Method for predicting deformation of supporting structure of deep foundation pit with silt of million square meters

Abstract

The disclosed ten-thousand-square-meter muddy deep foundation pit supporting structure deformation prediction method comprises the following steps: establishing a soil-supporting interaction rheological model, combining a Burgers rheological model to simulate the stress relaxation and creep effect of the muddy soil body, and realizing dynamic evolution prediction of the supporting structure deformation field and internal force field. The particle filtering algorithm is used to carry out denoising processing on the real-time monitoring data, the model parameters are updated through the Bayesian method, and the prediction accuracy is improved. At the same time, the random forest algorithm is used to establish a nonlinear mapping of the deformation rate and the internal force increment, and a failure risk probability distribution is generated. The present application sets multiple collapse early warning thresholds, uses a weighted average method to fuse the monitoring data and the prediction results, and identifies abnormal points in the high-risk area. For the high-risk area, the genetic algorithm is used to optimize the support system stiffness and prestress distribution, the construction parameters are dynamically adjusted, the deformation and internal force of the supporting structure are effectively controlled, and the safety of the deep foundation pit construction is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of deep foundation pit support, in particular to a method for predicting the deformation of a ten-thousand-square-meter deep foundation pit support structure in silt. BACKGROUND

[0002] In the field of deep foundation pit support construction in coastal beach rheological silt, predicting the deformation and stress state of the support structure is crucial to ensuring engineering safety.

[0003] Rheological silt has the characteristics of high compressibility, strong rheology and low strength, which leads to the deformation instability and even collapse of the foundation pit support structure during construction, seriously threatening the safety of the project and the stability of the surrounding environment.

[0004] Therefore, accurately predicting the deformation and internal force distribution of the support structure and optimizing the construction scheme based on the prediction results have become the core issue in this field.

[0005] Existing support structure prediction and construction methods mainly rely on traditional finite element analysis and empirical formulas, but these methods have significant limitations in dealing with the complex mechanical behavior of rheological silt.

[0006] Traditional models often ignore the rheological properties of soil, making it difficult to accurately capture the dynamic changes in soil-support interaction during construction.

[0007] In addition, existing methods lack effective integration of real-time monitoring data, resulting in large deviations between predicted results and actual construction conditions, delayed construction scheme adjustments, and increased safety risks.

[0008] Specifically, the strong rheology and high sensitivity of rheological silt make the prediction of the deformation and stress state of the support structure face three major challenges: first, how to accurately simulate the dynamic evolution process of soil-support interaction, second, how to correct the prediction model through real-time monitoring data to improve accuracy, and third, how to establish an effective collapse warning mechanism based on prediction and monitoring results.

[0009] These challenges directly lead to the inability to identify potential instability risks of the support structure in the construction process, making it difficult to dynamically optimize construction parameters, and thus affecting the safety and efficiency of the project.

[0010] Therefore, in the construction of a ten-thousand-square-meter deep foundation pit support in rheological silt, how to establish a dynamic soil-support interaction rheological model, integrate real-time monitoring data to correct the prediction results, and implement precise navigation of the construction process through a multi-level warning mechanism has become a key issue to ensure the safety and efficiency of the project. SUMMARY

[0011] The present application aims to solve the above problems, provide a method for predicting the deformation of a deep foundation pit supporting structure with a size of ten thousand square meters, establish a soil-support interaction rheological model, integrate real-time monitoring data to correct the prediction results, and realize precise navigation during construction by combining a multi-level early warning mechanism.

[0012] The technical solution adopted by the present application to solve its technical problems is:

[0013] The method for predicting the deformation of a deep foundation pit supporting structure with a size of ten thousand square meters comprises the following steps:

[0014] S101 obtains real-time monitoring data of rheological soil from the construction site, combines preset rheological model parameters, constructs a soil-support interaction rheological model, and obtains an initial deformation field and an internal force field;

[0015] S102 updates the soil-support interaction rheological model through real-time monitoring data to generate an optimized deformation field and internal force field;

[0016] S103 generates a warning signal if the optimized deformation field or internal force field exceeds a preset threshold;

[0017] S104 adjusts the soil-support interaction rheological model using an optimization algorithm according to the warning signal to obtain an optimized set of construction parameters;

[0018] S105 updates the soil-support interaction rheological model based on the optimized set of construction parameters to predict the deformation field and internal force field of the next construction stage.

[0019] Further, step S101 comprises:

[0020] Obtain stress, displacement, and pore water pressure data of rheological soil from construction site sensors;

[0021] Combine laboratory-determined Burgers model parameters, including viscoelastic modulus and viscosity coefficient, to construct an initial soil-support interaction rheological model through finite element analysis to calculate the initial deformation field and internal force field of the supporting structure.

[0022] Further, step S102 comprises:

[0023] Use the time stepping method to determine the time step according to the daily progress of the construction stage, simulate the stress relaxation and creep effect caused by soil rheology, and update the soil stiffness matrix and damping matrix to calculate the dynamically evolving deformation field and internal force field through finite element analysis;

[0024]

[0025] Fuse real-time monitoring data and prediction results through a Kalman filtering algorithm to obtain an optimized deformation field and internal force field.​

[0026] Further, step S103 comprises:

[0027] Extracting displacement sequence and internal force sequence of the supporting structure from real-time monitoring data, adopting particle filtering algorithm to carry out denoising processing, and obtaining denoised displacement sequence and internal force sequence;

[0028] Calculating the deviation of the denoised sequence from the predicted deformation field and internal force field, and if the deviation exceeds a preset threshold, a warning signal is generated;

[0029] Updating the rheological model parameters based on the normal prior distribution through the Bayesian method to generate a corrected soil-supporting interaction rheological model.

[0030] Further, step S104 comprises:

[0031] Obtaining abnormal deformation points and internal force points of the high-risk area from the warning signal, and analyzing the soil-supporting interaction characteristics;

[0032] Adopting genetic algorithm to minimize deformation and internal force as objective function, and construction cost and time as constraint, to optimize support system stiffness and prestress distribution, and generate an optimized construction parameter set.

[0033] Further, step S105 comprises:

[0034] Updating the soil-supporting interaction rheological model from the optimized construction parameter set, and adopting finite element method to simulate the dynamic evolution of the construction process to obtain an optimized deformation field and internal force field;

[0035] Extracting deformation rate and internal force change trend from the optimized deformation field and internal force field, and if it exceeds a preset threshold, correcting the model through real-time monitoring data to generate a dynamically adjusted construction control parameter set.

[0036] The beneficial effects of the present application are:

[0037] 1、The present application establishes a soil-supporting interaction rheological model, simulates the stress relaxation and creep effect of silt soil body by combining Burgers rheological model, realizes the dynamic evolution prediction of the deformation field and internal force field of the supporting structure. The present application adopts particle filtering algorithm to denoise the real-time monitoring data, updates the model parameters through Bayesian method, and improves the prediction accuracy. At the same time, the present application uses random forest algorithm to establish the nonlinear mapping of deformation rate and internal force increment, and generates instability risk probability distribution. The present application sets multiple collapse warning threshold values, adopts weighted average method to fuse monitoring data and prediction results, and identifies abnormal points in high-risk areas. For the high-risk area, the present application applies genetic algorithm to optimize the support system stiffness and prestress distribution, and realizes effective control of the deformation and internal force of the supporting structure by dynamically adjusting the construction parameters, ensuring the safety of deep foundation pit construction. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 Flow chart for the structure of the present application;

[0039] Figure 2 Flow chart for calculating the preliminary deformation field U0 and the internal force field F0 of the present application;

[0040] Figure 3 Schematic diagram for generating the soil-support interaction rheological model M1 of the present application;

[0041] Figure 4 Schematic diagram of the risk early warning system of the present application;

[0042] Figure 5 Flow chart for generating the construction parameter set of the present application. DETAILED DESCRIPTION

[0043] As Figure 1 shown, the ten thousand square meter level of silt deep foundation pit supporting structure deformation prediction method, including the following steps:

[0044] S101 obtains real-time monitoring data of rheological soil from the construction site, combines preset rheological model parameters, constructs a soil-support interaction rheological model, and obtains an initial deformation field and an internal force field;

[0045] S102 updates the soil-support interaction rheological model through real-time monitoring data, generates an optimized deformation field and an internal force field;

[0046] S103 generates a warning signal if the optimized deformation field or internal force field exceeds a preset threshold;

[0047] S104 adjusts the soil-support interaction rheological model using an optimization algorithm according to the warning signal to obtain an optimized construction parameter set;

[0048] S105 updates the soil-support interaction rheological model based on the optimized construction parameter set to predict the deformation field and internal force field of the next construction stage.

[0049] As Figure 2 shown, step S101 includes collecting stress, displacement and pore water pressure real-time data of rheological silt soil from construction site sensors, combining Burgers rheological model parameters determined by laboratory rheological tests, including viscoelastic modulus E1, E2 and viscosity coefficient η1, η2, establishing an initial soil-support interaction rheological model, and calculating the preliminary deformation field U0 and internal force field F0 of the supporting structure.

[0050] The real-time data of stress, displacement and pore water pressure of the rheological silt is acquired from the construction site sensor, and combined with the Burgers model parameters determined in the laboratory, including viscoelastic modulus E1, E2 and viscosity coefficient η1, η2, an initial soil-supporting interaction rheological model is constructed to obtain a preliminary deformation field U0 and an internal force field F0. The soil-supporting interaction rheological model is corrected by using a finite element analysis method, according to the preliminary deformation field U0 and the internal force field F0, combined with the real-time data, to obtain an updated deformation field U1 and an internal force field F1. Through the Kalman filtering algorithm, the real-time data and the updated deformation field U1 and internal force field F1 are fused to obtain an optimized deformation field U2 and internal force field F2. If the displacement value of the optimized deformation field U2 exceeds a preset threshold T1, or the stress value of the internal force field F2 exceeds a preset threshold T2, a first-level warning signal S1 is generated. According to the first-level warning signal S1, a Bayesian network algorithm is used to analyze the spatio-temporal variation trend of the deformation field U2 and the internal force field F2, to obtain a collapse risk assessment result R1. According to the collapse risk assessment result R1, combined with the real-time data, a numerical optimization method is used to adjust the boundary conditions of the finite element model to obtain an optimized construction scheme P1.

[0051] The pore water pressure sensor adopts a pressure type pore water pressure gauge: based on the principle of hydrostatic pressure, the pore water pressure is measured by using a water pressure sensitive integrated component. When the sensor is fixed at a certain point under water, the water column pressure above the measuring point acts on the water pressure sensitive integrated component, causing the resistance of the component to change, thereby causing the voltage to change, and then indirectly measuring the pore water pressure at the point.

[0052] The construction site sensor collects the stress, displacement and pore water pressure data of the rheological silt in real time, and provides dynamic input for the soil-supporting interaction model. The rheological silt needs to be described by the Burgers model parameters determined in the laboratory due to its viscoelastic characteristics. The Burgers model combines elastic modulus E1, E2 and viscosity coefficient η1, η2 to represent the deformation behavior of the silt under short-term and long-term loading.

[0053] The constitutive relation of the Burgers model can be expressed as:

[0054] Stress relationship:

[0055] σ(t) = E1ε1(t) + E2ε2(t)

[0056] Wherein, ε1(t) is the strain through the Maxwell unit, and ε2(t) is the strain through the Kelvin unit.

[0057] Strain relationship:

[0058] ε(t) = ε1(t) + ε2(t)

[0059] Constitutive relation of Maxwell unit:

[0060] σ1(t) = E1ε1(t)

[0061]

[0062] Constitutive relation of Kelvin unit:

[0063]

[0064] Physical meaning of parameters: E1 (elastic modulus 1): represents the stiffness of the elastic element in the Maxwell unit. The larger E1 is, the smaller the elastic deformation of the material under short-term loading is. η1 (viscosity coefficient 1): represents the viscosity of the viscous element in the Maxwell unit. The larger η1 is, the smaller the creep rate of the material is. E2 (elastic modulus 2): represents the stiffness of the elastic element in the Kelvin unit. The larger E2 is, the stronger the elastic recovery ability of the material under long-term loading is. η2 (viscosity coefficient 2): represents the viscosity of the viscous element in the Kelvin unit. The larger η2 is, the smaller the relaxation rate of the material is.

[0065] Based on S1, the spatio-temporal variation of U2 and F2 is analyzed using Bayesian network. Combined with historical data, the Bayesian network predicts that the stress of the side wall of the foundation pit will increase to 200 kPa within 24 hours, and the collapse probability R1 will reach 30%. This result guides the construction adjustment, such as reducing the excavation rate and reducing the disturbance of the soil body.

[0066] The numerical optimization method adjusts the boundary conditions of the finite element model to generate the optimized construction scheme P1.

[0067] The above method ensures dynamic updating of the model and accurate risk assessment through multi-link cooperation. Each technical link supports each other, reduces construction risk, and improves engineering efficiency and safety.

[0068] Step S102 includes: obtaining initial deformation field U0 and internal force field F0, establishing soil-support interaction rheological model through finite element analysis, and obtaining support book using time stepping method, determining time step Δt according to daily progress of construction stage, simulating stress relaxation and creep effect caused by soil rheology, updating stiffness matrix K and damping matrix C, and obtaining dynamically evolving deformation field U1 and internal force field F1.

[0069]

[0070] Δt represents the time step, L represents the length of the construction section, and v represents the daily progress rate of construction.

[0071] K(t) = K0e -αt

[0072] K(t) represents a stiffness matrix that varies with time, K0 represents an initial stiffness matrix, a represents a relaxation coefficient, and t represents a time variable.

[0073]

[0074] C(t) represents a damping matrix, and cij(t) represents a damping coefficient of the i-th degree of freedom on the j-th degree of freedom, which can vary with time t.

[0075]

[0076] U1 represents an updated deformation field, U0 represents an initial deformation field, v(t) represents a deformation rate, and t represents a time variable.

[0077]

[0078] F1 represents an updated internal force field, K(t) represents a current stiffness matrix, C(t) represents a current damping matrix, U1 represents a current displacement, and U1dot represents a current velocity.

[0079] The stiffness matrix K reflects the decay of the elastic modulus of the soil over time, and the damping matrix C represents the viscous dissipation.

[0080] The deformation and internal force data of the supporting structure are obtained through real-time monitoring. If the deviation between the monitoring data and the prediction result exceeds a preset threshold, the soil-supporting interaction rheological model parameters are corrected by the least square method to obtain optimized model parameters. According to the optimized model parameters, the deformation field U2 and the internal force field F2 of the supporting structure are recalculated, and the stress state of the supporting structure in the current construction stage is determined through finite element analysis. If the deformation field U2 or the internal force field F2 exceeds a preset multi-level warning threshold, the time step At is adjusted, the rheological effect of the soil is simulated again, and new deformation field U3 and internal force field F3 are obtained. Through the iterative comparison of the real-time monitoring data and the prediction result, the stability of the stress state of the supporting structure is judged, and the supporting design optimization scheme of the current construction stage is obtained. According to the supporting design optimization scheme, the soil-supporting interaction rheological model is updated, the deformation field U4 and the internal force field F4 of the next construction stage are calculated, and the dynamic evolution trend of the supporting structure is determined.

[0081] When obtaining the initial deformation field U0 and the internal force field F0, the stress and displacement data of the silt soil can be collected by the field sensor. The sensor is arranged at the interface between the supporting pile and the soil, records the earth pressure and pile displacement in the initial construction stage, obtains the initial settlement distribution of the soil represented by U0 and the internal force distribution of the pile represented by F0. The accuracy of the data depends on the sensor calibration and the laying density, and it is preferred to set a measuring point every meter to ensure that the data covers the entire supporting area.

[0082] When simulating the rheological effect of soil using the time-stepping method, the time step Δt can be determined according to the construction progress. For example, if a certain project excavates 0.5 meters per day, Δt is set to 24 hours, and the stress relaxation caused by soil creep is simulated by combining the Burgers model parameters. After updating the stiffness matrix K and the damping matrix C, the calculation obtains that the settlement of the support pile top represented by U1 increases by 2 millimeters, and the internal force of the pile represented by F1 increases by 10%.

[0083] The deviation correction of real-time monitoring data and prediction results can be achieved by the least squares method. For example, monitoring finds that the actual settlement of the support pile is 3 millimeters, while U1 predicts 2 millimeters, with a deviation exceeding the threshold of 5%. By adjusting the viscoelastic modulus E1 and the viscosity coefficient η1 in the model through the least squares method, the prediction result is closer to the actual situation. After adjustment, recalculation is performed to obtain U2, which shows that the settlement is 2.8 millimeters, and F2 shows that the internal force distribution is more uniform.

[0084] The setting of multi-level warning thresholds can be based on the design specifications of the supporting structure. The first-level threshold is set to 5 millimeters of settlement, and the second-level threshold is set to 8 millimeters. When U2 shows a settlement of 6 millimeters, the first-level warning is triggered, and Δt is adjusted to 12 hours. Re-simulation obtains U3, and the settlement is reduced to 4.5 millimeters. F3 shows that the internal force does not exceed the threshold, indicating that the structure is stable.

[0085] Shortening the time step can improve the simulation accuracy and timely detect potential risks. For example, when comparing the monitoring data and the prediction results iteratively, the stability can be verified through multiple simulations.

[0086] When updating the rheological model of soil-support interaction, the boundary conditions can be adjusted according to the optimization scheme. For example, after increasing the support, the boundary stiffness of the model is increased by 10%, and U4 and F4 are recalculated to predict that the long-term settlement of the support pile does not exceed the design limit. The dynamic evolution trend shows that the structure remains stable during subsequent construction, and the optimization scheme effectively improves the construction safety.

[0087] As shown in Figure 3 Step S103 includes: extracting the displacement sequence D and the internal force sequence S of the supporting structure from the real-time monitoring data, denoising the sequences D and S using the particle filtering algorithm to obtain the denoised displacement sequence D1 and the denoised internal force sequence S1, calculating the deviations ΔD and ΔF between the denoised sequences D1 and S1 and the dynamic evolution deformation field U1 and the internal force field F1, and if the deviation ΔD or ΔF exceeds the preset threshold of 5%, updating the Burgers model parameters E1, E2, η1, and η2 based on the normal prior distribution through the Bayesian method to generate the corrected soil-support interaction rheological model M1.

[0088] The displacement sequence D and the internal force sequence S of the supporting structure are obtained from a real-time monitoring system and stored as time series data to obtain an initial data set. The particle filter algorithm is used to denoise the displacement sequence D and the internal force sequence S to generate a denoised displacement sequence D1 and a denoised internal force sequence S1, and obtain a denoised data set. Through a finite element analysis method, the dynamic evolution deformation field U1 and the internal force field F1 are calculated based on the boundary conditions of the current construction stage to obtain predicted field data.

[0089]

[0090] U1 represents a deformation field vector, n represents the number of finite element nodes, α_i represents the weight coefficient of the i-th node, K_i represents the stiffness matrix, F_i represents the node force vector, β represents the damping coefficient, represents the Laplacian of displacement.

[0091]

[0092] F1 represents an internal force field vector, Ω represents a calculation domain, B represents a strain displacement matrix, D represents an elastic matrix, γ represents a time correlation coefficient, represents the partial derivative of displacement with respect to time.

[0093] The deviations ΔD of the denoised displacement sequence D1 and the deformation field U1, and the deviations ΔF of the denoised internal force sequence S1 and the internal force field F1 are calculated. If the deviations ΔD or ΔF exceed a preset threshold, it is marked as an abnormal state to obtain deviation state data. For the abnormal state data, the Burgers model parameters including the elastic modulus E1, the elastic modulus E2, the viscous coefficient η1, and the viscous coefficient η2 are updated based on the normal prior distribution through the Bayesian method to generate a corrected soil-supporting interaction rheological model M1, and the updated model parameters are obtained. According to the corrected soil-supporting interaction rheological model M1, in combination with the real-time monitoring data, the deformation field U2 and the internal force field F2 of the next construction stage are predicted to obtain predicted distribution data. If the extreme values of the deformation field U2 or the internal force field F2 in the predicted distribution data exceed a preset multi-level threshold, a warning signal is generated to obtain warning state data.

[0094] The displacement sequence D and the internal force sequence S of the supporting structure obtained from the real-time monitoring system are the basis for analyzing the soil-supporting interaction. In deep foundation pit construction, monitoring equipment such as inclinometers and stress meters record the displacement and internal force of the supporting pile every hour to form time series data. The initial data set may contain noise such as equipment error or environmental vibration interference. To ensure data reliability, the particle filter algorithm is used to denoise the displacement sequence D and the internal force sequence S.

[0095] Particle filtering simulates state distribution by a large number of particles, eliminates abnormal fluctuations, and generates denoised displacement sequence D1 and internal force sequence S1. Assuming that the displacement sequence D of a certain foundation pit monitoring point contains an abnormal peak value of 100 mm, the denoised D1 shows that it is stable within 10 mm, reflecting the true deformation trend. The denoised data set provides high-quality input for subsequent analysis. Through the finite element analysis method, the deformation field U1 and the internal force field F1 are calculated based on the construction stage boundary conditions.

[0096] The construction boundary conditions include the self-weight of the soil, the groundwater level, and the construction load. The finite element model divides the support structure and soil grid to predict U1 as the support pile top displacement of 5 mm and F1 as the internal force of 500 kN. The predicted field data is compared with the denoised data to calculate the deviation ΔD and ΔF.

[0097] If D1 is 4 mm and U1 is 5 mm, ΔD is 1 mm; if S1 is 480 kN and F1 is 500 kN, ΔF is 20 kN. If the preset threshold values are 2 mm and 30 kN respectively, ΔD is normal but ΔF is abnormal, marked as an abnormal state. The deviation state data guides the model correction. For abnormal state data, the Bayesian method updates the Burgers model parameters.

[0098] The Burgers model describes the rheological properties of the soil, including the elastic modulus E1, E2 and the viscosity coefficient η1, η2.

[0099] The initial E1 is 50 MPa and η1 is 1000 MPa·s. In the abnormal state, E1 is adjusted to 48 MPa and η1 is adjusted to 950 MPa·s through the normal prior distribution. The corrected model M1 more accurately reflects the stiffness and damping characteristics of the soil. Updating the model parameters improves the prediction accuracy. According to the model M1 combined with the monitoring data, U2 and F2 of the next stage are predicted.

[0100] For example, U2 predicts that the support pile top displacement increases to 6 mm and F2 is 520 kN. The predicted distribution data evaluates whether to trigger an early warning. If the multi-level threshold is set as displacement 7 mm or internal force 550 kN, the current U2 and F2 are not over limit, and no warning is needed. If F2 increases to 560 kN in the future, an early warning signal is generated. The early warning state data supports timely adjustment of the construction plan.

[0101] The above method forms a closed loop through monitoring, denoising, modeling, deviation analysis, parameter correction, and prediction. Each link supports each other, ensuring the accuracy of the support structure state evaluation and the timeliness of dynamic adjustment. Iterative optimization of real-time monitoring and prediction provides data support for construction safety.

[0102] For example, Figure 4As shown, step S103 includes: calculating the support structure deformation rate V and internal force increment ΔF1 of each construction stage from the revised soil-support interaction rheological model M1, training the historical monitoring data D, S and the prediction results V, ΔF1 using the random forest algorithm, establishing a nonlinear mapping function f(V, ΔF1) of the deformation rate V and the internal force increment ΔF1, and generating a support structure instability risk probability distribution P.

[0103] The deformation rate V and the internal force increment ΔF1 of each construction stage are obtained from the soil-support interaction rheological model M1. The deformation and internal force distribution of the support structure are calculated using the finite element analysis method to obtain the initial prediction results. The initial prediction results are revised according to the real-time monitoring data, and the soil-support interaction rheological model M1 is updated using the information-based construction method to obtain the revised deformation rate V and the internal force increment ΔF1.

[0104]

[0105] V(t) represents the deformation rate of the support structure at time t, u(t) represents the displacement function, α is the correction coefficient, d_i represents the displacement increment of the i th monitoring point, and t_i represents the corresponding time interval.

[0106]

[0107] ΔF1 represents the internal force increment of the support structure, Ks represents the structure stiffness matrix, Δε represents the strain increment, β is the stress correction coefficient, Pj represents the load value of the j th measuring point, and Aj represents the corresponding action area.

[0108] The historical monitoring data D, S and the revised deformation rate V, internal force increment ΔF1 are trained using the random forest algorithm to establish a nonlinear mapping function f(V, ΔF1), where V represents the deformation rate and ΔF1 represents the internal force increment, and the mapping relationship is obtained.

[0109]

[0110] f(V, ΔF1) represents the nonlinear mapping function, w_k represents the weight of the k th decision tree in the random forest, h_k represents the output of the k th decision tree, and T represents the total number of decision trees. The instability risk probability distribution P of the support structure is calculated according to the nonlinear mapping function f(V, ΔF1) to obtain the risk probability value.

[0111]

[0112] P represents the instability risk probability, M represents the sampling number, f_j represents the j th prediction result, θ represents the instability threshold, and I represents the indicator function.

[0113] If the risk probability value exceeds the preset threshold T1, the information-based construction method is used to adjust the support design parameters, update the soil-support interaction rheological model M1, recalculate the deformation rate V and the internal force increment ΔF1, and obtain the optimized prediction results. According to the optimized prediction results and the multi-level warning thresholds, the risk grade of the support structure instability is judged, and the corresponding warning signal is generated. The accuracy of the warning signal is verified according to the real-time monitoring data, the instability risk probability distribution P is updated using the finite element analysis method, and the final risk assessment results are obtained.

[0114] When obtaining the deformation rate V and the internal force increment ΔF1 of the support structure in the soil-support interaction rheological model M1, the displacement and internal force data of the support structure can be collected through the real-time monitoring system. Assuming that the support structure in a certain foundation pit project is a steel sheet pile, the displacement data of the monitoring point is recorded once every hour, and the displacement sequence D is obtained. Through time difference calculation, the deformation rate V is obtained, for example, the displacement change of a certain point in 24 hours is 12 millimeters, and V is 0.5 millimeters / hour. The internal force increment ΔF1 is measured by strain gauges to measure the stress change of the steel sheet pile, which is converted into internal force increment, for example, the internal force of a certain section of steel sheet pile increases from 1000 kN to 1200 kN, and ΔF1 is 200 kN. Such data collection method ensures the accuracy of model input.

[0115] When calculating the deformation and internal force distribution of the support structure using the finite element analysis method, a two-dimensional or three-dimensional numerical model can be established based on the M1 model. Assuming that the depth of the foundation pit is 10 meters and the soil is clay, the elastic modulus of the steel sheet pile and the shear modulus of the soil are set in the model. The finite element analysis generates the initial prediction results, for example, it predicts that the maximum deformation of a certain point is 15 millimeters and the internal force is 1300 kN.

[0116] When correcting the initial prediction results according to the real-time monitoring data, the information-based construction method adjusts the model parameters by comparing the monitoring data with the prediction results.

[0117] The monitoring data shows that the deformation of a certain point is 18 millimeters, which exceeds the predicted value of 15 millimeters, so the stiffness parameters of the soil are updated through an iterative optimization algorithm to generate the corrected V of 0.6 millimeters / hour and ΔF1 of 220 kN. Such correction improves the adaptability of the model.

[0118] When training the historical monitoring data D, S and the corrected V, ΔF1 using the random forest algorithm, a model containing 100 decision trees can be constructed. Assuming that the historical data contains 1000 samples, each including displacement, internal force, deformation rate and internal force increment. After training, the nonlinear mapping function f(V, ΔF1) is obtained, for example, input V is 0.5 millimeters / hour, ΔF1 is 200 kN, and output instability risk probability P is 0.3. This algorithm can capture complex data relationships and improve the accuracy of risk prediction.

[0119] In one embodiment, when calculating the instability risk probability distribution P, if P exceeds the threshold T1, for example T1 is set to 0.4, and a certain calculation result P is 0.45, the design parameter adjustment is triggered. The information construction method may increase the embedded depth of the steel sheet pile from 8 meters to 10 meters, and after updating the M1 model, the V is calculated to be 0.4 mm / h and the P is calculated to be 0.35. This adjustment reduces the risk of instability.

[0120] When judging the risk level of the supporting structure, based on the optimized V and ΔF1, multiple warning thresholds are set, for example, V exceeds 0.7 mm / h or ΔF1 exceeds 300 kN as high risk. Assuming that the V of a certain point is 0.8 mm / h, a high-risk warning signal is generated. Such a graded warning facilitates construction management.

[0121] When verifying the accuracy of the warning signal, real-time monitoring data shows that the V of a certain point is 0.75 mm / h, which is consistent with the warning. After updating P by finite element analysis, it is confirmed that the risk level is high risk, and the final evaluation result guides the reinforcement measures, such as increasing the support beam. This verification ensures the reliability of risk assessment.

[0122] As shown in Figure 4 , step S103 includes: obtaining the risk probability value from the instability risk probability distribution P, setting three levels of collapse warning thresholds T1=80%, T2=60%, T3=40%, if the probability value P exceeds the first level threshold T1, then based on the monitoring data variance σD, σS and the prediction model confidence α, the abnormal deformation point U2 and the abnormal internal force point F2 of the high-risk area and the warning signal A are generated by fusing the denoised displacement sequence D1, the denoised internal force sequence S1 and the predicted deformation field U1, the internal force field F1 through the weighted average method.

[0123] A risk probability value is obtained from a risk probability distribution P of instability, the risk probability value P representing a risk of instability of the supporting structure; the risk probability value P is compared with preset three-level early warning thresholds T1, T2 and T3 to determine whether T1 is exceeded, a trigger early warning condition is obtained, and the T1, T2 and T3 represent threshold values of different risk levels; if the risk probability value P exceeds T1, monitoring data variances σD and σS and a prediction model confidence level α are obtained, the σD represents a variance of displacement monitoring data, the σS represents a variance of internal force monitoring data, and the α represents a reliability of the prediction model; a weighted average method is used to fuse a denoised displacement sequence D1, a denoised internal force sequence S1, a predicted deformation field U1 and an internal force field F1 to obtain fused data sequences D2, S2, U2 and F2, the D1 and S1 represent denoised displacement and internal force sequences, and the U1 and F1 represent predicted deformation and internal force distributions; through a soil-supporting interaction rheological model, combined with finite element analysis, based on the fused data sequences D2, S2, U2 and F2, deformation distribution and internal force distribution of the supporting structure are calculated to obtain predicted deformation points U3 and internal force points F3; real-time monitoring data is used to perform deviation analysis on the predicted deformation points U3 and the internal force points F3, deviation values ΔU and ΔF are calculated, it is determined whether the deviation values ΔU and ΔF exceed preset threshold values, corrected deformation points U4 and internal force points F4 are obtained, the ΔU and ΔF represent deviations of predicted and actual deformations and internal forces; based on the corrected deformation points U4 and the internal force points F4, abnormal deformation points U5 and abnormal internal force points F5 are extracted, an early warning signal A1 is generated, the U5 and F5 represent deformation and internal force points exceeding a normal range; based on the early warning signal A1 and real-time monitoring data, supporting design parameters are optimized, and an adjusted supporting scheme B1 is generated; for the adjusted supporting scheme B1, the soil-supporting interaction rheological model is recalculated, predicted deformation field U6 and internal force field F6 are updated, it is determined whether a safety threshold value is met, and a final early warning signal A2 is generated, the U6 and F6 represent updated deformation and internal force distributions.

[0124] A risk probability value is obtained from a risk probability distribution P of instability, assuming that in a certain foundation pit supporting project, the P value is 0.75, indicating that the possibility of instability of the supporting structure is relatively high. In principle, the P value is obtained based on historical monitoring data and a prediction model through statistical analysis, reflecting the potential risk of the structure. The P value is compared with three-level early warning thresholds T1=0.6, T2=0.8 and T3=0.9, 0.75 exceeds T1 but is lower than T2, triggering a first-level early warning, indicating that further analysis is required but immediate shutdown is not required.

[0125] In a possible implementation, if the P value exceeds T1, monitoring data variances σD and σS and a prediction model confidence level α are obtained.

[0126] For example, the displacement monitoring data variance σD=0.02 mm 2, the internal force monitoring data variance σS=1.5kN 2 , the confidence level α=0.95, indicating that the data fluctuation is small and the model is reliable. The variance reflects the dispersion degree of the monitoring data, and the confidence level measures the stability of the prediction model, which helps to judge whether the data is reliable.

[0127] Specifically, the weighted average method is used to fuse the denoised displacement sequence D1, the denoised internal force sequence S1, the predicted deformation field U1 and the internal force field F1 to generate the fused data sequence D2, S2, U2 and F2.

[0128] For example, D1 contains the displacement sequence at the top of the foundation pit, S1 is the internal force sequence of the anchor rod, and the weights are 0.4, 0.3, 0.2 and 0.1 respectively. After fusion, D2 and S2 are smoother, reducing noise interference and improving data consistency.

[0129] Through the soil-support interaction rheological model combined with finite element analysis, the deformation and internal force distribution are calculated based on D2, S2, U2 and F2 to obtain the predicted deformation points U3 and internal force points F3.

[0130] For example, U3 shows that the maximum deformation in the middle of the foundation pit is 15mm, and F3 shows that the internal force of a certain anchor rod is 120kN. These points reflect the stress state of the supporting structure and provide a basis for subsequent correction.

[0131] For example, based on real-time monitoring data, deviation analysis is performed on U3 and F3 to calculate the deviation values ΔU and ΔF. Assuming that the actual monitoring deformation is 16mm and the internal force is 125kN, then ΔU=1mm and ΔF=5kN. If the preset threshold is ΔU≤2mm and ΔF≤10kN, the deviation is within the acceptable range, and after correction, U4=15.5mm and F4=122kN are obtained, indicating that the prediction result is relatively accurate.

[0132] Based on U4 and F4, abnormal points U5 and F5 are extracted, assuming that U5 is a point with local deformation exceeding 20mm and F5 is a point with internal force exceeding 150kN, generating a first warning signal A1, prompting that the local area needs attention. A1 is triggered based on abnormal points, which is beneficial to timely discover potential risks.

[0133] Based on A1 and real-time monitoring data, the supporting design parameters are optimized to generate an adjustment scheme B1.

[0134] For example, increasing the anchor rod density or adjusting the prestress generates B1 scheme. This optimization reduces the risk by adjusting the parameters and improves the stability of the structure.

[0135] The soil-support interaction rheological model is recalculated for B1 to update the predicted deformation field U6 and the internal force field F6.

[0136] For example, U6 shows that the maximum deformation is reduced to 12mm, and the internal force of F6 is reduced to 110kN, which meets the safety threshold, and generates the final warning signal A2 as "safe". This iterative update ensures that the prediction result is consistent with the actual working condition, and guarantees construction safety.

[0137] As shown in Figure 5 Step S104 includes: obtaining abnormal deformation points U2 and abnormal internal force points F2 of the high-risk area from the warning signal A, analyzing the soil-support interaction characteristics of the corresponding area, using genetic algorithm to minimize deformation U2 and internal force F2 as the objective function, and construction cost C and time T as the constraint, optimizing support system stiffness K1 and prestress distribution P1, and generating optimized construction parameter set Q.

[0138] Abnormal deformation points U2 and abnormal internal force points F2 are obtained from the warning signal A, and the deformation distribution V1 and the internal force distribution W1 of the supporting structure are calculated by the soil-support interaction rheological model and finite element analysis, to obtain the predicted deformation field Y1 and the internal force field Z1. Based on the predicted deformation field Y1 and the internal force field Z1, combined with the real-time monitoring data M1, the deformation deviation ΔY and the internal force deviation ΔZ are calculated. If the deviation ΔY or ΔZ exceeds the preset threshold T4, the real-time monitoring data M1 and the predicted deformation field Y1 and the internal force field Z1 are fused by using the weighted average method to generate the corrected deformation field Y2 and the internal force field Z2. Genetic algorithm is used to minimize the corrected deformation field Y2 and the internal force field Z2 as the objective function, and the construction cost C1 and the construction time T1 as the constraint condition, to optimize the support stiffness K2 and the prestress distribution P2 of the supporting structure, to obtain the optimized parameter set Q1. Based on the optimized parameter set Q1, the deformation distribution V2 and the internal force distribution W2 of the supporting structure are calculated by the soil-support interaction rheological model and finite element analysis, to obtain the updated deformation field Y3 and the internal force field Z3. The monitoring deformation points N1 and the monitoring internal force points N2 are obtained from the real-time monitoring data M1, the deviation ΔY1 between the updated deformation field Y3 and the monitoring deformation points N1 is calculated, and the deviation ΔZ1 between the internal force field Z3 and the monitoring internal force points N2 is calculated. If the deviation ΔY1 or ΔZ1 exceeds the preset threshold T5, the updated deformation field Y3, the internal force field Z3, and the monitoring deformation points N1, the monitoring internal force points N2 are fused by using the weighted average method to generate the corrected deformation field Y4 and the internal force field Z4. Based on the corrected deformation field Y4 and the internal force field Z4, the abnormal deformation points U3 and the abnormal internal force points F3 are extracted. If the abnormal deformation points U3 or the abnormal internal force points F3 exceed the warning threshold T6, the warning signal A3 is generated. By using the information construction method, based on the warning signal A3 and the corrected deformation field Y4 and the internal force field Z4, the supporting design parameters are adjusted to generate the optimized supporting scheme B2.

[0139] Specifically, obtaining abnormal deformation points U2 and abnormal internal force points F2 from the warning signal A is the core step of risk assessment of the supporting structure.

[0140] The early warning signal A is based on real-time monitoring data and prediction models, identifying areas in the support structure where instability may occur, such as points of excessive local deformation in the tunnel lining. Abnormal deformation points U2 represent coordinates with displacements outside the normal range, such as a monitoring point with a displacement of 5 mm, exceeding the threshold of 3 mm. Abnormal internal force points F2 may be cross-sections with stresses of 200 kPa, exceeding the safety value of 150 kPa. These points are generated through a combination of sensor data and prediction models, ensuring that high-risk areas are accurately located.

[0141] The deformation distribution V1 and internal force distribution W1 of the support structure are calculated using the soil-support interaction rheological model combined with finite element analysis, resulting in the predicted deformation field Y1 and internal force field Z1.

[0142] Based on Y1 and Z1 combined with real-time monitoring data M1, the deviations ΔY and ΔZ are calculated. For example, M1 shows that the actual deformation of a certain point is 4.5 mm, while Y1 predicts 4 mm, resulting in a ΔY of 0.5 mm. If ΔY exceeds the threshold T4, such as 0.3 mm, a correction is triggered.

[0143] The weighted average method is used to fuse M1 with Y1 and Z1, generating the corrected deformation field Y2 and internal force field Z2. The weighted average method assigns weights based on data reliability, such as M1 with a weight of 0.6 and Y1 with a weight of 0.4, resulting in a deformation point adjustment of 4.3 mm in Y2. This fusion improves the accuracy of the prediction.

[0144] The genetic algorithm is used to optimize the support stiffness K2 and prestress distribution P2, with Y2 and Z2 as the objective functions, and construction cost C1 and time T1 as constraints.

[0145] The initial K2 is 5 GPa, which is increased to 6 GPa after optimization, and the prestress P2 is adjusted from 100 kPa to 120 kPa. The genetic algorithm iteratively searches for the global optimal solution, ensuring that the support scheme is economical and efficient. The optimized parameters Q1 are used for finite element analysis to generate updated deformation field Y3 and internal force field Z3, with further reductions in deviations ΔY1 and ΔZ1, such as ΔY1 decreasing to 0.2 mm.

[0146] If ΔY1 or ΔZ1 exceeds the threshold T5, such as 0.15 mm, Y3 and Z3 are again fused with monitoring points N1 and N2 to generate Y4 and Z4. After fusion, abnormal deformation points U3 and internal force points F3 are extracted, such as U3 with a deformation of 5.5 mm, exceeding the T6 threshold of 5 mm, triggering the early warning signal A3.

[0147] Information-based construction adjusts design parameters based on A3 and Y4, Z4, such as increasing the density of anchor rods, generating an optimized scheme B2. This method ensures the safety and stability of the support structure through multiple rounds of correction and optimization.

[0148] The above process is progressive and interlocking from abnormal point extraction to optimization scheme generation. Each step is based on real-time data and prediction models, balancing accuracy and efficiency, to provide reliable support for support design under complex geological conditions.

[0149] Step 105, updating the soil-support interaction rheological model M1 from the optimized construction parameter set Q, using the finite element method to re-simulate the dynamic evolution of the construction process, calculating the optimized support structure deformation field U3 and internal force field F3, verifying the control effect of the optimized parameter set Q on the deformation field U3 and internal force field F3.

[0150] Obtain the initial construction parameter set Q and the soil-support interaction rheological model M1. Update the model M1 parameters to generate the updated model M2. Use the finite element method to simulate the dynamic evolution of the model M2, obtaining the deformation field U3 and internal force field F3 of the support structure. Through real-time monitoring data, extract the actual deformation data D1 and internal force data F1 of the support structure, and generate the monitoring data set S1. If the deviation between the monitoring data set S1 and the deformation field U3 and internal force field F3 exceeds the preset threshold, adjust the parameters of the model M2 through the information construction method to generate the corrected model M3. According to the corrected model M3, use the finite element method to simulate the construction process to obtain the optimized deformation field U4 and internal force field F4. Through the deformation field U4 and internal force field F4, combined with the preset multi-level warning threshold, judge the stress state of the support structure, and generate the warning signal W1.

[0151] Specifically, obtaining the initial construction parameter set Q and the soil-support interaction rheological model M1 is the basis for optimizing support design. The initial construction parameter set Q includes support stiffness, prestress distribution, and material parameters, etc.

[0152] For example, the support stiffness can be set to 1000 kN / m, the prestress is 500 kPa, and the concrete strength grade is C30. The soil-support interaction rheological model M1 describes the mechanical coupling relationship between the soil and the support structure, which is based on the elastic foundation beam theory, combined with the soil elastic modulus of 20 MPa and the Poisson's ratio of 0.3.

[0153] M1 is calibrated through field geological survey data to ensure that the initial model is close to the actual working condition. The updated model M1 generates M2 by incorporating real-time construction data.

[0154] The parameters of M1 are updated by monitoring the soil settlement and support internal force. For example, monitoring found that the soil settlement rate exceeded the expected value, reaching 5 mm / day, then adjusted the soil stiffness parameter in M1 to generate M2, the soil elastic modulus may be adjusted from 20 MPa to 18 MPa.

[0155] The updating process needs to ensure model convergence and avoid parameter adjustment leading to unstable calculation. Finite element method is adopted to simulate the dynamic evolution of M2, obtaining deformation field U3 and internal force field F3.

[0156] Finite element analysis can be based on commercial software such as ANSYS to simulate the stress evolution of the supporting structure during construction. Assuming that the simulation results show that the maximum deformation is 8mm and the maximum internal force is 600kN, the corresponding U3 and F3 fields are generated.

[0157] Dynamic evolution simulation considers a time step of 1 day, simulates the construction progress within 7 days, and captures the spatio-temporal distribution of deformation and internal force. By real-time monitoring data, actual deformation data D1 and internal force data F1 are extracted to generate monitoring dataset S1.

[0158] The monitoring equipment includes displacement meters and stress sensors, which record the deformation of a certain point of the supporting structure as 7.5mm and the internal force as 620kN, forming S1. The monitoring frequency can be set to 2 times per day to ensure the timeliness of the data.

[0159] The accuracy of S1 directly affects the reliability of subsequent deviation analysis. If the deviation of S1 from U3 and F3 exceeds the preset threshold, for example, the deformation deviation threshold is 2mm and the internal force deviation threshold is 50kN, then adjust the parameters of M2 to generate M3.

[0160] For example, deviation analysis shows that the deformation deviation is 2.5mm, triggering adjustment.

[0161] M3 is generated by fusing the parameters of S1 and M2 through weighted average method, and the adjusted soil stiffness may be 17.5MPa. M3 improves the fitting degree of the model to the actual working conditions. Based on M3, the construction process is simulated by finite element method, obtaining optimized U4 and F4.

[0162] The simulation shows that the maximum deformation of U4 is reduced to 6.5mm and the maximum internal force of F4 is 580kN, indicating that the stress of the supporting structure is more balanced after optimization.

[0163] The simulation process can refine the grid division to improve the calculation accuracy. Through the combination of U4 and F4 with multi-level warning thresholds, the stress state is judged to generate warning signal W1.

[0164] The warning thresholds are divided into three levels: deformation less than 5mm is safe, 5-7mm is attention, and more than 7mm is dangerous. Assuming that U4 shows that the deformation of a certain area is 6.8mm, triggering the W1 signal of attention level.

[0165] W1 can guide the construction team to locally increase the support stiffness or adjust the construction progress to reduce potential risks.

[0166] Step S105, the deformation rate V1 of the support structure and the internal force change trend AF2 are extracted from the optimized deformation field U3 and the internal force field F3, if the deformation rate V1 exceeds 2mm / day or the internal force change trend AF2 exceeds 10% of the design value, the model M1 is further corrected by using real-time monitoring data D, S, the high-risk area abnormal point generation, construction parameter optimization and dynamic evolution simulation are circularly executed, and the dynamically adjusted construction control parameter set Q1 is generated.

[0167] The deformation rate V1 of the support structure is extracted from the deformation field U3 of the finite element analysis, and the internal force change trend AF2 is extracted from the internal force field F3, to obtain the initial deformation rate V1 and the internal force change trend AF2. If the deformation rate V1 exceeds the preset threshold or the internal force change trend AF2 exceeds the design value, the real-time monitoring data D and S are obtained, the real-time monitoring data D and S are fused by using the weighted average method to obtain the fused monitoring data D1. Error analysis is performed on the fused monitoring data D1 and the prediction model M1, the parameters of the prediction model M1 are corrected by using the least square method to obtain the corrected model M2. According to the corrected model M2, the finite element analysis is performed to generate the abnormal point set P1 of the high-risk area, and the distribution characteristics of the abnormal point set P1 are determined. Through the correlation analysis of the abnormal point set P1 and the construction parameter Q1, the construction parameter Q1 is optimized by using the genetic algorithm to obtain the optimized construction parameter Q2. According to the optimized construction parameter Q2, the dynamic evolution simulation is performed to generate the prediction result R1 of the deformation and internal force distribution of the support structure. Through the comparative analysis of the prediction result R1 and the fused monitoring data D1, the corrected model M2 is updated to obtain the updated model M3.

Claims

1. A method for predicting the deformation of support structures in deep silty foundation pits covering tens of thousands of square meters, characterized in that... Includes the following steps: S101 obtains real-time monitoring data of rheological soil from the construction site, and constructs a soil-structure interaction model by combining it with preset rheological model parameters to obtain the initial deformation field and internal force field; S102 updates the soil-structure interaction model by real-time monitoring data, generating optimized deformation and internal force fields; S103 If the optimized deformation field or internal force field exceeds a preset threshold, an early warning signal is generated; S104 Based on the warning signal, an optimization algorithm is used to adjust the soil-structure interaction model to obtain an optimized set of construction parameters; S105 updates the soil-structure interaction model based on the optimized set of construction parameters to predict the deformation field and internal force field of the next construction stage; Step S103 includes: Displacement and internal force sequences of the support structure are extracted from real-time monitoring data, and denoising is performed using a particle filter algorithm to obtain denoised displacement and internal force sequences. The deviation between the denoised sequence and the predicted deformation field and internal force field is calculated. If the deviation exceeds a preset threshold, an early warning signal is generated. By updating the rheological model parameters based on the normal prior distribution using the Bayesian method, a modified soil-structure interaction model is generated.

2. The deformation prediction method for deep foundation pit support structures with siltation up to 10,000 square meters as described in claim 1, characterized in that, Step S101 includes: Data on stress, displacement, and pore water pressure of rheological soil were obtained from sensors at the construction site. Based on the Burgers model parameters determined in the laboratory, including the viscoelastic modulus and viscosity coefficient, an initial soil-structure interaction model was constructed using the finite element analysis method, and the initial deformation field and internal force field of the support structure were calculated.

3. The deformation prediction method for deep foundation pit support structures with siltation up to 10,000 square meters as described in claim 1, characterized in that, Step S102 includes: The time step method is adopted to determine the time step based on the daily progress of the construction stage, and to simulate the stress relaxation and creep effects caused by soil rheology. Update the soil stiffness matrix and damping matrix, and calculate the dynamically evolving deformation field and internal force field through finite element analysis; By fusing real-time monitoring data and prediction results using the Kalman filter algorithm, the optimized deformation field and internal force field are obtained.

4. The deformation prediction method for deep foundation pit support structures with siltation up to 10,000 square meters as described in claim 1, characterized in that, Step S104 includes: Abnormal deformation points and internal force points in high-risk areas are obtained from early warning signals to analyze soil-structure interaction characteristics; A genetic algorithm is used to optimize the stiffness and prestress distribution of the support system, with the objective function of minimizing deformation and internal force, and the constraints of construction cost and time, to generate an optimized set of construction parameters.

5. The method for predicting the deformation of a 10,000-square-meter-scale silty deep foundation pit support structure as described in claim 1, characterized in that, Step S105 includes: The soil-structure interaction model is updated by updating the optimized set of construction parameters, and the dynamic evolution of the construction process is simulated by the finite element method to obtain the optimized deformation field and internal force field. The deformation rate and internal force change trend are extracted from the optimized deformation field and internal force field. If they exceed the preset threshold, the model is corrected by real-time monitoring data to generate a set of dynamically adjusted construction control parameters.

Images