Foundation pit construction method and system based on BIM

Abstract

The invention discloses a foundation pit construction method and system based on BIM, and relates to foundation pit monitoring. The foundation pit construction method comprises the steps that a foundation pit BIM model is established; parameter distribution is initialized, the elastic modulus, the cohesive force and the internal friction angle of the soil body unit and the contact rigidity parameter of the contact surface unit serve as parameters to be updated, and when n is equal to 1, initial prior distribution of the parameters to be updated is established according to engineering geological survey parameters; when ngt; when 1 is greater than 1, taking the posterior distribution of the parameters obtained in the (n-1) th excavation step as the prior distribution of the nth excavation step; for the nth excavation step, according to the parameter expected value in the prior distribution, iteratively calculating the soil pressure distribution and the support structure deformation through the foundation pit BIM model, and obtaining a foundation pit deformation predicted value of the nth excavation step; constructing a comprehensive likelihood function according to the accumulated foundation pit deformation measured values of the first excavation step to the nth excavation step, and obtaining posterior distribution of to-be-updated parameters through Bayesian updating; the deep foundation pit deformation monitoring precision is improved.

Description

Technical Field

[0001] The present application relates to the field of foundation pit detection, and in particular to a BIM-based foundation pit construction method and system. Background Art

[0002] With the rapid development of urban underground space, deep foundation pit projects are increasing in number, characterized by large excavation depths, complex surrounding environments, and high construction risks. Deformation control during foundation pit excavation is crucial for ensuring project safety. Accurately predicting foundation pit deformation is crucial for optimizing design options, guiding construction decisions, and ensuring the safety of surrounding structures. Traditional methods for predicting foundation pit deformation rely primarily on empirical formulas and numerical simulations. Numerical simulations, which can account for complex geological conditions, support structure types, and construction procedures, have become the primary prediction tool in engineering practice. However, the accuracy of numerical simulations is highly dependent on the reliability of input parameters, particularly soil physical and mechanical parameters (such as elastic modulus, cohesion, and internal friction angle) and soil-structure interface parameters. In practical projects, these parameters are typically obtained through limited geological surveys. Due to the spatial variability of geological conditions and the limitations of sampling and testing, significant uncertainty exists in the initial parameters. Furthermore, as the excavation depth increases, errors in the initial parameters lead to cumulative and amplified prediction biases, resulting in significant inaccuracy in deformation prediction during the deep excavation phase, posing a significant risk to project safety.

[0003] In recent years, Building Information Modeling (BIM) technology has been widely used in foundation pit projects, providing a platform for the integrated management of multi-source information, including design, construction, and monitoring. Furthermore, with the advancement of construction monitoring technology, the acquisition of foundation pit deformation monitoring data has become more convenient and accurate. However, how to fully utilize this monitoring data accumulated during construction, dynamically update model parameters, and improve prediction accuracy has become a pressing technical challenge.

[0004] The existing parameter updating methods have the following main problems:

[0005] Most methods only use the monitoring data of the current excavation step for parameter correction, failing to fully utilize historical monitoring information, resulting in poor stability in parameter identification. When using single-point optimization or simple back-analysis methods to update parameters, overfitting is prone to occur, causing the parameters to oscillate repeatedly between different excavation steps, affecting the continuity and reliability of the prediction. Traditional methods treat parameters as fixed values, ignoring the uncertainty characteristics of the parameters. They are unable to quantify the credibility of the prediction results, nor can they achieve the reasonable transfer of uncertainty throughout the construction process. As the number of excavation steps increases, if all parameters need to be re-identified each time, the amount of calculation will increase dramatically, affecting the real-time performance of engineering applications.

[0006] Therefore, it is urgent to develop a foundation pit construction method that can make full use of the monitoring information of the entire construction process, realize the reasonable transfer of parameter uncertainty, and ensure the stability and accuracy of prediction to meet the actual needs of safety control of deep foundation pit projects. Summary of the Invention

[0007] In view of the fact that the prediction of deformation during deep foundation pit construction suffers from the cumulative amplification of parameter uncertainty, resulting in low prediction accuracy, this application provides a BIM-based foundation pit construction method and system. By using the posterior distribution of the parameters of the previous excavation step as the prior distribution of the subsequent steps, and using the accumulated monitoring data to construct a comprehensive likelihood function for parameter update, the gradual convergence of parameter uncertainty can be achieved, thereby improving the accuracy of deep foundation pit deformation monitoring.

[0008] One aspect of the present application provides a BIM-based foundation pit construction method, comprising: S1, establishing a foundation pit BIM model including soil units, support structure units, and soil-structure contact surface units; S2, initializing parameter distribution, and taking the elastic modulus, cohesion, internal friction angle of the soil unit, and the contact stiffness parameter of the contact surface unit as parameters to be updated: when n=1, establishing an initial prior distribution of the parameters to be updated based on engineering geological survey parameters; when n>1, taking the posterior distribution of the parameters obtained in the n-1th excavation step as the prior distribution of the nth excavation step; S3, for the nth excavation step, iteratively calculating the soil pressure distribution and the support structure deformation through the foundation pit BIM model based on the expected value of the parameters in the prior distribution, and obtaining a foundation pit deformation prediction for the nth excavation step. value; S4, perform the actual construction of the nth excavation step, and obtain the measured value of the foundation pit deformation through the preset monitoring points; S5, calculate the root mean square error between the predicted value of the foundation pit deformation and the measured value of the deformation at each monitoring point, keep the current parameters unchanged, and use the prior distribution determined in S2 directly as the posterior distribution; it indicates that the current parameter prediction accuracy meets the engineering requirements and there is no need to update the parameters; when the root mean square error is greater than the preset threshold: construct a comprehensive likelihood function based on the cumulative measured values ​​of the foundation pit deformation from the 1st to the nth excavation steps; according to the prior distribution determined in S2 and the comprehensive likelihood function, obtain the posterior distribution of the parameters to be updated through Bayesian updating; extract the parameter expected value from the posterior distribution, and update the corresponding parameters in the foundation pit BIM model; S6, when n is less than the total excavation steps, let , return to step S2; S7, store the predicted deformation value, measured deformation value and parameter posterior distribution of each excavation step in the foundation pit BIM model. Where n is a positive integer, that is, the number of excavation steps.

[0009] Soil elements are finite element units used to simulate the mechanical behavior of soil in the foundation pit BIM model. In numerical analysis, the continuous soil mass is discretized into numerous small units, each of which carries the soil's basic mechanical properties, including elastic modulus, Poisson's ratio, specific gravity, cohesion, and angle of internal friction. Soil elements are interconnected through nodes to simulate the stress-strain relationship, strength characteristics, and deformation behavior of the soil during excavation.

[0010] Support structure elements are finite element units used to simulate the mechanical properties of foundation pit support systems. They primarily include retaining wall elements (such as underground diaphragm walls and SMW piles), support elements (such as reinforced concrete supports and steel supports), and anchor elements. Each support structure element has corresponding material properties, such as elastic modulus, flexural stiffness, and axial stiffness, which are used to simulate the stress-deformation response of the support structure under earth pressure.

[0011] The soil-structure interface element is a special element placed between the soil element and the supporting structure element. It is used to simulate the interaction mechanism at the soil-support interface. This element accounts for nonlinear contact behaviors such as relative slip and separation that may occur between the soil and the structure. It can transmit normal and tangential forces and allows for interface failure or slip when certain conditions are met.

[0012] Contact stiffness parameters are key parameters describing the mechanical properties of soil-structure interface elements. They primarily include normal and tangential contact stiffness. Normal contact stiffness controls the deformation resistance of the contact surface in the vertical direction, influencing the normal stress transfer between the soil and the structure. Tangential contact stiffness controls the shear deformation resistance of the contact surface, influencing the development of interfacial friction. These parameters directly influence the interaction between the soil and the supporting structure and are crucial factors influencing the accuracy of foundation pit deformation calculations.

[0013] Furthermore, when n=1, the initial prior distribution of the parameters to be updated is established based on the engineering geological survey parameters, including: calculating the mean and standard deviation of each parameter to be updated based on multiple drilling data from the engineering geological survey; using the truncated normal distribution as the prior distribution of the elastic modulus and contact stiffness parameters; and using the transformed form of the β distribution as the prior distribution of the cohesion and internal friction angle.

[0014] Among them, the truncated normal distribution refers to a probability distribution form that truncates the distribution by setting upper and lower boundaries on the basis of the traditional normal distribution. In this application, the truncated normal distribution is used to establish the prior distribution of the elastic modulus and contact stiffness parameters. The value range of the normal distribution itself is from negative infinity to positive infinity, but engineering parameters such as elastic modulus and contact stiffness must be positive in the physical sense, and usually have reasonable upper and lower limits. Through truncation processing, the parameters can be restricted to a physically reasonable range, such as the truncation interval of the elastic modulus is set to ,in , .

[0015] The transformation form of β distribution refers to the method of modeling by using β distribution after mapping the original parameters to the interval [0, 1] through linear transformation. In this application, it is used to establish the prior distribution of cohesion and internal friction angle parameters. First, the original parameters are normalized, such as the normalization of cohesion is , so that its value range becomes [0, 1]; then the normalized parameters are treated with β distribution Modeling is performed and the distribution parameters α and β are determined by the moment estimation method.

[0016] Furthermore, the truncated normal distribution is used as the prior distribution of elastic modulus and contact stiffness parameters, including: obtaining the original parameter data set D from engineering geological survey parameters, ,in, is the elastic modulus of the ith drill hole, For cohesion, is the internal friction angle; calculate the sample mean of the elastic modulus and standard deviation ; Calculate the sample mean of cohesion and standard deviation ; Calculate the sample mean of the internal friction angle and standard deviation ; Set truncated normal distribution ,in, , ; The probability density function is: ; where Z is the normalization constant, ; Establish the truncated normal distribution prior of the contact stiffness parameter k: Estimate the initial value of the contact stiffness based on the soil stiffness: , ; Set truncated normal distribution ,in , .

[0017] Furthermore, a transformed form of β distribution is used as the prior distribution of cohesion and internal friction angle, including normalizing the cohesion: ,in, , ; Using beta distribution Modeling ,in: ; ; , ; Establish the internal friction angle Transformation of the β distribution prior: Normalize the internal friction angle: ,in, , ; Using beta distribution Modeling , the parameter calculation method is the same as ; The prior distribution parameters of each parameter As the initial prior distribution information.

[0018] Furthermore, when n>1, the posterior distribution of the parameters obtained in the n-1th excavation step is used as the prior distribution of the n-1th excavation step, including: calculating the information entropy of the posterior distribution of the n-1th step; when the information entropy is less than the set threshold, it indicates that the parameter is a converged parameter, the uncertainty of the corresponding parameter has been reduced, and the fixed corresponding parameter is no longer updated; the converged parameter is removed from the set of parameters to be updated to reduce the calculation dimension.

[0019] Among them, in this application, uncertainty specifically refers to the epistemic uncertainty in the parameter estimation process, that is, the degree to which the true value of the parameter is unknown due to insufficient information. The level of uncertainty is quantified by calculating the information entropy of the posterior distribution of the parameter. The smaller the information entropy, the more concentrated the parameter distribution and the lower the uncertainty; the larger the information entropy, the more dispersed the parameter distribution and the higher the uncertainty. When the information entropy of the posterior distribution of the n-1th step is less than the set threshold, it is determined that the corresponding parameter has converged and its uncertainty has been reduced to an acceptable level, and it can be fixed and no longer updated. This mechanism realizes the automatic identification of the parameter convergence state, effectively reducing the dimension and complexity of subsequent calculations.

[0020] Furthermore, based on the cumulative measured values ​​of foundation pit deformation from the 1st to the nth excavation steps, a comprehensive likelihood function is constructed, including: extracting the measured values ​​of foundation pit deformation from the 1st to the nth excavation steps and the corresponding predicted values ​​of foundation pit deformation from the foundation pit BIM model to form a cumulative data set; setting a time weight for the data of the i-th excavation step , where i = 1, 2, ..., n, so that the data of the recent excavation steps have higher weights; the weighted prediction error is calculated based on the accumulated data set and the time weight; the error covariance matrix is ​​constructed according to the monitoring point type and the corresponding measurement accuracy. The monitoring point types include support structure horizontal displacement measurement points, surface settlement measurement points, and soil displacement measurement points; based on the weighted prediction error and the error covariance matrix, a comprehensive likelihood function is constructed. The comprehensive likelihood function reflects the matching degree between the parameters to be updated and the accumulated monitoring data.

[0021] The error covariance matrix describes the correlation between measurement errors and the differences in measurement accuracy between different monitoring points. The diagonal elements of the matrix reflect the measurement variance of each monitoring point, reflecting the differences in accuracy of different types of monitoring equipment (for example, support structure horizontal displacement measurement points, surface settlement measurement points, and soil displacement measurement points have different measurement accuracies). The off-diagonal elements reflect the covariance of measurement errors between different monitoring points, taking into account the influence of spatial correlation and systematic errors. By constructing the error covariance matrix, the reliability of different monitoring data can be reasonably weighed during the Bayesian update process, giving high-precision monitoring data greater weight. At the same time, the correlation between monitoring points is considered to avoid the impact of repeated calculation of related information on parameter update results.

[0022] The comprehensive likelihood function is a statistical function constructed based on the cumulative monitoring data from the first to the nth excavation steps. It quantifies the degree of match between the updated parameters and the historical monitoring data. This function comprehensively considers the time weighting effect (recent data is given a higher weight), the accuracy differences among multiple types of monitoring data (reflected by the error covariance matrix), and the fusion of information from the cumulative data. Compared to traditional likelihood functions based solely on single-step monitoring data, the comprehensive likelihood function utilizes information from the entire construction process.

[0023] Furthermore, the comprehensive likelihood function expression is:

[0024] ;

[0025] Among them, θ is the parameter vector to be updated, including the unconverged soil unit parameters and contact surface unit parameters; Y is the cumulative foundation pit deformation measured value matrix, , is the p-dimensional measured value vector of the i-th excavation step; n is the number of excavation steps currently completed; p is the total number of monitoring points of a single excavation step; W is the time weight matrix, , , is the p×p identity matrix; is the error covariance matrix, reflecting the measurement accuracy differences of different types of monitoring points; is the weighted prediction error vector, , is the cumulative foundation pit deformation prediction value matrix, vec(*) represents the matrix vectorization operation; Represents the Kronecker product operation of matrices.

[0026] Furthermore, according to the prior distribution determined in S2 and the comprehensive likelihood function, the posterior distribution of the parameter to be updated is obtained through Bayesian updating, including: according to Bayesian theorem, combining the prior distribution and the comprehensive likelihood function, constructing the posterior distribution expression of the parameter to be updated ,in, is the prior distribution, is a comprehensive likelihood function; a Markov chain Monte Carlo method is used to generate a parameter sample set from the posterior distribution; a rejection sampling strategy is used to perform constraint testing on the parameter sample set generated by sampling so that the sampling results meet the physical constraints; based on the parameter sample set that meets the physical constraints, the statistical characteristics of the posterior distribution are calculated as the posterior distribution of the parameters to be updated, and the statistical characteristics include the posterior mean and the posterior covariance.

[0027] In this application, the Markov Chain Monte Carlo method is a numerical calculation method for extracting parameter samples from a complex posterior distribution. Specifically, the Metropolis-Hastings algorithm is used to generate candidate parameters based on the proposed distribution starting from the current parameter value, and the acceptance-rejection criterion is used to decide whether to accept the candidate value, thereby generating a parameter sample sequence. , where each Contains elastic modulus E, contact stiffness k, normalized cohesion and normalized internal friction angle .

[0028] The rejection sampling strategy is a quality control measure to ensure that the parameter samples generated by MCMC meet the engineering physics meaning. Parameter boundary constraint test: The elastic modulus must be within range, ensuring that no negative or abnormally large values ​​appear; the contact stiffness must be within The normalized parameters must be within the range of [0, 1] to ensure that the physical parameters after the inverse transformation are meaningful. The negative correlation constraint between cohesion and internal friction angle is: , reflecting the physical relationship between soil strength parameters. This constraint, based on the Mohr-Coulomb strength theory, ensures that the combination of soil strength parameters conforms to the actual mechanical properties of the soil. Samples that do not meet the constraints are directly eliminated, and only physically reasonable samples are retained for posterior statistical calculations.

[0029] Furthermore, the rejection sampling strategy is used to perform constraint checks on the parameter sample set generated by sampling, so that the sampling results meet the physical constraints, including: for each parameter sample obtained by sampling , perform the following constraint checks: Check the elastic modulus constraint: If or , then reject the sample; check the contact stiffness constraint: if or , then reject the sample; test the normalized cohesion constraint: if or , then reject the sample; test the normalized internal friction angle constraint: if or , then reject the sample; for samples that pass the constraint test, further test the correlation constraints between parameters: convert the normalized parameters into original parameters: ; ; Check the negative correlation constraint between cohesion and internal friction angle: If , then reject the sample, where is the empirical coefficient, is the upper limit value. , the specific value is determined according to the soil type: Clay soil: ; Soft plastic clay: ;Hard plastic clay: ;Silty soil: Silty clay: ; Sandy silt: ; Sandy soil: ; Fine sand: Medium coarse sand: ; ,in, is the initial mean value of cohesion.

[0030] Physical constraints refer to the engineering mechanics and physical laws that soil parameters must meet, including: Elastic modulus E: must be positive and within a reasonable range ( arrive ), to avoid unreasonable soils that are too soft or too hard; cohesion c: must be non-negative, with an upper limit of 2 times the initial mean, in line with the strength characteristics of common soils; internal friction angle φ: limited to between 0° and 45°, in line with the typical range of soil friction angles in soil mechanics; contact stiffness k: kept within a reasonable range, usually 0.1 to 2 times the soil stiffness. Strength parameter constraints: There is a negative correlation between cohesion and internal friction angle, that is, soils with high cohesion tend to have a small internal friction angle (such as clay), while soils with a large internal friction angle have a small cohesion (such as sand). This constraint is expressed through the empirical formula implementation, where is the empirical coefficient, These constraints ensure that the updated parameters not only mathematically make the predicted values ​​consistent with the measured values, but more importantly, ensure that the parameter combination is reasonable within the theoretical framework of soil mechanics, avoiding parameter combinations that are actually infeasible.

[0031] Another aspect of the present application also provides a BIM-based foundation pit construction system, including: a BIM modeling module, which establishes a foundation pit BIM model including soil units, support structure units and soil-structure contact surface units; a parameter setting module, which initializes parameter distribution, takes the elastic modulus, cohesion, internal friction angle of the soil unit, and the contact stiffness parameter of the contact surface unit as parameters to be updated, and when n=1, establishes an initial prior distribution of the parameters to be updated based on engineering geological survey parameters; when n>1, takes the posterior distribution of parameters obtained in the n-1th excavation step as the prior distribution of the nth excavation step; a deformation prediction module, which, for the nth excavation step, iteratively calculates the soil pressure distribution and support structure deformation through the foundation pit BIM model according to the parameter expected value in the prior distribution, and obtains the nth The predicted value of foundation pit deformation in the excavation step; the construction monitoring module, which obtains the measured value of foundation pit deformation through preset monitoring points during the actual construction process of the nth excavation step; the parameter updating module, which calculates the root mean square error between the predicted value of foundation pit deformation and the measured value of deformation at each monitoring point. When the root mean square error is greater than the preset threshold, a comprehensive likelihood function is constructed according to the cumulative measured values ​​of foundation pit deformation from the 1st to the nth excavation steps. According to the prior distribution and the comprehensive likelihood function, the posterior distribution of the parameters to be updated is obtained through Bayesian updating, and the expected value of the parameter is extracted from the posterior distribution to update the corresponding parameters in the foundation pit BIM model; the data storage module, which stores the predicted value of foundation pit deformation, the measured value of deformation and the posterior distribution of parameters of each excavation step in the foundation pit BIM model; the processor, which executes the calculation and control of each module.

[0032] Compared with the existing technology, the advantages of this application are:

[0033] In traditional foundation pit deformation prediction methods, the initial parameters are determined based on limited geological survey data and are subject to significant uncertainty. As excavation depth increases, the prediction deviation caused by parameter errors gradually increases, seriously affecting the safety control of deep excavation. This application establishes a sequential Bayesian parameter update mechanism, using the posterior distribution of the parameters of the previous excavation step as the prior distribution of the next step, thereby achieving continuous transmission of parameter uncertainty information and avoiding information loss. At the same time, as construction progresses and monitoring data continues to accumulate, the variance of the posterior distribution of the parameters gradually decreases, achieving asymptotic convergence of parameter uncertainty.

[0034] This application constructs a comprehensive likelihood function to fully utilize all the accumulated monitoring data from the 1st to the nth excavation steps for parameter update, rather than relying solely on the single monitoring data of the current step, thereby greatly improving the stability and reliability of parameter identification. The time weight mechanism is combined to give more weight to recent data, and the rejection sampling strategy is adopted to ensure that the parameters meet the physical constraints, effectively avoiding the overfitting phenomenon that is prone to occur in a single parameter update and the problem of repeated oscillation of parameters between different excavation steps. In this way, the accuracy of foundation pit deformation prediction gradually improves as construction progresses, providing a reliable technical basis for the safety control of the entire process of deep foundation pit construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The present application will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein:

[0036] Figure 1 is an exemplary flow chart of a BIM-based foundation pit construction method according to some embodiments of the present application;

[0037] Figure 2 is an exemplary flow chart of constructing an initial prior distribution according to some embodiments of the present application;

[0038] Figure 3 This is an exemplary flowchart of constructing a comprehensive likelihood function according to some embodiments of the present application. DETAILED DESCRIPTION

[0039] The method and system provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0040] like Figure 1 As shown in Figure 1, S1, establishes a foundation pit BIM model including soil units, support structure units and soil-structure interface units; soil units: based on geological survey data, the soil within the influence range of the foundation pit is divided into multiple calculation units, and each unit is assigned initial physical and mechanical parameters; support structure units: including structural components such as underground continuous walls and support systems, their geometric dimensions and material properties are defined; soil-structure interface units: contact interface units are set between the soil and the support structure to simulate soil-structure interaction; monitoring point layout: monitoring points are preset in the model, including support structure horizontal displacement measurement points, surface settlement measurement points and soil deep displacement measurement points.

[0041] S2, the elastic modulus E, cohesion c, internal friction angle φ of the soil unit, and the contact stiffness parameter k of the contact surface unit are used as the parameters to be updated. Different initialization strategies are adopted according to the different excavation steps n:

[0042] like Figure 2 As shown in Figure 1, when n=1, the initial prior distribution of the parameters to be updated is established based on the engineering geological survey parameters;

[0043] S2-1: Extract soil physical and mechanical parameter test data of m boreholes from the engineering geological survey report to obtain the original parameter data set ,in, is the elastic modulus of the i-th drill hole, For cohesion, is the internal friction angle;

[0044] S2-2: Perform statistical processing on the original parameter data set D:

[0045] Calculate the sample mean of the elastic modulus and standard deviation , the sample mean of cohesion and standard deviation , and the sample mean of the internal friction angle and standard deviation These statistics will serve as the basis for constructing the prior distribution parameters.

[0046] S2-3: Establish a truncated normal distribution prior for the elastic modulus E. Considering that the elastic modulus must be positive and cannot deviate too much from the actual value, a truncated normal distribution is used to model it. Set the truncation boundary is 0.1 times the mean, The probability density function is constructed and the normalization constant Z is calculated to ensure that the probability density integral is 1.

[0047] Specifically, set the truncated normal distribution ,in, , ; The probability density function is: ; where Z is the normalization constant, ;

[0048] S2-4: Establish the transformation β distribution prior of cohesion c. To ensure that the cohesion changes within a bounded range, first perform a normalization transformation to map the cohesion to the interval [0, 1]. Set the transformation boundary is 0, The normalized cohesion is modeled using β, and the shape parameters of β are calculated using the moment matching method. and .

[0049] Specifically, the cohesion is normalized: ,in, , ; Using beta distribution Modeling ,in: ; ; , ;

[0050] S2-5: Establishing the internal friction angle The transformation β distribution prior of the internal friction angle is treated similarly to the cohesion. Set physically reasonable transformation boundaries is 0°, The angle is 45°, and β modeling is adopted after normalization transformation.

[0051] Specifically, the internal friction angle is normalized: , using the beta distribution Modeling , parameter calculation method is the same as S2-4;

[0052] S2-6: Establish a truncated normal distribution prior for the contact stiffness parameter k. The contact stiffness parameter is estimated based on the soil stiffness, with an initial value set to 0.5 times the mean elastic modulus and a standard deviation set to 30% of the initial value. Use a truncated normal distribution model to ensure that the parameter is within a reasonable range.

[0053] Specifically, the initial contact stiffness is estimated based on the soil stiffness: , , set the truncated normal distribution ,in, , ;

[0054] S2-7: The prior distribution parameters of each parameter As the initial prior distribution information, it is used for deformation prediction calculation in S3.

[0055] When n>1, the posterior distribution of the parameters obtained in the n-1th excavation step is used as the prior distribution of the nth excavation step;

[0056] S2-8: Read the posterior distribution information of the parameters to be updated obtained by S5 in the n-1th excavation step from the foundation pit BIM model, including the posterior mean vector and the posterior covariance matrix ;

[0057] S2-9: Calculate the information entropy of each parameter to be updated to assess parameter uncertainty. This calculation quantifies the degree of uncertainty in each parameter. For parameters of a truncated normal distribution, the differential entropy is calculated taking into account the truncation effect; for parameters of a beta distribution, the information entropy is calculated directly. Lower information entropy indicates lower parameter uncertainty.

[0058] Specifically, for the elastic modulus E that obeys the truncated normal distribution, its information entropy ,in, for where is the variance of E, and Φ is the cumulative function of the standard normal distribution;

[0059] For the contact stiffness k that obeys the truncated normal distribution, the information entropy The calculation method is the same as ;

[0060] For the normalized cohesion obeying the β distribution , its information entropy , where B is the β function and ψ is the digamma function;

[0061] For the normalized internal friction angle that follows the β distribution , information entropy The calculation method is the same as ;

[0062] S2-10: Determine parameter convergence and update the set of parameters to be updated. Set the information entropy threshold to 1.0. When the information entropy of a parameter falls below this threshold, the corresponding parameter is considered to have fully converged. Fix the converged parameter to its posterior mean and remove it from the set of parameters to be updated to reduce the dimensionality of subsequent calculations.

[0063] Specifically, set the information entropy threshold ;

[0064] when When the elastic modulus E is fixed to , removed from the set of parameters to be updated;

[0065] when When the contact stiffness k is fixed to , removed from the set of parameters to be updated;

[0066] when When the cohesion c is fixed at , removed from the set of parameters to be updated;

[0067] when When the internal friction angle φ is fixed at , removed from the set of parameters to be updated;

[0068] S2-11: For parameters that have not converged, a new prior distribution is constructed based on the posterior distribution information. Keeping the distribution type unchanged, update the distribution parameters to the posterior values ​​and construct the prior distribution of the nth excavation step:

[0069] For the unconverged elastic modulus E, and Construct a new truncated normal distribution ;

[0070] For the unconverged contact stiffness k, and Construct a new truncated normal distribution ;

[0071] For the non-convergent cohesion c, the β distribution parameters are updated according to the posterior distribution parameters , ;

[0072] For the non-convergent internal friction angle , update the beta distribution parameters according to the posterior distribution parameters , ;

[0073] S2-12: The updated prior distribution parameters and the converged fixed parameter values ​​are passed to S3 for deformation prediction calculation of the nth excavation step.

[0074] S3, for the nth excavation step, based on the expected values ​​of the parameters in the prior distribution, the soil pressure distribution and the deformation of the support structure are iteratively calculated through the foundation pit BIM model to obtain the predicted value of the foundation pit deformation for the nth excavation step, specifically including:

[0075] S3-1: Extract the expected values ​​of the parameters from the prior distribution determined by S2. Extract the expected values ​​of each parameter from the prior distribution determined by S2. Normalized parameters need to be converted to their original physical parameters through an inverse transformation. For parameters that have converged, their fixed values ​​are used directly.

[0076] Specifically, for the unconverged elastic modulus E, extract the expected value ; For the unconverged contact stiffness k, extract the expected value ; For the unconverged normalized cohesion , extract the expected value and convert it: ; For the non-converged normalized internal friction angle , extract the expected value and convert it: ;

[0077] For converged parameters, use the fixed values ​​determined by S2-10;

[0078] S3-2: Assign the extracted parameter values ​​to the corresponding units of the foundation pit BIM model. Assign the soil parameters to the soil units within the excavation influence range, and assign the contact stiffness to the soil-structure interface unit: 、 、 Assign to the soil units within the influence range of the nth excavation step; Assign values ​​to the soil-structure interface element;

[0079] S3-3: Calculate the active earth pressure distribution at the nth excavation step based on the updated parameters: Calculate the active earth pressure coefficient using Coulomb earth pressure theory ; Considering the influence of cohesion, calculate the active earth pressure at depth z: ; Wherein, γ is the soil mass, which is determined by the soil properties in the foundation pit BIM model;

[0080] S3-4: Construct the overall stiffness matrix considering the soil, support structure and contact surface. Calculate the equivalent node load vector based on the soil pressure distribution and form the finite element equilibrium equation: Construct the overall stiffness matrix ;in, is the soil stiffness matrix, Sure; is the stiffness matrix of the supporting structure; is the contact surface stiffness matrix, Determine; apply earth pressure load vector , is the shape function;

[0081] S3-5: Iterative solution of support structure deformation and soil displacement: Solve the equation K×u = F to obtain the node displacement vector u; use the Newton-Raphson iterative method to deal with nonlinear problems, and the convergence criterion is ; Update the earth pressure distribution and recalculate until convergence;

[0082] S3-6: Extract the predicted value of foundation pit deformation from the converged solution: interpolate and calculate the deformation value at the preset monitoring points defined in S4; form the predicted value vector of foundation pit deformation for the nth excavation step ;Will Stored in the foundation pit BIM model for error calculation in S5.

[0083] S4, performing actual construction of the nth excavation step and obtaining actual measured values ​​of foundation pit deformation through preset monitoring points;

[0084] like Figure 3 As shown, S5 calculates the root mean square error between the predicted value and the measured value of the foundation pit deformation at each monitoring point. When the root mean square error is greater than the preset threshold:

[0085] S5-1: Extract historical data of the 1st to nth excavation steps from the foundation pit BIM model: For the i-th excavation step , extract the measured value vector of the foundation pit deformation , where p is the number of monitoring points; extract the corresponding foundation pit deformation prediction value vector ; Form a cumulative measured data matrix and the cumulative prediction data matrix ;

[0086] S5-2: Set the time weighting coefficient matrix: assign weight coefficients to the data of the i-th excavation step , so that recent data has higher weight; construct a time weight matrix ,in, is the p×p identity matrix;

[0087] S5-3: Calculating weighted prediction error: Calculating the original error matrix ; Calculate the weighted error vector , where vec represents a matrix vectorization operation; the vector reflects the overall deviation of the model prediction.

[0088] S5-4: Construct the error covariance matrix, considering the measurement accuracy of different types of monitoring data: the monitoring points are divided into three categories: support structure horizontal displacement measurement points (accuracy ), surface settlement measuring points (accuracy ), soil displacement measurement points (accuracy ); Construct measurement accuracy matrix ,in, are the unit matrices of the number of various measurement points; forming a complete covariance matrix , where kron represents the Kronecker product; this matrix takes into account the differences in measurement uncertainty of different monitoring items.

[0089] S5-5: Constructing an integrated likelihood function based on the multivariate normal distribution assumption: assuming weighted errors Obey multivariate normal distribution ,in, represents the Kronecker product; the integrated likelihood function is: ;in, is the parameter vector to be updated, including the non-converged parameters determined by S2-11;

[0090] S5-6: Take the logarithm of the likelihood function to obtain the log-likelihood function: ; Use the log-likelihood function for subsequent Bayesian update calculations. Convert the likelihood function to logarithmic form to facilitate numerical calculations and avoid numerical underflow.

[0091] S5-7: According to Bayes' theorem, establish the relationship between the posterior distribution, prior distribution and likelihood function:

[0092] Posterior probability density: ;in, The prior distribution determined in S2 includes the following parameters that have not converged:

[0093] Truncated normal distribution of elastic modulus E ;

[0094] Truncated normal distribution of contact stiffness k ;

[0095] Normalized cohesion of ;

[0096] Normalized internal friction angle of ;

[0097] S5-8: Using Markov Chain Monte Carlo (MCMC) methods to sample from the posterior distribution:

[0098] Use the Metropolis-Hastings algorithm and set the proposed distribution to multivariate normal distribution;

[0099] Initialization parameter values is the expected value of the prior distribution;

[0100] Generate M = 10000 samples and set the burn-in period to B = 2000;

[0101] Calculate the acceptance rate and adjust the covariance matrix of the proposal distribution to ensure that the acceptance rate is between 20% and 50%;

[0102] S5-9: Calculating statistical characteristics of the posterior distribution based on MCMC samples:

[0103] Compute the posterior mean: ;

[0104] Compute the posterior covariance matrix: ;

[0105] The rejection sampling strategy is used to perform constraint checks on the parameter sample set generated by sampling, so that the sampling results meet the physical constraints, including: for each parameter sample obtained by sampling , perform the following constraint checks:

[0106] Check elastic modulus constraint: If or , then reject the sample;

[0107] Check contact stiffness constraint: If or , then reject the sample;

[0108] Test the normalized cohesion constraint: If or , then reject the sample;

[0109] Check the normalized internal friction angle constraint: If or , then reject the sample;

[0110] For samples that pass the constraint test, further test the correlation constraints between parameters:

[0111] Convert normalized parameters to original parameters: , ;

[0112] Check the negative correlation constraint between cohesion and internal friction angle: If , then reject the sample, where is the empirical coefficient, is the upper limit value;

[0113] S5-10: Convert the continuous distribution parameters back to the original physical parameters:

[0114] elastic modulus ,in, is the posterior mean of E;

[0115] Contact stiffness ,in, is the posterior mean of k;

[0116] Cohesion ;

[0117] internal friction angle ;

[0118] S5-11: Store the posterior distribution information for the next excavation step:

[0119] Save the posterior distribution parameters ;

[0120] Record parameter update history, including parameter value changes before and after the update;

[0121] The updated parameter value Replace the corresponding parameters in the foundation pit BIM model.

[0122] In step S6, if n is less than the total number of excavation steps, n is increased by 1 and the process returns to step S2 to continue the next iteration. This iterative mechanism enables the model parameters to be gradually optimized as construction progresses and monitoring data accumulates.

[0123] S7 stores the following information for each excavation step in the foundation pit BIM model: time-history curves of predicted and measured foundation pit deformation; the evolution of the posterior distribution of parameters; the changing trend of prediction errors; and the reduction of parameter uncertainty. Using the BIM platform's visualization capabilities, this information is intuitively displayed in charts, cloud maps, and other formats to support engineering decision-making. A data interface is also established to facilitate data exchange with other engineering management systems.

[0124] The invention of the present application and its implementation methods are described schematically above. This description is not restrictive. Without departing from the spirit or basic features of the present application, the present application can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention of the present application, and the actual structure is not limited to this. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and embodiment similar to the technical solution are designed without creativity, which should all fall within the scope of protection of the present application. In addition, the word "including" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" elements. Words such as first and second are used to indicate names and do not indicate any specific order.

Claims

1. A foundation pit construction method based on BIM, characterized in that: include: S1, establish a foundation pit BIM model including soil units, support structure units and soil-structure interface units; S2, initialize the parameter distribution, and use the elastic modulus, cohesion, internal friction angle of the soil unit, and the contact stiffness parameters of the contact surface unit as the parameters to be updated: When n=1, the initial prior distribution of the parameters to be updated is established based on the engineering geological survey parameters; When n>1, the posterior distribution of the parameters obtained in the n-1th excavation step is used as the prior distribution of the nth excavation step; S3, for the nth excavation step, based on the expected values ​​of the parameters in the prior distribution, the soil pressure distribution and the deformation of the supporting structure are iteratively calculated through the foundation pit BIM model to obtain the predicted value of the foundation pit deformation for the nth excavation step; where n is a positive integer; S4, performing actual construction of the nth excavation step and obtaining actual measured values ​​of foundation pit deformation through preset monitoring points; S5, calculate the root mean square error between the predicted value and the measured value of the foundation pit deformation at each monitoring point. When the root mean square error is greater than the preset threshold: Based on the measured values ​​of the accumulated pit deformation from the 1st to the nth excavation steps, a comprehensive likelihood function is constructed; According to the prior distribution determined in S2 and the comprehensive likelihood function, the posterior distribution of the parameters to be updated is obtained through Bayesian updating; Extract the expected value of the parameter from the posterior distribution and update the corresponding parameters in the foundation pit BIM model; S6, when n is less than the total excavation steps, let , return to step S2; S7, storing the predicted value of foundation pit deformation, the measured value of deformation, and the posterior distribution of parameters for each excavation step in the foundation pit BIM model.

2. The BIM-based foundation pit construction method according to claim 1, characterized in that: When n=1, the initial prior distribution of the parameters to be updated is established based on the engineering geological survey parameters, including: Calculate the mean and standard deviation of each parameter to be updated based on multiple drilling data from engineering geological surveys; The truncated normal distribution is used as the prior distribution of elastic modulus and contact stiffness parameters; The transformed form of β distribution is used as the prior distribution of cohesion and internal friction angle.

3. The BIM-based foundation pit construction method according to claim 2, characterized in that: The truncated normal distribution is used as the prior distribution of elastic modulus and contact stiffness parameters, including: Obtain the original parameter data set D from the engineering geological survey parameters, ,in, is the elastic modulus of the ith drill hole, For cohesion, is the internal friction angle; Calculate the sample mean of the elastic modulus and standard deviation ; Calculate the sample mean of cohesion and standard deviation ; Calculate the sample mean of the internal friction angle and standard deviation ; Setting truncated normal distribution ,in, , ; The probability density function is: ; Where Z is the normalization constant, ; Establish a truncated normal distribution prior for the contact stiffness parameter k: Estimate the initial contact stiffness based on the soil stiffness: , ; Setting truncated normal distribution ,in , .

4. The BIM-based foundation pit construction method according to claim 3, characterized in that: The transformed form of β distribution is used as the prior distribution of cohesion and internal friction angle, including: Normalize the cohesion: ,in, , ; Using beta distribution Modeling ,in: ； ； ， ； Establishing the internal friction angle The transformed beta distribution prior: Normalize the internal friction angle: ,in, , ; Using beta distribution Modeling , the parameter calculation method is the same as ; The prior distribution parameters of each parameter As the initial prior distribution information.

5. The BIM-based foundation pit construction method according to claim 4, characterized in that: When n>1, the posterior distribution of the parameters obtained in the n-1th excavation step is used as the prior distribution of the nth excavation step, including: Calculate the information entropy of the posterior distribution at step n-1; When the information entropy is less than the set threshold, it indicates that the parameter is a converged parameter, the uncertainty of the corresponding parameter has been reduced, and the fixed corresponding parameter is no longer updated; Remove the converged parameters from the set of parameters to be updated to reduce the computational dimension.

6. The BIM-based foundation pit construction method according to claim 5, characterized in that: According to the measured values ​​of the cumulative pit deformation from the 1st to the nth excavation steps, a comprehensive likelihood function is constructed, including: Extract the measured values ​​of foundation pit deformation from the 1st to the nth excavation steps and the corresponding predicted values ​​of foundation pit deformation from the foundation pit BIM model to form a cumulative data set; Set time weight for the data of the i-th excavation step , where i = 1, 2, ..., n, giving higher weight to the data of the recent excavation steps; Calculate the weighted prediction error based on the accumulated data set and time weight; According to the monitoring point type and the corresponding measurement accuracy, the error covariance matrix is ​​constructed. The monitoring point types include support structure horizontal displacement measurement points, surface settlement measurement points, and soil displacement measurement points. According to the weighted prediction error and the error covariance matrix, a comprehensive likelihood function is constructed, which reflects the matching degree between the parameters to be updated and the accumulated monitoring data.

7. The BIM-based foundation pit construction method according to claim 6, characterized in that: The comprehensive likelihood function expression is: ； Where θ is the parameter vector to be updated, including the unconverged soil unit parameters and contact surface unit parameters; Y is the matrix of accumulated foundation pit deformation measurements; n is the number of excavation steps currently completed; p is the total number of monitoring points in a single excavation step; W is the time weight matrix; is the error covariance matrix; is the weighted prediction error vector; Represents the Kronecker product operation of matrices.

8. The BIM-based foundation pit construction method according to claim 7, characterized in that: Obtain the posterior distribution of the parameters to be updated, including: According to Bayes' theorem, combining the prior distribution and the comprehensive likelihood function, the posterior distribution expression of the parameters to be updated is constructed ,in, is the prior distribution, is the comprehensive likelihood function; The Markov chain Monte Carlo method is used to generate parameter sample sets from the posterior distribution; The rejection sampling strategy is used to perform constraint checks on the parameter sample set generated by sampling, so that the sampling results meet the physical constraints. Based on the parameter sample set that meets the physical constraints, the statistical characteristics of the posterior distribution are calculated as the posterior distribution of the parameters to be updated. The statistical characteristics include the posterior mean and the posterior covariance.

9. The BIM-based foundation pit construction method according to claim 8, characterized in that: The rejection sampling strategy is used to perform constraint tests on the parameter sample set generated by sampling, including: For each parameter sample obtained , perform the following constraint checks: Check elastic modulus constraint: If or , then reject the sample; Check contact stiffness constraint: If or , then reject the sample; Test the normalized cohesion constraint: If or , then reject the sample; Check the normalized internal friction angle constraint: If or , then reject the sample; For samples that pass the constraint test, test the correlation constraints between parameters: Convert normalized parameters to original parameters: ; ; Check the negative correlation constraint between cohesion and internal friction angle: If , then reject the sample, where is the empirical coefficient, is the upper limit value.

10. A foundation pit construction system based on BIM, characterized in that: include: BIM modeling module, which establishes a foundation pit BIM model including soil units, support structure units, and soil-structure interface units; The parameter setting module initializes the parameter distribution and takes the elastic modulus, cohesion, internal friction angle of the soil unit, and the contact stiffness parameters of the contact surface unit as the parameters to be updated. When n=1, the initial prior distribution of the parameters to be updated is established based on the engineering geological survey parameters; when n>1, the posterior distribution of the parameters obtained in the n-1th excavation step is used as the prior distribution of the nth excavation step; The deformation prediction module calculates the soil pressure distribution and support structure deformation through the foundation pit BIM model according to the expected values ​​of the parameters in the prior distribution for the nth excavation step, and obtains the predicted value of the foundation pit deformation for the nth excavation step; The construction monitoring module obtains the measured values ​​of foundation pit deformation at preset monitoring points during the actual construction of the nth excavation step; The parameter update module calculates the root mean square error (RMS) between the predicted and measured values ​​of the foundation pit deformation at each monitoring point. When the RMS error is greater than a preset threshold, a comprehensive likelihood function is constructed based on the cumulative measured values ​​of the foundation pit deformation from the 1st to the nth excavation steps. Based on the prior distribution and the comprehensive likelihood function, the posterior distribution of the parameters to be updated is obtained through Bayesian updating. The expected values ​​of the parameters are extracted from the posterior distribution to update the corresponding parameters in the foundation pit BIM model. The data storage module stores the predicted deformation value, measured deformation value and parameter posterior distribution of each excavation step into the foundation pit BIM model; Processor, performs calculation and control of each module.

Images