperature, pore water pressure and soil deformation data of each monitoring point.

[0030] Step B2: Model Prediction

[0031] Actual output data Input a multiphysics coupled digital twin model, and use the multiphysics coupled digital twin model to predict the future in the time domain based on the current state. The system output sequence includes the cumulative amount of calcium carbonate, total head, pore water pressure, and volumetric strain at each future spatial location.

[0032] Step B3: Scrolling Optimization

[0033] Solve a nonlinear optimization problem in a finite time domain: with future control over the time domain. The sequence of control variables within the system is the optimization object, with the minimum objective function J. The control variables include the injection rate and slurry ratio of each grouting hole, and the pumping or reinjection flow rate of each multi-functional control well. The optimization process satisfies the physical limit constraints of the equipment and the engineering safety threshold constraints.

[0034] Minimum objective function J:

[0035] ;

[0036] in, This represents the spatial uniformity variance of the curtain thickness within the prediction time domain. This represents the seepage risk coefficient of the foundation pit at key points within the predicted time domain. This represents the sum of the energy consumption of all grouting wells injected and the energy consumption of all multi-functional control wells pumping or reinjecting water within the predicted time domain. , , This represents a preset weighting coefficient used to coordinate the priorities among curtain wall quality, engineering safety, and economy.

[0037] The sequence of control variables is represented as follows:

[0038] ;

[0039] in, This represents the injection rate and slurry ratio of the i-th grouting hole, where the slurry ratio is the volume ratio of microbial inoculum solution, calcium source solution, and urea solution. This represents the flow rate of the j-th multi-functional control well; a positive value indicates pumping mode, and a negative value indicates reinjection mode. Indicates the total number of grouting holes; Indicates the total number of multi-functional control wells;

[0040] The constraints include: the injection rate of each grouting hole is between zero and the maximum design injection rate; the total water head at any point in the foundation pit is between the preset minimum water level and the preset maximum water level; and the soil deformation rate does not exceed the preset safety threshold.

[0041] Step B4: Execution Control

[0042] The first step of the optimal control variable sequence obtained in step B3 is used as the control command at the current moment and sent to the physical execution layer for execution through the closed-loop control layer.

[0043] Step B5: Scroll Time Shift

[0044] At the next sampling time, repeat steps B1 to B4 to achieve forward scrolling of the control time domain.

[0045] Preferably, the multiphysics coupled digital twin model includes the following sub-models:

[0046] The groundwater seepage sub-model is used to describe the flow law of groundwater in porous media. It considers the disturbance of the seepage field caused by the pumping or recharge of the multi-functional control well and the local source term caused by the grouting pressure of the grouting hole, and outputs the total head and pore water velocity at each spatial location.

[0047] The solute transport sub-model is used to describe the convection, diffusion, and reaction consumption processes of microbial inoculum, calcium ions, and urea in groundwater. The pore water flow velocity is obtained from the groundwater seepage sub-model as input, and the output is the concentration of various solutes at each spatial location.

[0048] A kinetic sub-model for microbial-induced calcium carbonate deposition reaction is used to describe the kinetic process of calcium ions reacting with urease to form calcium carbonate. The calcium ion concentration is obtained from the solute transport sub-model as input, and the formation rate and cumulative amount of calcium carbonate are output, and the soil porosity is updated accordingly.

[0049] The soil elastoplastic deformation sub-model is used to describe the changes in effective stress and corresponding volumetric strain of soil caused by microbial-induced calcium carbonate deposition grouting and groundwater pumping. The pore water pressure is obtained from the groundwater seepage sub-model to calculate the effective stress, and the cumulative amount of calcium carbonate generated is obtained from the reaction kinetics sub-model to update the soil stiffness. The volumetric strain at each spatial location is output.

[0050] Preferably, the process of updating the model state of a multiphysics coupled digital twin model includes:

[0051] Step C1: Current state generation

[0052] Starting from the optimal state of the previous time step, the multiphysics coupled digital twin model is used to predict the state at the next time step according to the change patterns described by each sub-model, thus obtaining the predicted state vector for the current time step. The predicted state vector includes at least the total water head, calcium ion concentration, calcium carbonate accumulation, soil porosity, and effective stress at the spatial location; at the same time, the state covariance matrix updated at the previous time step is propagated through the model to obtain the predicted state covariance matrix at the current time step.

[0053] Step C2: Obtain measured data

[0054] Obtain the measured data vector at the current moment from the multi-source sensing layer. It includes at least the pore water pressure value of each piezometer monitoring point, the temperature value of each temperature monitoring point, and the deformation value of each inclinometer monitoring point.

[0055] Step C3: Calculate the Kalman gain

[0056] Calculate the Kalman gain matrix using the following formula:

[0057] ;

[0058] in, H represents the predicted state covariance matrix; H represents the observation matrix, which is used to map the state vector to the observation space; R represents the observation noise covariance matrix, which is a diagonal matrix, and the diagonal elements are the variances of the measurement noise of each sensor. The variances are determined in advance through sensor calibration experiments.

[0059] Step C4: State fusion update

[0060] The predicted state vector and the measured data vector are weighted and fused according to the following formula to obtain the assimilated optimal state estimate:

[0061] ;

[0062] in, This represents the updated optimal state estimation vector; Represents the predicted state vector; Represents the Kalman gain matrix; Represents a vector of measured data;

[0063] Step: Update the error covariance

[0064] Update the state covariance matrix according to the following formula:

[0065] ;

[0066] in, This represents the updated state covariance matrix; Represents the identity matrix;

[0067] The updated state covariance matrix is ​​then used as the output covariance matrix for the predicted state generation in step C1 at the next time step.

[0068] Preferably, the process of predicting future states using a time-progression solution method in a multiphysics coupled digital twin model includes:

[0069] Step D1: Set initial conditions

[0070] The optimal state estimate obtained through data assimilation at the current moment As the starting point for prediction calculations;

[0071] Step D2: Set the future control scheme

[0072] Obtain the control input sequence in the future control time domain. The control input sequence includes the injection rate and slurry ratio of each grouting hole at each time, and the pumping or reinjection flow rate of each multi-functional control well at each time.

[0073] Step D3: Solving by Time Advancement

[0074] The time-progressed solution of the multiphysics coupled digital twin model is performed according to the following sub-steps:

[0075] Step D3-1: Initialization

[0076] The optimal state estimate obtained through data assimilation at the current time k is used as the initial state for the prediction calculation, and the current time is set. ;

[0077] Step D3-2: Advance step by step over time

[0078] For each time step in the prediction time domain Repeat the following operations:

[0079] (a) Determine the control input for the current step: Based on the future control input sequence obtained in step D2, determine the injection rate and slurry ratio of each grouting hole and the pumping or reinjection flow rate of each multi-functional control well within the current time step;

[0080] (b) Calculate the updated state value at the current step: Input the total head and control input sequence in the current state value into the groundwater seepage sub-model to obtain the total head and pore water velocity components after one time step;

[0081] The pore water velocity components and solute concentrations in the current state value obtained by the solution are input into the solute transporter model to obtain the solute concentrations after one time step, including at least the calcium ion concentration.

[0082] The calcium ion concentration obtained from the solution is input into the microbial-induced calcium carbonate deposition reaction kinetics sub-model to obtain the cumulative calcium carbonate generation after one time step.

[0083] Based on the total head, the pore water pressure is solved, and the pore water pressure and the cumulative amount of calcium carbonate generated are input into the soil elastic-plastic deformation sub-model to obtain the volumetric strain after one time step.

[0084] (c) Update the state to the next time step: Use the calculated state values ​​after one time step as the state values ​​for the next time step;

[0085] (d) Update the coupling parameters between sub-models: update the soil porosity based on the newly generated cumulative amount of calcium carbonate, update the permeability coefficient based on the updated soil porosity, and feed the updated permeability coefficient back into the seepage calculation for the next time step;

[0086] Step D3-3: Loop until the end of the prediction time domain

[0087] Update the time to the previous moment plus one time step, use the updated state value as the current state, and repeat step D3-2 until the end of the prediction time domain is reached.

[0088] The system state prediction values ​​at each time point in the prediction time domain are output and used as future state prediction results for the model predictive controller to calculate the objective function.

[0089] Preferably, the process by which the risk assessment module calculates the curtain underdevelopment index includes:

[0090] Step E1: Identification of Valid Curtain Area

[0091] The spatial distribution of calcium carbonate production is obtained from the calcium carbonate production inversion module, and the calcium carbonate content threshold is obtained. Regions that meet the calcium carbonate content threshold are marked as effective curtain regions.

[0092] Step E2: Extract curtain thickness

[0093] In the direction perpendicular to the sidewall of the foundation pit, for each horizontal position, the distance from the outer boundary to the inner boundary of the effective curtain area is defined as the curtain thickness of that area. ;

[0094] Step E3: Calculate the thickness statistics

[0095] Calculate the average curtain thickness at all locations using the following formula. and standard deviation :

[0096] , ;

[0097] Where N represents the total number of curtain thickness sampling points; This represents the curtain thickness at the i-th data collection point; This represents the average thickness of the curtain. The standard deviation of the curtain thickness;

[0098] Step E4: Calculate the effective area coverage rate

[0099] Calculate the effective curtain area using the following formula. With the design target curtain area The ratio:

[0100] ;

[0101] Step E5: Calculate the curtain underdevelopment index

[0102] Calculate the curtain underdevelopment index using the following formula. :

[0103] ;

[0104] in, It represents the coefficient of variation in thickness, reflecting the uniformity of the curtain thickness; , This indicates the set weighting coefficient.

[0105] Preferably, the process of calculating the seepage risk coefficient of the foundation pit by the risk assessment module includes:

[0106] Step F1: Extract the predicted total head distribution

[0107] Extract the total head h at each spatial location in the predicted time domain from the future state predicted by the multiphysics coupled digital twin model;

[0108] Step F2: Calculate the hydraulic gradient

[0109] Calculate the hydraulic gradient at each location using the following formula. :

[0110] ;

[0111] in, , , This represents the rate of change of total head in the three directions of x, y, and z.

[0112] Step F3: Identify the out-of-limit area

[0113] Setting a safety threshold for hydraulic gradients The statistical hydraulic gradient satisfies regional volume ;

[0114] Step F4: Determine the volume of the critical influence zone

[0115] Determine the total volume of the critical influence zone of the foundation pit. Specifically, it is defined as the volume of soil within a depth of one foundation pit below the bottom of the foundation pit.

[0116] Step F5: Determine the maximum hydraulic gradient

[0117] Determine the maximum hydraulic gradient within the monitoring area. ;

[0118] Step F6: Calculate the seepage risk factor

[0119] Calculate the seepage risk factor using the following formula. :

[0120] .

[0121] This invention also discloses a method for the formation of foundation pit curtain and groundwater regulation based on digital twin and risk feedback control, the specific steps of which are as follows:

[0122] Multi-dimensional physical field data of the construction site are collected in real time through a multi-source sensing layer;

[0123] By leveraging digital twins and intelligent decision-making layers, multidimensional physical field data is assimilated to update the digital twin model. Based on the state of the digital twin model, risk assessment and rolling optimization of model prediction are performed to generate a set of control instructions.

[0124] Through the closed-loop control layer, the control command set is sent to the physical execution layer to coordinate the adjustment of grouting operation parameters and groundwater regulation strategies, thus forming a closed-loop control.

[0125] The present invention, which adopts the above technical solution, has the following prominent features compared with the prior art:

[0126] (1) It pioneers a new paradigm of integrated intelligent closed-loop control: This invention is not a simple superposition of existing monitoring and construction technologies, but rather a creative integration and closed-loop coupling of biochemical reaction process control (MICP) and geotechnical hydraulic process control (groundwater pumping and injection) through a unified digital twin platform and model predictive control algorithm. While forming a curtain, the injection pressure and volume of MICP grouting activities also become an effective means of regulating the local groundwater field; conversely, the regulation decision based on deformation and seepage risk (such as initiating recharge) also directly optimizes the grout migration path and reaction environment. This scheme, which unifies the design and synergistic optimization of control loops for two completely different physical processes, is groundbreaking in the field of geotechnical engineering.

[0127] (2) Solved the quality "black box" problem of traditional MIP process: By introducing distributed temperature sensing and advanced inversion algorithm, the "visual" online monitoring and quantitative evaluation of the hidden underground biomineralization reaction process was realized for the first time, so that the quality of the curtain changed from "result acceptance" to "process controllability", ensuring the reliable formation of the seepage prevention body.

[0128] (3) It has achieved a leap from passive early warning to active feedforward control: Through model prediction, the system can adjust the construction parameters and hydraulic boundary conditions in advance before obvious signs of foundation pit deformation or seepage risk appear, changing "post-event remediation" to "pre-event prevention", which greatly improves the safety margin of the project.

[0129] (4) Possesses strong adaptive optimization capabilities: The system can automatically cope with complex factors such as formation heterogeneity and reaction rate uncertainty. By updating the model through real-time data assimilation and continuously optimizing control commands, it always dynamically tracks the optimal construction path, significantly reducing the reliance on manual experience. It is estimated that the system can improve construction efficiency by about 15%-30% and reduce material waste and energy consumption by about 20%. Attached Figure Description

[0130] Figure 1 This is a schematic diagram of the system structure in an embodiment of the present invention;

[0131] Figure 2 This is a schematic diagram of the rolling optimization principle in an embodiment of the present invention;

[0132] Figure 3 This is a schematic diagram of the process for retrieving calcium carbonate production in an embodiment of the present invention. Detailed Implementation

[0133] The present invention will be further illustrated below with reference to specific embodiments. The purpose of this illustration is solely to provide a better understanding of the invention. Therefore, the examples given do not limit the scope of protection of the present invention.

[0134] like Figure 1 As shown in the figure, this embodiment presents a foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control. The system constructs an intelligent closed-loop collaborative control system integrating perception, simulation, decision-making, and execution, specifically including the following levels:

[0135] The physical execution layer includes an array of microbial-induced calcium carbonate deposition grouting holes and a multi-functional control well array deployed in the strata surrounding the foundation pit. The grouting hole array is used to inject microbial liquid, calcium source solution and urea solution. The multi-functional control well array is connected to hydraulic control equipment and has the ability to switch between pumping mode and reinjection mode.

[0136] The multi-source sensing layer includes distributed temperature sensing optical fibers, inclinometer arrays and piezometer networks embedded in the soil of the target curtain formation area and the foundation pit influence area, which are used to collect temperature field data, deformation field data and pore water pressure field data in real time.

[0137] Digital twins and intelligent decision-making layers include:

[0138] A multi-physics coupled digital twin model is used to assimilate data collected by a multi-source sensing layer in real time to update the model state, and to predict the future state through a time-progression solution method.

[0139] The calcium carbonate formation inversion module, based on temperature field data monitored by distributed temperature sensing fiber optics, uses a physical calibration model or a convolutional neural network data-driven model to invert and calculate the rate and spatial distribution of calcium carbonate formation within the curtain.

[0140] The risk assessment module is used to calculate the curtain underdevelopment index based on the rate and spatial distribution of calcium carbonate formation within the curtain, and to calculate the pit seepage risk coefficient based on the future state predicted by the digital twin model.

[0141] The model predictive controller, based on the nonlinear model predictive control method, takes the multi-objective collaborative optimization of curtain thickness uniformity, seepage risk and construction cost as the criterion, and solves the optimal control instruction set for future moments through rolling optimization.

[0142] The closed-loop control layer is used to automatically send control command sets to the physical execution layer to dynamically adjust the grouting pressure, slurry ratio, and pumping or reinjection flow rate of each grouting hole and each multi-functional control well.

[0143] During implementation, see Figure 3 The calcium carbonate formation inversion module calculates the rate and spatial distribution of calcium carbonate formation within the curtain wall, including:

[0144] Step A1: Data Preprocessing

[0145] Raw temperature field data is obtained from distributed temperature sensing optical fibers. The raw temperature field data is then cleaned and calibrated to remove abnormal noise and unify timestamps and spatial coordinates.

[0146] Step A2: Calculate the spatiotemporal variation characteristics of the temperature field

[0147] From the preprocessed temperature field data, the temperature rise rate of each monitoring point is extracted as the key feature. The temperature rise rate is the amount of temperature change per unit time.

[0148] Step A3: Select and apply the inversion model

[0149] Input the temperature rise rate into the preset physical calibration model. The physical calibration model uses the following formula to calculate the mass of calcium carbonate formed per unit volume of soil:

[0150] ;

[0151] in, Indicates the mass of calcium carbonate produced. Indicates soil density. Indicates the specific heat capacity of soil. This indicates the temperature difference at the monitoring points before and after grouting. This represents the molar enthalpy change of the microbially induced calcium carbonate deposition reaction. Indicates soil porosity;

[0152] Step A4: Output the spatiotemporal distribution of calcium carbonate formation

[0153] The physical calibration model outputs the calcium carbonate formation rate and amount at different times at various spatial locations within the curtain, quantitatively describing the curtain formation process in a three-dimensional spatial and temporal form.

[0154] In another implementation, when inverting the calculation of the rate and spatial distribution of calcium carbonate formation within the curtain using a convolutional neural network data-driven model, a black-box model trained using historical experimental or field data can be used based on machine learning (such as convolutional neural networks CNN or long short-term memory networks LSTM) to achieve a nonlinear mapping from temperature characteristics to formation amount.

[0155] In implementation, the model predictive controller, within each control cycle, see [reference needed]. Figure 2 The process of finding the optimal control instruction set for future time steps through rolling optimization includes:

[0156] Step B1: Measuring Sensing

[0157] At time k, the actual output data of the system at the current time is obtained from the multi-source sensing layer. The actual output data should include at least the temperature, pore water pressure and soil deformation data of each monitoring point.

[0158] Step B2: Model Prediction

[0159] Actual output data Input a multiphysics coupled digital twin model, and use the multiphysics coupled digital twin model to predict the future in the time domain based on the current state. The system output sequence includes the cumulative amount of calcium carbonate at spatial locations at future time points to evaluate the formation quality of the curtain, the total water head to evaluate the seepage risk of the foundation pit, the pore water pressure to monitor the risk of sudden surge, and the volumetric strain to monitor the deformation of the foundation pit.

[0160] Step B3: Scrolling Optimization

[0161] Solve a nonlinear optimization problem in a finite time domain: with future control over the time domain. The sequence of control variables within the system is the optimization object, with the minimum objective function J. The control variables include the injection rate and slurry ratio of each grouting hole, and the pumping or reinjection flow rate of each multi-functional control well. The optimization process satisfies the physical limit constraints of the equipment and the engineering safety threshold constraints.

[0162] Minimum objective function J:

[0163] ;

[0164] in, The variance of the spatial uniformity of the curtain thickness in the prediction time domain is calculated as follows: at the end of the prediction time domain, from the spatial distribution of calcium carbonate accumulation, the area with calcium carbonate content greater than a preset threshold (taken as 50 kg per cubic meter) is set as the effective curtain. The thickness value of the curtain body in the direction perpendicular to the side wall of the foundation pit is extracted, and the variance of these thickness values ​​is calculated. The smaller the variance, the more uniform and continuous the curtain is.

[0165] This represents the seepage risk coefficient of the foundation pit at key points in the prediction time domain. It is calculated as follows: the hydraulic gradient at each location is calculated from the spatial distribution of the total water head, and the probability of the hydraulic gradient exceeding the preset safety threshold (with a value of 1.0) is statistically analyzed. The higher the probability, the greater the seepage risk.

[0166] This represents the sum of the energy consumption of all grouting wells injected and the energy consumption of all multi-functional control wells pumping or reinjecting water within the predicted time domain. , , This represents a preset weighting coefficient used to coordinate the priorities among curtain wall quality, engineering safety, and economy.

[0167] The sequence of control variables is represented as follows:

[0168] ;

[0169] in, This represents the injection rate and slurry ratio of the i-th grouting hole, where the slurry ratio is the volume ratio of microbial inoculum solution, calcium source solution, and urea solution. This represents the flow rate of the j-th multi-functional control well; a positive value indicates pumping mode, and a negative value indicates reinjection mode. Indicates the total number of grouting holes; Indicates the total number of multi-functional control wells;

[0170] The constraints include: the injection rate of each grouting hole is between zero and the maximum design injection rate; the total water head at any point in the foundation pit is between the preset minimum water level and the preset maximum water level; and the soil deformation rate does not exceed the preset safety threshold.

[0171] Step B4: Execution Control

[0172] The first step of the optimal control variable sequence obtained in step B3 is used as the control command at the current moment and sent to the physical execution layer for execution through the closed-loop control layer.

[0173] Step B5: Scroll Time Shift

[0174] At the next sampling time, repeat steps B1 to B4 to achieve forward scrolling of the control time domain.

[0175] In practice, the model predictive controller uses a sequential quadratic programming algorithm to solve the objective function, with a control period ranging from five to fifteen minutes, and the prediction time domain... The value is set to six hours, controlling the time domain. The value is two hours.

[0176] In practice, the multiphysics coupled digital twin model includes the following sub-models:

[0177] The groundwater seepage sub-model is used to describe the flow law of groundwater in porous media. It considers the disturbance of the seepage field caused by the pumping or recharge of the multi-functional control well and the local source term caused by the grouting pressure of the grouting hole, and outputs the total head and pore water velocity at each spatial location.

[0178] Specifically, the groundwater seepage sub-model can be described using both the unsteady seepage equation and Darcy's law:

[0179] Unsteady seepage equation:

[0180] ;

[0181] Darcy's Law:

[0182] , , ;

[0183] in, Indicates the water storage coefficient; , , These are the permeability coefficients in the x, y, and z directions, respectively. This indicates the source and sink items caused by pumping or reinjection from a multi-functional control well; This indicates a local source term caused by the grouting pressure in the grouting hole; , , These represent the pore water velocity components in three directions, respectively. , , These represent the hydraulic gradients in three directions.

[0184] The solute transport sub-model is used to describe the convection, diffusion, and reaction consumption processes of microbial inoculum, calcium ions, and urea in groundwater. The pore water flow velocity is obtained from the groundwater seepage sub-model as input, and the output is the concentration of various solutes at each spatial location.

[0185] Specifically, the solute transporter model can be described using convection-dispersion-reaction equations:

[0186]

[0187] ;

[0188] in, This indicates the concentration of the i-th solute, which includes microbial inoculum, calcium ions, and urea. , , Indicates the hydrodynamic dispersion coefficient in three directions; , , The pore water flow velocity in the three directions is represented by Darcy's law from the groundwater seepage sub-model; The rate of consumption or formation of the i-th solute in the microbial-induced calcium carbonate deposition reaction, wherein the rate of calcium ion consumption can be expressed as: , Represents the reaction rate constant. This indicates the concentration of urease activity.

[0189] A kinetic sub-model for microbial-induced calcium carbonate deposition reaction is used to describe the kinetic process of calcium ions reacting with urease to form calcium carbonate. The calcium ion concentration is obtained from the solute transport sub-model as input, and the formation rate and cumulative amount of calcium carbonate are output, and the soil porosity is updated accordingly.

[0190] Specifically, the kinetic sub-model of microbial-induced calcium carbonate deposition can be achieved using reaction kinetic equations:

[0191] ;

[0192] in, This indicates the amount of calcium carbonate produced. This represents the temperature and pH inhibition factor, with a value ranging from 0 to 1.

[0193] The soil elastic-plastic deformation sub-model is used to describe the changes in effective stress and corresponding volumetric strain of soil caused by microbial-induced calcium carbonate deposition grouting and groundwater pumping. The pore water pressure is obtained from the groundwater seepage sub-model to calculate the effective stress, and the cumulative amount of calcium carbonate generated is obtained from the reaction kinetics sub-model to update the soil stiffness. The volumetric strain at each spatial location is output.

[0194] Specifically, the soil elastic-plastic deformation sub-model is described using the modified Cambridge model:

[0195] ;

[0196] in, Indicates volumetric strain; Indicates the average effective stress; Indicates the compression index; Indicates the rebound index; Indicates the volume ratio.

[0197] In implementation, the coupling relationship between the sub-models is as follows:

[0198] The amount of calcium carbonate produced reduces the soil porosity, and the reduced porosity leads to a decrease in the permeability coefficient.

[0199] The decrease in permeability coefficient has a reaction effect on the seepage field, changing the pore water velocity and pore water pressure;

[0200] Changes in pore water pressure alter the effective stress, causing soil deformation.

[0201] Soil deformation reacts to soil porosity, further affecting seepage and solute transport.

[0202] Specifically, the updated soil porosity is expressed as follows: ;

[0203] The updated permeability coefficient is expressed as: ;

[0204] Effective stress calculation formula: ;

[0205] The relationship between pore water pressure and total head is as follows: ;

[0206] In the formula, Indicates the initial porosity; Indicates the volume factor; This indicates the cumulative amount of calcium carbonate produced. This represents the updated penetration coefficient; Indicates the initial permeability coefficient; This represents the exponential coefficient, with a value between 2.5 and 3.5. This represents the total stress, calculated from the weight of the overlying soil. Pore ​​water pressure represents the pressure exerted on the water within the pores of the soil. The density of water, It is the acceleration due to gravity; Total head; The position head represents the height of the calculation point relative to the reference datum. Pressure head indicates the height of the water column that pressure can drive.

[0207] In practice, the process of updating the model state of a multiphysics coupled digital twin model includes:

[0208] Step C1: Current state generation

[0209] Starting from the optimal state of the previous time step, the multiphysics coupled digital twin model is used to predict the state at the next time step according to the change patterns described by each sub-model, thus obtaining the predicted state vector for the current time step. The predicted state vector includes at least the total water head, calcium ion concentration, calcium carbonate accumulation, soil porosity, and effective stress at the spatial location; at the same time, the state covariance matrix updated at the previous time step is propagated through the model to obtain the predicted state covariance matrix at the current time step.

[0210] Step C2: Obtain measured data

[0211] Obtain the measured data vector at the current moment from the multi-source sensing layer. It should include at least the pore water pressure values ​​at each piezometer monitoring point. Temperature values ​​at each temperature monitoring point and deformation values ​​at each inclinometer monitoring point ;

[0212] Step C3: Calculate the Kalman gain

[0213] Calculate the Kalman gain matrix using the following formula:

[0214] ;

[0215] in, H represents the predicted state covariance matrix; H represents the observation matrix, used to map the state vector to the observation space. In this invention, the non-zero elements of the observation matrix only appear in the pore water pressure corresponding to the total head h. The mapping position, the specific mapping relationship is as follows: , Let represent the density of water, g represent the gravitational acceleration, z represent the position head, and R represent the observation noise covariance matrix. R is a diagonal matrix, with the diagonal elements representing the variance of the measurement noise of each sensor. The variance is predetermined through sensor calibration experiments; for the piezometer, the value is 0.25; for the temperature sensing fiber optic cable, the value is 0.01; and for the inclinometer, the value is 0.01.

[0216] Step C4: State fusion update

[0217] The predicted state vector and the measured data vector are weighted and fused according to the following formula to obtain the assimilated optimal state estimate:

[0218] ;

[0219] in, This represents the updated optimal state estimation vector; Represents the predicted state vector; Represents the Kalman gain matrix; Represents a vector of measured data;

[0220] Step: Update the error covariance

[0221] Update the state covariance matrix according to the following formula:

[0222] ;

[0223] in, This represents the updated state covariance matrix; Represents the identity matrix;

[0224] The updated state covariance matrix is ​​then used as the output covariance matrix for the predicted state generation in step C1 at the next time step.

[0225] In implementation, the process of predicting future states using a time-progressive solution method in a multiphysics coupled digital twin model includes:

[0226] Step D1: Set initial conditions

[0227] The optimal state estimate obtained through data assimilation at the current moment As the starting point for prediction calculations;

[0228] Step D2: Set the future control scheme

[0229] Obtain the control input sequence in the future control time domain. The control input sequence includes the injection rate and slurry ratio of each grouting hole at each time, and the pumping or reinjection flow rate of each multi-functional control well at each time.

[0230] Step D3: Time-Advanced Solution: Starting from the current state value, the solution is advanced step by step according to the changing patterns of each sub-model, until the end of the prediction time domain is reached. Within each time step, the groundwater seepage sub-model (unsteady seepage equation + Darcy's law), solute transport sub-model (convection-dispersion-reaction equation), MICP reaction kinetics sub-model (reaction kinetic equation), and soil elastoplastic deformation sub-model (modified Cambridge model) are solved sequentially, and the coupling parameters are updated.

[0231] Specifically, the time-progression solution of the multiphysics coupled digital twin model is performed according to the following sub-steps:

[0232] Step D3-1: Initialization

[0233] The optimal state estimate obtained through data assimilation at the current time k is used as the initial state for the prediction calculation, and the current time is set. ;

[0234] Step D3-2: Advance step by step over time

[0235] For each time step in the prediction time domain Repeat the following operations:

[0236] (a) Determine the control input for the current step: Based on the future control input sequence obtained in step D2, determine the injection rate and slurry ratio of each grouting hole and the pumping or reinjection flow rate of each multi-functional control well within the current time step;

[0237] (b) Calculate the updated state value at the current step: Input the total head and control input sequence in the current state value into the groundwater seepage sub-model to obtain the total head and pore water velocity components after one time step;

[0238] The pore water velocity components and solute concentrations in the current state value obtained by the solution are input into the solute transporter model to obtain the solute concentrations after one time step, including at least the calcium ion concentration.

[0239] The calcium ion concentration obtained from the solution is input into the microbial-induced calcium carbonate deposition reaction kinetics sub-model to obtain the cumulative calcium carbonate generation after one time step.

[0240] Based on the total head, the pore water pressure is solved, and the pore water pressure and the cumulative amount of calcium carbonate generated are input into the soil elastic-plastic deformation sub-model to obtain the volumetric strain after one time step.

[0241] (c) Update the state to the next time step: Use the calculated state values ​​after one time step as the state values ​​for the next time step;

[0242] (d) Update the coupling parameters between sub-models: update the soil porosity based on the newly generated cumulative amount of calcium carbonate, update the permeability coefficient based on the updated soil porosity, and feed the updated permeability coefficient back into the seepage calculation for the next time step;

[0243] Step D3-3: Loop until the end of the prediction time domain

[0244] Update the time to the previous moment plus one time step, use the updated state value as the current state, and repeat step D3-2 until the end of the prediction time domain is reached.

[0245] The system state prediction values ​​at each time point in the prediction time domain are output and used as future state prediction results for the model predictive controller to calculate the objective function.

[0246] In implementation, the process of calculating the curtain underdevelopment index in the risk assessment module includes:

[0247] Step E1: Identification of Valid Curtain Area

[0248] The spatial distribution of calcium carbonate production is obtained from the calcium carbonate production inversion module, and the calcium carbonate content threshold is obtained. Regions that meet the calcium carbonate content threshold are marked as effective curtain regions.

[0249] Step E2: Extract curtain thickness

[0250] In the direction perpendicular to the sidewall of the foundation pit, for each horizontal position, the distance from the outer boundary to the inner boundary of the effective curtain area is defined as the curtain thickness of that area. ;

[0251] Step E3: Calculate the thickness statistics

[0252] Calculate the average curtain thickness at all locations using the following formula. and standard deviation :

[0253] , ;

[0254] Where N represents the total number of curtain thickness sampling points; This represents the curtain thickness at the i-th data collection point; This represents the average thickness of the curtain. The standard deviation of the curtain thickness;

[0255] Step E4: Calculate the effective area coverage rate

[0256] Calculate the effective curtain area using the following formula. With the design target curtain area The ratio:

[0257] ;

[0258] Step E5: Calculate the curtain underdevelopment index

[0259] Calculate the curtain underdevelopment index using the following formula. :

[0260] ;

[0261] in, It represents the coefficient of variation in thickness, reflecting the uniformity of the curtain thickness; , This indicates the set weighting coefficient, which can be 0.6 or 0.4 respectively.

[0262] During implementation, the process of calculating the seepage risk coefficient of the foundation pit in the risk assessment module includes:

[0263] Step F1: Extract the predicted total head distribution

[0264] Extract the total head h at each spatial location in the predicted time domain from the future state predicted by the multiphysics coupled digital twin model;

[0265] Step F2: Calculate the hydraulic gradient

[0266] Calculate the hydraulic gradient at each location using the following formula. :

[0267] ;

[0268] in, , , This represents the rate of change of total head in the three directions of x, y, and z.

[0269] Step F3: Identify the out-of-limit area

[0270] Setting a safety threshold for hydraulic gradients It can be set to 1.0, and the statistical hydraulic gradient satisfies... regional volume ;

[0271] Step F4: Determine the volume of the critical influence zone

[0272] Determine the total volume of the critical influence zone of the foundation pit. Specifically, it is defined as the volume of soil within a depth of one foundation pit below the bottom of the foundation pit.

[0273] Step F5: Determine the maximum hydraulic gradient

[0274] Determine the maximum hydraulic gradient within the monitoring area. ;

[0275] Step F6: Calculate the seepage risk factor

[0276] Calculate the seepage risk factor using the following formula. :

[0277] .

[0278] In practice, two risk indicators can be compared with preset thresholds: less than 0.3 is considered green and safe, 0.3 to 0.6 is a yellow warning, and greater than or equal to 0.6 is a red warning. When either indicator reaches a yellow or red warning level, the model prediction controller is triggered to perform rolling optimization adjustments.

[0279] This invention also discloses a method for the formation of foundation pit curtain and groundwater regulation based on digital twin and risk feedback control, the specific steps of which are as follows:

[0280] Multi-dimensional physical field data of the construction site are collected in real time through a multi-source sensing layer;

[0281] By leveraging digital twins and intelligent decision-making layers, multidimensional physical field data is assimilated to update the digital twin model. Based on the state of the digital twin model, risk assessment and rolling optimization of model prediction are performed to generate a set of control instructions.

[0282] Through the closed-loop control layer, the control command set is sent to the physical execution layer to coordinate the adjustment of grouting operation parameters and groundwater regulation strategies, thus forming a closed-loop control.

[0283] Example

[0284] The intelligent foundation pit curtain formation and groundwater control system includes a physical system deployed on-site and a cloud computing platform located in the project command center.

[0285] The physical system includes:

[0286] Grouting hole array: arranged in a ring around the perimeter of the foundation pit according to the design spacing, each grouting hole is equipped with an independently controlled precision grouting pump and a three-liquid (bacterial solution, calcium solution, urea solution) mixing and injection device.

[0287] Multifunctional control well array: It is arranged at intervals on the inner and outer sides of the grouting hole array. Each well is equipped with a reversible submersible pump and a high-precision flow meter, which can seamlessly switch between pumping and reinjection modes according to instructions.

[0288] Sensor network: including distributed temperature sensing optical fibers laid out in layers along the center line of the designed curtain body, and inclinometers and piezometers buried in the foundation pit retaining structure and key soil.

[0289] The cloud computing platform includes a data receiving and preprocessing module, a digital twin model updating module, a risk assessment module, a nonlinear MPC optimization solution module, and an encrypted command delivery module. Among these, the Zisheng model couples non-Darcy flow, solute transport considering the inhibitory effect of biochemical reactions, and the Cambridge elastoplastic model.

[0290] On the fifth day of curtain wall construction, the sensor network detected that the temperature rise rate of the DTS fiber in section A was only 60% of that in the adjacent section B. At the same time, the piezometer on the outside of this section showed that the pore water pressure was accumulating at a rate of 0.5 kPa per hour.

[0291] The data is transmitted to the cloud platform in real time. The model update module quickly assimilates the data. The Zisheng model simulation shows that in section A, due to the low permeability coefficient of the local soil layer, the grout diffusion is hindered, the reaction is weak, and a local water-blocking zone is formed, leading to an increase in water pressure.

[0292] The risk assessment module outputs that the "insufficient curtain development index" and "local surge risk coefficient" for this section both exceed the yellow warning line.

[0293] The MPC optimization module initiates rolling calculations. Within the current control period (e.g., the next 6 hours), the main optimization variables are the adjustment of the operation of the three grouting wells in section A and the two nearby control wells. After solving, the system generates and issues the following instructions:

[0294] Increase the grouting pressure of the three grouting holes in section A by 15% and increase the bacterial solution concentration by 20% to enhance its penetration and reactivity.

[0295] Switch the nearest control well on the outer side of section A to "low-speed reinjection" mode, reinjecting at a rate of 2 cubic meters per hour. This aims to slightly raise the water head on the periphery, drive the slurry to diffuse inward (to the weak reaction zone), and alleviate pressure accumulation.

[0296] After the command is executed, the model shows that the temperature field in section A begins to develop in a balanced manner and the pore water pressure tends to stabilize through subsequent sensing data feedback. The system automatically enters the next round of optimization cycle until the risk indicators in this area are eliminated and the curtain quality parameters meet the standards.

[0297] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the present invention. All equivalent changes made based on the description and drawings of the present invention are included within the scope of the present invention.

Claims

1. A foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control, characterized in that, The system constructs an intelligent closed-loop collaborative control system integrating perception, simulation, decision-making, and execution, specifically including the following levels: The physical execution layer includes an array of microbial-induced calcium carbonate deposition grouting holes and a multi-functional control well array deployed in the strata surrounding the foundation pit. The grouting hole array is used to inject microbial liquid, calcium source solution and urea solution. The multi-functional control well array is connected to hydraulic control equipment and has the ability to switch between pumping mode and reinjection mode. The multi-source sensing layer includes distributed temperature sensing optical fibers, inclinometer arrays and piezometer networks embedded in the soil of the target curtain formation area and the foundation pit influence area, which are used to collect temperature field data, deformation field data and pore water pressure field data in real time. Digital twins and intelligent decision-making layers include: A multi-physics coupled digital twin model is used to assimilate data collected by a multi-source sensing layer in real time to update the model state, and to predict the future state through a time-progression solution method. The calcium carbonate formation inversion module, based on temperature field data monitored by distributed temperature sensing fiber optics, uses a physical calibration model or a convolutional neural network data-driven model to invert and calculate the rate and spatial distribution of calcium carbonate formation within the curtain. The risk assessment module is used to calculate the curtain underdevelopment index based on the rate and spatial distribution of calcium carbonate formation within the curtain, and to calculate the pit seepage risk coefficient based on the future state predicted by the digital twin model. The model predictive controller, based on the nonlinear model predictive control method, takes the multi-objective collaborative optimization of curtain thickness uniformity, seepage risk and construction cost as the criterion, and solves the optimal control instruction set for future moments through rolling optimization. The closed-loop control layer is used to automatically send control command sets to the physical execution layer to dynamically adjust the grouting pressure, slurry ratio, and pumping or reinjection flow rate of each grouting hole and each multi-functional control well.

2. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 1, characterized in that, The calcium carbonate formation inversion module calculates the rate and spatial distribution of calcium carbonate formation within the curtain wall, including: Step A1: Data Preprocessing Raw temperature field data is obtained from distributed temperature sensing optical fibers. The raw temperature field data is then cleaned and calibrated to remove abnormal noise and unify timestamps and spatial coordinates. Step A2: Calculate the spatiotemporal variation characteristics of the temperature field From the preprocessed temperature field data, the temperature rise rate of each monitoring point is extracted as a key feature. The temperature rise rate is the amount of temperature change per unit time. Step A3: Select and apply the inversion model Input the temperature rise rate into the preset physical calibration model. The physical calibration model uses the following formula to calculate the mass of calcium carbonate formed per unit volume of soil: ； in, Indicates the mass of calcium carbonate produced. Indicates soil density. Indicates the specific heat capacity of soil. This indicates the temperature difference at the monitoring points before and after grouting. This represents the molar enthalpy change of the microbially induced calcium carbonate deposition reaction. Indicates the porosity of the soil; Step A4: Output the spatiotemporal distribution of calcium carbonate formation The physical calibration model outputs the calcium carbonate formation rate and amount at different times at various spatial locations within the curtain, quantitatively describing the curtain formation process in a three-dimensional spatial and temporal form.

3. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 2, characterized in that, The process by which a model predictive controller (MMC) seeks the optimal set of control instructions for future time periods through rolling optimization within each control cycle includes: Step B1: Measuring Sensation At time k, the actual output data of the system at the current time is obtained from the multi-source sensing layer. The actual output data should include at least the temperature, pore water pressure and soil deformation data of each monitoring point. Step B2: Model Prediction Actual output data Input a multiphysics coupled digital twin model, and use the multiphysics coupled digital twin model to predict the future in the time domain based on the current state. The system output sequence includes the cumulative amount of calcium carbonate, total head, pore water pressure, and volumetric strain at each future spatial location. Step B3: Scrolling Optimization Solve a nonlinear optimization problem in a finite time domain: with future control over the time domain. The sequence of control variables within the system is the optimization object, with the minimum objective function J. The control variables include the injection rate and slurry ratio of each grouting hole, and the pumping or reinjection flow rate of each multi-functional control well. The optimization process satisfies the physical limit constraints of the equipment and the engineering safety threshold constraints. Minimum objective function J: ； in, This represents the spatial uniformity variance of the curtain thickness within the prediction time domain. This represents the seepage risk coefficient of the foundation pit at key points within the predicted time domain. This represents the sum of the energy consumption of all grouting wells injected and the energy consumption of all multi-functional control wells pumping or reinjecting water within the predicted time domain. , , This represents a preset weighting coefficient used to coordinate the priorities among curtain wall quality, engineering safety, and economy. The sequence of control variables is represented as follows: ； in, This represents the injection rate and slurry ratio of the i-th grouting hole, where the slurry ratio is the volume ratio of microbial inoculum solution, calcium source solution, and urea solution. This represents the flow rate of the j-th multi-functional control well; a positive value indicates pumping mode, and a negative value indicates reinjection mode. Indicates the total number of grouting holes; Indicates the total number of multi-functional control wells; The constraints include: the injection rate of each grouting hole is between zero and the maximum design injection rate; the total water head at any point in the foundation pit is between the preset minimum water level and the preset maximum water level; and the soil deformation rate does not exceed the preset safety threshold. Step B4: Execution Control The first step of the optimal control variable sequence obtained in step B3 is used as the control command at the current moment and sent to the physical execution layer for execution through the closed-loop control layer. Step B5: Scroll Time Shift At the next sampling time, repeat steps B1 to B4 to achieve forward scrolling of the control time domain.

4. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 3, characterized in that, The multiphysics coupled digital twin model includes the following sub-models: The groundwater seepage sub-model is used to describe the flow law of groundwater in porous media. It considers the disturbance of the seepage field caused by the pumping or recharge of the multi-functional control well and the local source term caused by the grouting pressure of the grouting hole, and outputs the total head and pore water velocity at each spatial location. The solute transport sub-model is used to describe the convection, diffusion, and reaction consumption processes of microbial inoculum, calcium ions, and urea in groundwater. The pore water flow velocity is obtained from the groundwater seepage sub-model as input, and the output is the concentration of various solutes at each spatial location. A kinetic sub-model for microbial-induced calcium carbonate deposition reaction is used to describe the kinetic process of calcium ions reacting with urease to form calcium carbonate. The calcium ion concentration is obtained from the solute transport sub-model as input, and the formation rate and cumulative amount of calcium carbonate are output, and the soil porosity is updated accordingly. The soil elastoplastic deformation sub-model is used to describe the changes in effective stress and corresponding volumetric strain of soil caused by microbial-induced calcium carbonate deposition grouting and groundwater pumping. The pore water pressure is obtained from the groundwater seepage sub-model to calculate the effective stress, and the cumulative amount of calcium carbonate generated is obtained from the reaction kinetics sub-model to update the soil stiffness. The volumetric strain at each spatial location is output.

5. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 4, characterized in that, The process of updating the model state in a multiphysics coupled digital twin model includes: Step C1: Current state generation Starting from the optimal state of the previous time step, the multiphysics coupled digital twin model is used to predict the state at the next time step according to the change patterns described by each sub-model, thus obtaining the predicted state vector for the current time step. The predicted state vector includes at least the total water head, calcium ion concentration, calcium carbonate accumulation, soil porosity, and effective stress at the spatial location; at the same time, the state covariance matrix updated at the previous time step is propagated through the model to obtain the predicted state covariance matrix at the current time step. Step C2: Obtain measured data Obtain the measured data vector at the current moment from the multi-source sensing layer. It includes at least the pore water pressure value of each piezometer monitoring point, the temperature value of each temperature monitoring point, and the deformation value of each inclinometer monitoring point. Step C3: Calculate the Kalman gain Calculate the Kalman gain matrix using the following formula: ； in, H represents the predicted state covariance matrix; H represents the observation matrix, which is used to map the state vector to the observation space; R represents the observation noise covariance matrix, which is a diagonal matrix, and the diagonal elements are the variances of the measurement noise of each sensor. The variances are determined in advance through sensor calibration experiments. Step C4: State fusion update The predicted state vector and the measured data vector are weighted and fused according to the following formula to obtain the assimilated optimal state estimate: ； in, This represents the updated optimal state estimation vector; Represents the predicted state vector; Represents the Kalman gain matrix; Represents a vector of measured data; Step: Update the error covariance Update the state covariance matrix according to the following formula: ； in, This represents the updated state covariance matrix; Represents the identity matrix; The updated state covariance matrix is ​​then used as the output covariance matrix for the predicted state generation in step C1 at the next time step.

6. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 5, characterized in that, The process by which a multiphysics coupled digital twin model predicts future states using a time-progressive solution method includes: Step D1: Set initial conditions The optimal state estimate obtained through data assimilation at the current moment As the starting point for prediction calculations; Step D2: Set the future control scheme Obtain the control input sequence in the future control time domain. The control input sequence includes the injection rate and slurry ratio of each grouting hole at each time, and the pumping or reinjection flow rate of each multi-functional control well at each time. Step D3: Solving by Time Advancement The time-progressed solution of the multiphysics coupled digital twin model is performed according to the following sub-steps: Step D3-1: Initialization The optimal state estimate obtained through data assimilation at the current time k is used as the initial state for the prediction calculation, and the current time is set. ; Step D3-2: Advance step by step over time For each time step in the prediction time domain Repeat the following operations: (a) Determine the control input for the current step: Based on the future control input sequence obtained in step D2, determine the injection rate and slurry ratio of each grouting hole and the pumping or reinjection flow rate of each multi-functional control well within the current time step; (b) Calculate the updated state value at the current step: Input the total head and control input sequence in the current state value into the groundwater seepage sub-model to obtain the total head and pore water velocity components after one time step; The pore water velocity components and solute concentrations in the current state value obtained by the solution are input into the solute transporter model to obtain the solute concentrations after one time step, including at least the calcium ion concentration. The calcium ion concentration obtained from the solution is input into the microbial-induced calcium carbonate deposition reaction kinetics sub-model to obtain the cumulative calcium carbonate generation after one time step. Based on the total head, the pore water pressure is solved, and the pore water pressure and the cumulative amount of calcium carbonate generated are input into the soil elastic-plastic deformation sub-model to obtain the volumetric strain after one time step. (c) Update the state to the next time step: Use the calculated state values ​​after one time step as the state values ​​for the next time step; (d) Update the coupling parameters between sub-models: update the soil porosity based on the newly generated cumulative amount of calcium carbonate, update the permeability coefficient based on the updated soil porosity, and feed the updated permeability coefficient back into the seepage calculation for the next time step; Step D3-3: Loop until the end of the prediction time domain Update the time to the previous moment plus one time step, use the updated state value as the current state, and repeat step D3-2 until the end of the prediction time domain is reached. The system state prediction values ​​at each time point in the prediction time domain are output and used as future state prediction results for the model predictive controller to calculate the objective function.

7. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 6, characterized in that, The process of calculating the curtain underdevelopment index in the risk assessment module includes: Step E1: Identification of Valid Curtain Area The spatial distribution of calcium carbonate production is obtained from the calcium carbonate production inversion module, and the calcium carbonate content threshold is obtained. Regions that meet the calcium carbonate content threshold are marked as effective curtain regions. Step E2: Extract curtain thickness In the direction perpendicular to the sidewall of the foundation pit, for each horizontal position, the distance from the outer boundary to the inner boundary of the effective curtain area is defined as the curtain thickness of that area. ; Step E3: Calculate the thickness statistics Calculate the average curtain thickness at all locations using the following formula. and standard deviation : ， ； Where N represents the total number of curtain thickness sampling points; This represents the curtain thickness at the i-th data collection point; This represents the average thickness of the curtain. The standard deviation of the curtain thickness; Step E4: Calculate the effective area coverage rate Calculate the effective curtain area using the following formula. With the design target curtain area The ratio: ； Step E5: Calculate the curtain underdevelopment index Calculate the curtain underdevelopment index using the following formula. : ； in, It represents the coefficient of variation in thickness, reflecting the uniformity of the curtain thickness; , This indicates the set weighting coefficient.

8. The foundation pit curtain formation and groundwater regulation system based on digital twin and risk feedback control according to claim 7, characterized in that, The process of calculating the seepage risk coefficient in the foundation pit using the risk assessment module includes: Step F1: Extract the predicted total head distribution Extract the total head h at each spatial location in the predicted time domain from the future state predicted by the multiphysics coupled digital twin model; Step F2: Calculate the hydraulic gradient Calculate the hydraulic gradient at each location using the following formula. : ； in, , , This represents the rate of change of total head in the three directions of x, y, and z. Step F3: Identify the out-of-limit area Setting a safety threshold for hydraulic gradients The statistical hydraulic gradient satisfies regional volume ; Step F4: Determine the volume of the critical influence zone Determine the total volume of the critical influence zone of the foundation pit. Specifically, it is defined as the volume of soil within a depth of one foundation pit below the bottom of the foundation pit. Step F5: Determine the maximum hydraulic gradient Determine the maximum hydraulic gradient within the monitoring area. ; Step F6: Calculate the seepage risk factor Calculate the seepage risk factor using the following formula. : 。 9. A method for foundation pit curtain formation and groundwater regulation based on digital twin and risk feedback control, characterized in that, The procedure described in any one of claims 1 to 7 is as follows: Multi-dimensional physical field data of the construction site are collected in real time through a multi-source sensing layer; By leveraging digital twins and intelligent decision-making layers, multidimensional physical field data is assimilated to update the digital twin model. Based on the state of the digital twin model, risk assessment and rolling optimization of model prediction are performed to generate a set of control instructions. Through the closed-loop control layer, the control command set is sent to the physical execution layer to coordinate the adjustment of grouting operation parameters and groundwater regulation strategies, thus forming a closed-loop control.

Images