Clay parameter back analysis method based on IFC attribute closed loop and intelligent dual-drive model

Abstract

The invention discloses a clay parameter back analysis method based on an IFC attribute closed loop and an intelligent dual-drive model. The clay parameter back analysis method comprises the steps that a foundation pit three-dimensional BIM model is built, and an IFC attribute set is generated; carrying out orthogonal processing on the soil body core parameters by adopting an orthogonal test design method to obtain a soil body core parameter set, and carrying out finite element simulation analysis through rock-soil finite element calculation software to obtain an analysis result; constructing an RF model, and performing optimization processing through a grey wolf cuckoo optimization algorithm to obtain an optimal RF training model; carrying out classification processing by adopting the optimal RF training model to obtain back analysis soil body core parameters of the target foundation pit; and the back analysis soil body core parameters are updated to soil body core parameters in the IFC attribute set, and the foundation pit three-dimensional BIM model is updated through the IFC attribute set. According to the clay parameter back analysis method based on the IFC attribute closed loop and the intelligent dual-drive model, the actual soil body core parameters of the target foundation pit can be accurately obtained, and then a reliable data basis is provided for safety evaluation and construction optimization of foundation pit engineering.

Description

Technical Field

[0001] This invention relates to the field of intelligent analysis and BIM parametric management technology in geotechnical engineering, and in particular to a method for inverse analysis of clay parameters based on an IFC attribute closed-loop and intelligent dual-drive model. Background Technology

[0002] In foundation and underground engineering design, mastering the engineering parameters of clay is crucial to ensuring the accuracy of calculations, predictions, and safety assessments. However, traditional laboratory verification and empirical formula-based methods are often costly, time-consuming, and fail to reflect site-scale effects and soil heterogeneity. Pure numerical inversion methods are also frequently affected by initial assumptions and boundary condition sensitivities, leading to uncertainties and non-uniqueness in the inversion results.

[0003] With the widespread application of BIM (Building Information Modeling) technology in engineering, although geometric information has been digitized, there are still gaps in the integration of engineering design and monitoring data with the BIM model. Traditional parametric back-analysis processes are often independent of the BIM model: numerical simulation results and parametric inversion values ​​usually exist in isolated documents or tables, which cannot be efficiently and standardizedly updated back into the BIM model. This results in a break in the information flow between design parameters and construction monitoring data, and parametric design and intelligent construction lack a unified data foundation.

[0004] While several parameter identification schemes based on machine learning or optimization algorithms have emerged in recent years, relying solely on data-driven approaches is prone to overfitting and insufficient generalization when physical constraints are lacking. Currently, no systematic solution has been developed that balances physical constraints, adaptive data learning, and the BIM / IFC data standard closed loop in engineering applications. Summary of the Invention

[0005] This invention addresses the problems existing in the prior art by proposing a clay parameter inverse analysis method based on an IFC attribute closed-loop and intelligent dual-drive model. The technical means employed in this invention are as follows: A method for inverse analysis of clay parameters based on IFC property closed-loop and intelligent dual-drive model includes the following steps: Step 1: Based on the construction drawings of the target foundation pit project, construct a three-dimensional BIM model of the foundation pit using building information modeling software, and store and export the IFC attribute set generated from the three-dimensional BIM model of the foundation pit. The IFC attribute set includes the prior range of the core soil parameters. Step 2: Map the IFC attribute set generated from the 3D BIM model of the foundation pit to the geotechnical finite element calculation software to form a foundation pit model dataset for the geotechnical finite element calculation software. Import the foundation pit model dataset into the geotechnical finite element calculation software to form a foundation pit finite element analysis model. Use the orthogonal experimental design method to orthogonally process multiple different soil core parameters to obtain multiple sets of soil core parameter sets. Use the geotechnical finite element calculation software to perform finite element simulation analysis on the foundation pit finite element analysis model in sequence using each set of soil core parameter sets to obtain multiple sets of finite element simulation analysis results for the foundation pit finite element analysis model. Step 3: Obtain the on-site measured monitoring data of the target foundation pit, construct the RF model, train the RF model based on multiple sets of soil core parameter sets and corresponding finite element simulation analysis results to obtain the initial RF training model, and optimize the initial RF training model based on the on-site measured monitoring data of the target foundation pit and the Grey Wolf Cuckoo optimization algorithm to obtain the optimal RF training model. Step 4: Based on the on-site measured monitoring data of the target foundation pit, the optimal RF training model is used for classification processing to obtain the core parameters of the back-analysis soil of the target foundation pit; Step 5: Based on the obtained back-analysis soil core parameters of the target foundation pit, update the soil core parameters in the IFC attribute set, and update the 3D BIM model of the foundation pit constructed by the building information modeling software through the updated IFC attribute set.

[0006] Furthermore, it also includes the following steps: Step 6: Generate a dynamic IFC attribute set based on the updated 3D BIM model of the foundation pit; Step 7: The dynamic foundation pit model dataset is formed by mapping the dynamic IFC attribute set to the geotechnical finite element calculation software data. Step 8: Import the dynamic foundation pit model dataset into the geotechnical finite element calculation software to generate a dynamic foundation pit finite element analysis model. Perform finite element simulation analysis on the dynamic foundation pit finite element analysis model using the geotechnical finite element calculation software to obtain the finite element simulation analysis results of the dynamic foundation pit finite element analysis model. Step 10: Compare the finite element simulation analysis results of the dynamic foundation pit finite element analysis model with the on-site measured monitoring data to determine whether the error between the two is less than the set threshold. If so, collect the on-site measured monitoring data again at the set time interval and compare them again. If not, use the optimal RF training model to classify and process the on-site measured monitoring data of the current target foundation pit to obtain the core parameters of the back analysis soil of the current target foundation pit, and return to step 5 to update the foundation pit 3D BIM model for the next time.

[0007] Furthermore, step 3 includes the following steps: Step 30: Construct the RF model and perform initialization settings. Set the input of the RF training model to the soil core parameter set and the finite element simulation analysis results of the foundation pit finite element analysis model. The decision variable to be optimized in the RF training model is the core parameter of the RF training model: the number of trees. and feature sampling dimensions for each tree ; Step 31: Train the RF model based on multiple sets of soil core parameters and corresponding finite element simulation analysis results to obtain an initial RF training model, and obtain the initial population of core parameters of the RF training model based on the initial RF training model. ; Step 32: Construct a cuckoo training model and a gray wolf training model, setting a maximum number of iterations for both models. Similarly, the initial population was trained using both the cuckoo training model and the gray wolf training model. Training is conducted to calculate the bird's nest fitness and obtain the current optimal bird's nest location. And calculate the gray wolf fitness to obtain the current optimal position of the three-headed wolf. ; Step 33: Determine if the maximum number of iterations has been reached or the convergence condition has been met. If so, output the current optimal Bird's Nest position. If the optimal solution is not found, proceed to the next step. Step 34: Update the current optimal bird's nest location. And the current optimal position of the three-headed wolf ; Step 35: Adopt the current optimal position of the three-headed wolf. For the current optimal Bird's Nest location Update to obtain the optimal Bird's Nest. ; Step 36: Generate random numbers using the cuckoo training model and determine if the random numbers are greater than the set probability of the nest being discovered. If not, keep the current nest position unchanged; if yes, update the nest position. Step 37: Calculate the fitness values ​​of all nests after the update to obtain the optimal nest location after the update. with optimal fitness Then proceed to the next iteration.

[0008] Furthermore, the fitness function is: (1) in: To calculate the fitness value using the mean squared error; Represents the number of data sets; The on-site measured monitoring data representing the target foundation pit; These are the predicted values ​​obtained from the current RF training model.

[0009] Furthermore, the optimal position of the three-headed wolf is adopted. For the current optimal Bird's Nest location Update to obtain the optimal Bird's Nest. Specifically as follows: (2) This formula represents the updated optimal bird's nest. The current best Bird's Nest The current optimal position of the three-headed wolf Union of the two sets; updated optimal bird's nest The calculation is as follows: (3) In the formula, This represents the current optimal three-headed wolf in the Grey Wolf Algorithm. Average position of the guide; A uniformly distributed random factor is used to adjust the perturbation intensity of Levi's flight. These are random numbers that follow a Lévy distribution. The solution is the current location of the bird's nest; The current global optimal solution; The best Bird's Nest after the update The optimal solution.

[0010] Furthermore, the convergence condition formula in step 33 is as follows: (4) in, This represents the current error evaluation index function value in the optimal RF training model; This represents the previous error evaluation index function value in the optimal RF training model; The set convergence threshold; The error evaluation index function in the optimal RF training model is: (5) in, The data is the prediction data based on the optimal RF training model; The monitoring data measured on-site for the target foundation pit; The set physical-driven weight values ​​and data-driven weight values; The empirical risk term is the prediction error of the optimal RF training model.

[0011] Furthermore, the core parameters of the soil include cohesion c, internal friction angle φ, elastic modulus E, and Poisson's ratio ν; the monitoring data / finite element analysis results include stress and displacement. Compared with existing technologies, the clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model disclosed in this invention has the following beneficial effects: The clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model disclosed in this invention constructs a target foundation pit BIM model and generates an IFC attribute set, then generates a foundation pit finite element analysis model through the IFC attribute set, performs simulation analysis on the finite element analysis model through finite element analysis software, and obtains the optimal RF training model by training the RF model based on the simulation analysis results and actual monitoring results. At the same time, based on the optimal RF training model and actual monitoring results, the actual core parameters of the soil in the target foundation pit can be accurately obtained, thus providing a reliable data foundation for the safety assessment and construction optimization of foundation pit engineering. Attached Figure Description

[0012] Figure 1 This is a flowchart of the clay parameter back analysis method based on IFC attribute closed loop and intelligent dual-drive model disclosed in this invention. Figure 2 The flowchart shows the clay parameter inverse analysis method based on IFC attribute closed loop and intelligent dual-drive model disclosed in this invention. Figure 3 This is a flowchart of the gray wolf-cuckoo optimization algorithm disclosed in this invention. Detailed Implementation

[0013] like Figure 1 and Figure 2 As shown, this invention discloses a method for inverse analysis of clay parameters based on an IFC property closed-loop and intelligent dual-drive model, comprising the following steps: Step 1: Based on the construction drawings of the target foundation pit project, construct a three-dimensional BIM model of the foundation pit using building information modeling software, and store and export the IFC attribute set generated from the three-dimensional BIM model of the foundation pit. The IFC attribute set includes the prior range of the core soil parameters. Specifically, in this embodiment, the building information modeling software used is Revit. In this step, based on the construction drawings of the target foundation pit project, a three-dimensional BIM model of the target foundation pit is constructed using Revit software. The Revit software then generates an IFC attribute set from the three-dimensional BIM model of the foundation pit, and stores and exports the IFC attribute set. The IFC attribute set includes the prior range of the soil core parameters that need to be obtained. In this embodiment, the soil core parameters include cohesion c, internal friction angle φ, elastic modulus E, and Poisson's ratio ν. Step 2: Map the IFC attribute set generated from the 3D BIM model of the foundation pit to the geotechnical finite element calculation software to form a foundation pit model dataset for the geotechnical finite element calculation software. Import the foundation pit model dataset into the geotechnical finite element calculation software to form a foundation pit finite element analysis model. Use the orthogonal experimental design method to orthogonally process multiple different soil core parameters to obtain multiple sets of soil core parameter sets. Use the geotechnical finite element calculation software to perform finite element simulation analysis on the foundation pit finite element analysis model in sequence using each set of soil core parameter sets to obtain multiple sets of finite element simulation analysis results for the foundation pit finite element analysis model. Specifically, the IFC attribute set generated from the 3D BIM model of the foundation pit can be directly mapped to the data of the geotechnical finite element analysis software to form a foundation pit model dataset for the geotechnical finite element analysis software. In this embodiment, the geotechnical finite element analysis software used is PLAXIS3D. By importing the foundation pit model dataset from the geotechnical finite element analysis software into PLAXIS3D software, a foundation pit finite element analysis model can be obtained. At the same time, orthogonal experimental design method is used to orthogonally process multiple different soil core parameters to obtain multiple sets of soil core parameter sets. Then, the foundation pit finite element analysis model is set with different soil core parameter sets in sequence using PLAXIS3D software and finite element simulation analysis is performed to obtain multiple sets of finite element analysis results for the foundation pit finite element analysis model. Step 3: Obtain the on-site measured monitoring data of the target foundation pit, construct the RF model, train the RF model based on multiple sets of soil core parameter sets and corresponding finite element simulation analysis results to obtain the initial RF training model, and optimize the initial RF training model based on the on-site measured monitoring data of the target foundation pit and the Grey Wolf Cuckoo optimization algorithm to obtain the optimal RF training model. Specifically, based on the principles of engineering importance, sensitivity, and significant mechanical response, key locations are selected as virtual "monitoring points" in the numerical model (finite element analysis model), and the finite element analysis results of the corresponding "monitoring points" are obtained. Simultaneously, on-site measured monitoring data of the corresponding "monitoring points" at the target foundation pit are obtained through various sensors at the site. Then, an RF (random forest) model is constructed. After initializing the RF model, the initial RF training model is trained using multiple sets of soil core parameters and the corresponding finite element analysis results from step 2. Then, based on the on-site measured monitoring data of the target foundation pit collected by sensors, the initial RF training model is further optimized using the Grey Wolf Cuckoo optimization algorithm to obtain the optimal RF training model. In this embodiment, the on-site measured monitoring data / finite element analysis results include stress and displacement.

[0014] Step 4: Based on the on-site measured monitoring data of the target foundation pit, the optimal RF training model is used for classification processing to obtain the core parameters of the back-analysis soil of the target foundation pit; Specifically, after obtaining the optimal RF training model, the on-site measured monitoring data of the target foundation pit collected by the sensor is input into the optimal RF training model for training and classification, so as to obtain the core parameters of the soil in the reverse analysis of the target foundation pit, that is, the real core parameters of the soil of the target foundation pit can be analyzed through the optimal RF training model. Step 5: Based on the obtained back analysis soil core parameters of the target foundation pit, update the soil core parameters in the IFC attribute set, and update the foundation pit 3D BIM model constructed by the building information modeling software through the updated IFC attribute set. Specifically, after obtaining the true soil core parameters of the target foundation pit through the optimal RF training model, the true soil core parameters of the target foundation pit are written back into the IFC attribute set (updating the IFC attribute set). Then, the 3D BIM model of the foundation pit constructed by the building information modeling software is updated using the updated IFC attribute set (at this time, the soil core parameters in the IFC attribute set are the true soil core parameters of the target foundation pit obtained through the optimal RF training model).

[0015] This invention discloses a clay parameter back analysis method based on IFC attribute closed-loop and intelligent dual-drive model. This method constructs a target foundation pit BIM model and generates an IFC attribute set. Then, it generates a finite element analysis (FEM) model of the foundation pit using the IFC attribute set. The FEM model is then simulated using FEM software. Based on the simulation results and actual monitoring results, an optimal RF training model is obtained through training. Simultaneously, based on the optimal RF training model and actual monitoring results, the actual core soil parameters of the target foundation pit can be accurately obtained, thus providing a reliable data foundation for the safety assessment and construction optimization of foundation pit engineering. Furthermore, this application first acquires initial training data using FEM software, and then trains the data using the RF training model. This effectively reduces the time required for obtaining core soil parameters through finite element simulation analysis (which is time-consuming), while ensuring the diversity and speed of data when using the RF model for training, thereby improving the accuracy of the final obtained core soil parameters.

[0016] Furthermore, it also includes the following steps: Step 6: Generate a dynamic IFC attribute set based on the updated 3D BIM model of the foundation pit; Specifically, after updating the 3D BIM model of the foundation pit constructed by the Building Information Modeling (BIM) software in step 5, the updated 3D BIM model of the foundation pit is then used again by the BIM software to generate an IFC attribute set, i.e., a dynamic IFC attribute set. The IFC attribute set includes various information about the target foundation pit, such as its geometric information (IfcProduct, IfcSolidModel), physical properties (IfcMaterial, IfcPropertySet), boundary constraints (IfcBoundaryCondition), and construction step information (IfcTask). A BIM physical information set can be generated by parsing the IFC file. (1) in, Represents the geometric information of the component. Indicates the material and its mechanical properties. This indicates the mapping information of monitoring points. Indicates boundary and load conditions.

[0017] Step 7: The dynamic foundation pit model dataset is formed by mapping the dynamic IFC attribute set to the geotechnical finite element calculation software data. Specifically, mapping the dynamic IFC attribute set to the geotechnical finite element calculation software data can directly form a dynamic foundation pit model dataset for geotechnical finite element calculation software. Step 8: Import the dynamic foundation pit model dataset into the geotechnical finite element calculation software to generate a dynamic foundation pit finite element analysis model. Perform finite element simulation analysis on the dynamic foundation pit finite element analysis model using the geotechnical finite element calculation software to obtain the finite element simulation analysis results of the dynamic foundation pit finite element analysis model. Specifically, the dynamic foundation pit model dataset is imported into the geotechnical finite element calculation software to generate a dynamic foundation pit finite element analysis model, and then finite element simulation analysis is performed to obtain the finite element simulation analysis results of the dynamic foundation pit finite element analysis model. Step 10: Compare the finite element simulation analysis results of the dynamic foundation pit finite element analysis model with the on-site measured monitoring data to determine whether the error between the two is less than the set threshold. If so, collect the on-site measured monitoring data again at the set time interval and compare them again. If not, use the optimal RF training model to classify and process the on-site measured monitoring data of the current target foundation pit to obtain the core parameters of the back analysis soil of the current target foundation pit, and return to step 5 to update the foundation pit 3D BIM model for the next time. Specifically, the finite element simulation analysis results of the dynamic foundation pit finite element analysis model are compared with the on-site measured monitoring data to determine whether the error between the two is less than a set threshold. If so, the on-site measured monitoring data is collected again at the set time interval and compared again. If not, the optimal RF training model is used to classify and process the on-site measured monitoring data of the current target foundation pit to obtain the core parameters of the back analysis soil of the current target foundation pit, and then the process returns to step 5 to update the foundation pit 3D BIM model for the next time.

[0018] In the actual construction of foundation pits, as time progresses and the construction of the target foundation pit advances, different soil layers at different locations within the target foundation pit correspond to different core soil parameters. This application, through continuous simulation, FR training analysis, and actual measurement of foundation pit data, ensures the synchronous evolution of the BIM visualization model and the FEM analysis model throughout the entire construction process. This ensures that the model prediction results are dynamically corrected over time, constructing a digital twin foundation pit system with self-learning and self-correction characteristics. This invention constructs a dynamic closed loop of "monitoring-simulation-correction" through steps 6 to 10, using the IFC standard to break down the data silos between BIM and finite element analysis, realizing the automated synchronous evolution of the foundation pit digital twin model. This process can capture the spatiotemporal effects of soil parameters in real time, and through error feedback, endow the system with self-diagnosis and self-correction capabilities, ensuring that model predictions are dynamically corrected with the construction progress, thereby significantly improving the robustness of the back analysis results and the foresight of engineering safety early warning.

[0019] Furthermore, such as Figure 3 As shown, step 3 includes the following steps: Step 30: Construct the RF model and perform initialization settings. Set the input of the RF training model to the soil core parameter set and the finite element simulation analysis results of the foundation pit finite element analysis model. The decision variable to be optimized in the RF training model is the core parameter of the RF training model: the number of trees. and feature sampling dimensions for each tree ; Specifically, when constructing the RF model, the RF model is initialized according to actual needs, and the input of the RF training model is set as the core parameter set of the soil and the finite element simulation analysis results of the foundation pit finite element analysis model. The decision variables to be optimized in the RF training model are the core parameters of the RF training model: the number of trees. and feature sampling dimensions for each tree That is, the data and results of the finite element analysis of the foundation pit are used as the initial population for training, and the core parameters of the RF training model are used as the optimization results of the training. Step 31: Train the RF model based on multiple sets of soil core parameters and corresponding finite element simulation analysis results to obtain an initial RF training model, and obtain the initial population of core parameters of the RF training model based on the initial RF training model. ; Specifically, in this step, the RF model is initially trained using data from the finite element analysis to obtain an initial RF training model. Then, the initial population of the core parameters of the RF training model can be obtained from the initial RF training model. ; Step 32: Construct a cuckoo training model and a gray wolf training model, setting a maximum number of iterations for both models. Similarly, the initial population was trained using both the cuckoo training model and the gray wolf training model. Training is conducted to calculate the bird's nest fitness and obtain the current optimal bird's nest location. And calculate the gray wolf fitness to obtain the current optimal position of the three-headed wolf. ; Specifically, in this step, a cuckoo training model and a gray wolf training model are constructed, and routine initialization settings are performed on each, such as controlling parameters, defining population size, maximum number of iterations, number of decision variables, and upper and lower bound vectors of the solution space; at the same time, the maximum number of iterations for both models is set. Once the initialization settings are identical, the initial population containing the core parameters of the RF training model will be obtained from the initial RF training model. The data is fed into two separate models for training, and the nest fitness is calculated to obtain the current optimal nest location. And calculate the gray wolf fitness to obtain the current optimal position of the three-headed wolf. In this embodiment, the fitness function is: (2) in: To calculate the fitness value using the mean squared error; Represents the number of data sets; The on-site measured monitoring data representing the target foundation pit; These are the predicted values ​​obtained from the current RF training model.

[0020] Step 33: Determine if the maximum number of iterations has been reached or the convergence condition has been met. If so, output the current optimal Bird's Nest position. If the optimal solution is not found, proceed to the next step. Specifically, the output is the current optimal location of the Bird's Nest. The corresponding solution is the optimal solution, which is the core parameter of the RF training model that needs to be sought: the number of tree trees. and feature sampling dimensions for each tree The optimal solution; in this embodiment, the convergence condition formula is as follows: (3) in, This represents the current error evaluation index function value in the optimal RF training model; This represents the previous error evaluation index function value in the optimal RF training model; The set convergence threshold; The error evaluation index function in the optimal RF training model is: (4) in, The data is the prediction data based on the optimal RF training model; The monitoring data measured on-site for the target foundation pit; The set physical-driven weight values ​​and data-driven weight values; The empirical risk term is the prediction error of the optimal RF training model.

[0021] Step 34: Update the current optimal bird's nest location. And the current optimal position of the three-headed wolf ; Step 35: Adopt the current optimal position of the three-headed wolf. For the current optimal Bird's Nest location Update to obtain the optimal Bird's Nest. ; Specifically, the adoption of the current optimal position of the three-headed wolf For the current optimal Bird's Nest location Update to obtain the optimal Bird's Nest. Specifically as follows: (5) This formula represents the updated optimal bird's nest. The current best Bird's Nest The current optimal position of the three-headed wolf Union of the two sets; updated optimal bird's nest The calculation is as follows: (6) In the formula, This represents the current optimal three-headed wolf in the Grey Wolf Algorithm. Average position of the guide; A uniformly distributed random factor is used to adjust the perturbation intensity of Levi's flight. These are random numbers that follow a Lévy distribution. The solution is the current location of the bird's nest; The current global optimal solution; The best Bird's Nest after the update The optimal solution.

[0022] Step 36: Generate random numbers using the cuckoo training model and determine if the random numbers are greater than the set probability of the nest being discovered. If not, keep the current nest position unchanged; if yes, update the nest position. Step 37: Calculate the fitness values ​​of all nests after the update to obtain the optimal nest location after the update. with optimal fitness Then proceed to the next iteration.

[0023] Through the above process, the Grey Wolf algorithm improves the local convergence ability of the Cuckoo Search, while the Cuckoo Search enhances the global exploration ability. The two algorithms synergistically optimize the random forest parameters, achieving coordinated control of global and local search capabilities and preventing the algorithm from getting trapped in local optima. This enables the random forest model to possess higher accuracy and stability in complex clay back-analysis problems. Furthermore, by combining error feedback from field monitoring data, a complete intelligent inversion closed loop from parameter optimization to model prediction to back-analysis correction is achieved.

[0024] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for inverse analysis of clay parameters based on an IFC attribute closed-loop and intelligent dual-drive model, characterized in that, Includes the following steps: Step 1: Based on the construction drawings of the target foundation pit project, construct a three-dimensional BIM model of the foundation pit using building information modeling software, and store and export the IFC attribute set generated from the three-dimensional BIM model of the foundation pit. The IFC attribute set includes the prior range of the core soil parameters. Step 2: Map the IFC attribute set generated from the 3D BIM model of the foundation pit to the geotechnical finite element calculation software to form a foundation pit model dataset for the geotechnical finite element calculation software. Import the foundation pit model dataset into the geotechnical finite element calculation software to form a foundation pit finite element analysis model. Use the orthogonal experimental design method to orthogonally process multiple different soil core parameters to obtain multiple sets of soil core parameter sets. Use the geotechnical finite element calculation software to perform finite element simulation analysis on the foundation pit finite element analysis model in sequence using each set of soil core parameter sets to obtain multiple sets of finite element simulation analysis results for the foundation pit finite element analysis model. Step 3: Obtain the on-site measured monitoring data of the target foundation pit, construct the RF model, train the RF model based on multiple sets of soil core parameter sets and corresponding finite element simulation analysis results to obtain the initial RF training model, and optimize the initial RF training model based on the on-site measured monitoring data of the target foundation pit and the Grey Wolf Cuckoo optimization algorithm to obtain the optimal RF training model. Step 4: Based on the on-site measured monitoring data of the target foundation pit, the optimal RF training model is used for classification processing to obtain the core parameters of the back-analysis soil of the target foundation pit; Step 5: Based on the obtained back-analysis soil core parameters of the target foundation pit, update the soil core parameters in the IFC attribute set, and update the 3D BIM model of the foundation pit constructed by the building information modeling software through the updated IFC attribute set.

2. The clay parameter back analysis method based on IFC attribute closed-loop and intelligent dual-drive model according to claim 1, characterized in that, It also includes the following steps: Step 6: Based on the updated 3D BIM model of the foundation pit, generate a dynamic IFC attribute set; Step 7: The dynamic foundation pit model dataset is formed by mapping the dynamic IFC attribute set to the geotechnical finite element calculation software data. Step 8: Import the dynamic foundation pit model dataset into the geotechnical finite element calculation software to generate a dynamic foundation pit finite element analysis model. Perform finite element simulation analysis on the dynamic foundation pit finite element analysis model using the geotechnical finite element calculation software to obtain the finite element simulation analysis results of the dynamic foundation pit finite element analysis model. Step 10: Compare the finite element simulation analysis results of the dynamic foundation pit finite element analysis model with the on-site measured monitoring data to determine whether the error between the two is less than the set threshold. If so, collect the on-site measured monitoring data again at the set time interval and compare them again. If not, use the optimal RF training model to classify and process the on-site measured monitoring data of the current target foundation pit to obtain the core parameters of the back analysis soil of the current target foundation pit, and return to step 5 to update the foundation pit 3D BIM model for the next time.

3. The clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model according to claim 1, characterized in that: Step 3 includes the following steps: Step 30: Construct the RF model and perform initialization settings. Set the input of the RF training model to the soil core parameter set and the finite element simulation analysis results of the foundation pit finite element analysis model. The decision variable to be optimized in the RF training model is the core parameter of the RF training model: the number of trees. and feature sampling dimensions for each tree ; Step 31: Train the RF model based on multiple sets of soil core parameters and corresponding finite element simulation analysis results to obtain an initial RF training model, and obtain the initial population of core parameters of the RF training model based on the initial RF training model. ; Step 32: Construct a cuckoo training model and a gray wolf training model, setting a maximum number of iterations for both models. Similarly, the initial population was trained using both the cuckoo training model and the gray wolf training model. Training is conducted to calculate the bird's nest fitness and obtain the current optimal bird's nest location. And calculate the gray wolf fitness to obtain the current optimal position of the three-headed wolf. ; Step 33: Determine if the maximum number of iterations has been reached or the convergence condition has been met. If so, output the current optimal Bird's Nest position. If the optimal solution is not found, proceed to the next step. Step 34: Update the current optimal bird's nest location. And the current optimal position of the three-headed wolf ; Step 35: Adopt the current optimal position of the three-headed wolf. For the current optimal Bird's Nest location Update to obtain the optimal Bird's Nest. ; Step 36: Generate random numbers using the cuckoo training model and determine if the random numbers are greater than the set probability of the nest being discovered. If not, keep the current nest position unchanged; if yes, update the nest position. Step 37: Calculate the fitness values ​​of all nests after the update to obtain the optimal nest location after the update. with optimal fitness Then proceed to the next iteration.

4. The clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model according to claim 3, characterized in that: The fitness function is: （1） in: To calculate the fitness value using the mean squared error; Represents the number of data sets; The on-site measured monitoring data representing the target foundation pit; These are the predicted values ​​obtained from the current RF training model.

5. The clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model according to claim 4, characterized in that: The optimal position for the three-headed wolf is adopted. For the current optimal Bird's Nest location Update to obtain the optimal Bird's Nest. Specifically as follows: （2） This formula represents the updated optimal bird's nest. The best Bird's Nest at present The current optimal position of the three-headed wolf Union of the two sets; updated optimal bird's nest The calculation is as follows: （3） In the formula, This represents the current optimal three-headed wolf in the Grey Wolf Algorithm. Average position of the guide; A uniformly distributed random factor is used to adjust the perturbation intensity of Levi's flight. These are random numbers that follow a Lévy distribution; The solution is the current location of the bird's nest; The current global optimal solution; The best Bird's Nest after the update The optimal solution.

6. The clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model according to claim 5, characterized in that: The convergence condition formula in step 33 is as follows: （4） in, This represents the current error evaluation index function value in the optimal RF training model; This represents the previous error evaluation index function value in the optimal RF training model. The set convergence threshold; The error evaluation index function in the optimal RF training model is: （5） in, The data is the prediction data based on the optimal RF training model; The monitoring data measured on-site for the target foundation pit; The set physical-driven weight values ​​and data-driven weight values; The empirical risk term is the prediction error of the optimal RF training model.

7. The clay parameter inverse analysis method based on IFC attribute closed-loop and intelligent dual-drive model according to claim 1, characterized in that: The core parameters of the soil include cohesion c, internal friction angle φ, elastic modulus E, and Poisson's ratio ν; the monitoring data / finite element analysis results include stress and displacement.

Images