A method and device for predicting foundation pit excavation deformation based on finite element

Abstract

The present invention provides a method and device for predicting deformation of foundation pit excavation based on finite elements, which relates to the field of foundation pit construction technology. The method first builds a digital geological model based on geological survey drilling data, and constructs a finite element numerical simulation model of foundation pit excavation in combination with the foundation pit support design drawing of the design institute; combines its orthogonal design parameters to obtain multiple sets of simulation data and establish a twin database; constructs an initial prediction model and uses the twin database to train a target prediction model; obtains foundation pit monitoring data through sensors, extracts data corresponding to the twin database and combines them into an effective parameter combination, which is input into the target model to predict the deformation result of the foundation pit excavation in the next stage; uses the deformation result to determine whether the excavation has reached the base, and if not, continues to predict until it reaches the base. The present invention combines geological and design data modeling, uses a twin database to train the model, can accurately predict deformation, and also combines monitoring data to achieve dynamic monitoring and real-time prediction to ensure construction safety.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of foundation pit construction, and in particular to a method and device for predicting foundation pit excavation deformation based on finite element. Background Art

[0002] With the acceleration of modernization, the utilization of urban aboveground space is nearing saturation. To address the shortage of urban land resources, the rational and efficient use of underground space has become an effective solution. The construction of underground projects, such as high-rise building basements, underground shopping malls, underground parking lots, large-scale drainage and sewage treatment systems, underground substations, and large subway stations, all rely on the fundamental project of foundation pit excavation. As the number of foundation pits increases, the potential construction risks are also increasing. As the foundation and key step of underground space development, foundation pit engineering involves numerous factors, and the construction process is unpredictable and risky.

[0003] Although some studies have attempted to combine machine learning with finite element methods in recent years, there are still problems such as a lack of model training data and a lack of real-time feedback mechanisms. For example, existing finite element analysis is mostly based on static geological models and fixed parameters, which makes it difficult to reflect the spatial variability of geological conditions. As a result, the prediction lags behind the actual excavation process, making it difficult to guide dynamic construction decisions. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and device for predicting foundation pit excavation deformation based on finite element method in view of the deficiencies in the prior art.

[0005] The present invention solves the above-mentioned technical problem with the following technical solution: a method for predicting foundation pit excavation deformation based on finite element method, comprising:

[0006] S1. Build a digital geological model based on pre-collected geological survey drilling data. Construct a finite element numerical simulation model for foundation pit excavation based on the foundation pit support design drawings of the design institute and in combination with the digital geological model.

[0007] S2. Perform orthogonal design parameter combination on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and establish a twin database based on the multiple sets of finite element numerical simulation data;

[0008] S3. Build an initial prediction model and train the initial prediction model using the established twin database to obtain a target prediction model.

[0009] S4. Obtaining monitoring data of the foundation pit in real time through pre-deployed sensors, extracting data corresponding to the twin database from the monitoring data and combining it with the data to obtain an effective parameter combination;

[0010] S5. Inputting the effective parameter combination as input parameters into the target prediction model to predict the deformation result of the next stage of foundation pit excavation;

[0011] S6. Determine whether the excavation has reached the base of the foundation pit based on the deformation result. If not, return to S4 to predict the next stage of foundation pit excavation until the base of the target building is reached.

[0012] Another technical solution of the present invention to solve the above technical problems is as follows: a foundation pit excavation deformation prediction device based on finite element method, comprising:

[0013] The finite element model construction module is used to establish a digital geological model based on the pre-collected geological survey drilling data. According to the foundation pit support design drawings of the design institute and combined with the digital geological model, a finite element numerical simulation model of the foundation pit excavation is constructed;

[0014] A twin database establishment module is used to perform orthogonal design parameter combination on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and to establish a twin database based on the multiple sets of finite element numerical simulation data;

[0015] The target prediction model construction module is used to build an initial prediction model. The initial prediction model is trained through the established twin database to obtain the target prediction model.

[0016] An effective parameter combination module is used to obtain the monitoring data of the foundation pit in real time through pre-deployed sensors, extract the data corresponding to the twin database from the monitoring data, and combine it with the data to obtain an effective parameter combination;

[0017] A deformation prediction module is used to input the effective parameter combination as an input parameter into the target prediction model to predict the deformation result of the next stage of foundation pit excavation;

[0018] It is also used to determine whether the excavation has reached the base of the foundation pit based on the deformation result. If not, the next stage of foundation pit excavation is predicted until the base of the target building is reached.

[0019] Another technical solution of the present invention to solve the above-mentioned technical problem is as follows: a foundation pit excavation deformation prediction device based on finite elements, comprising a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the foundation pit excavation deformation prediction method based on finite elements as described above is implemented.

[0020] Another technical solution of the present invention to solve the above technical problems is as follows: a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the finite element-based foundation pit excavation deformation prediction method as described above.

[0021] The beneficial effects of the present invention are: by combining geological survey data and foundation pit support design drawings to construct a finite element numerical simulation model, and using the twin database to train the prediction model, the deformation during foundation pit excavation can be predicted more accurately, providing a reliable reference basis for construction. By combining real-time monitoring data with the twin database to generate an effective parameter combination, the target prediction model is input for prediction, realizing dynamic monitoring and real-time prediction of the foundation pit excavation process, and being able to timely discover potential problems and take corresponding measures to ensure construction safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A schematic flow chart of a method for predicting foundation pit excavation deformation provided by an embodiment of the present invention;

[0023] Figure 2 A flowchart of the method for predicting deformation during foundation pit excavation provided by an embodiment of the present invention;

[0024] Figure 3 This is a functional module block diagram of the foundation pit excavation deformation prediction device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0026] like Figure 1 As shown, an embodiment of the present invention provides a method for predicting foundation pit excavation deformation based on finite element method, comprising:

[0027] S1. Build a digital geological model based on pre-collected geological survey drilling data. Construct a finite element numerical simulation model for foundation pit excavation based on the foundation pit support design drawings of the design institute and in combination with the digital geological model.

[0028] S2. Perform orthogonal design parameter combination on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and establish a twin database based on the multiple sets of finite element numerical simulation data;

[0029] S3. Build an initial prediction model and train the initial prediction model using the established twin database to obtain a target prediction model.

[0030] S4. Obtaining monitoring data of the foundation pit in real time through pre-deployed sensors, extracting data corresponding to the twin database from the monitoring data and combining it with the data to obtain an effective parameter combination;

[0031] S5. Inputting the effective parameter combination as input parameters into the target prediction model to predict the deformation result of the next stage of foundation pit excavation;

[0032] S6. Determine whether the excavation has reached the base of the foundation pit based on the deformation result. If not, return to S4 to predict the next stage of foundation pit excavation until the base of the target building is reached.

[0033] The following describes the entire process from geological survey to foundation pit excavation deformation prediction through a specific foundation pit project. Figure 2 As shown:

[0034] 1) Geological survey and digital geological model establishment:

[0035] Collect geological survey drilling data;

[0036] Build a digital geological model.

[0037] 2) Foundation pit support design:

[0038] Based on the foundation pit support design drawings of the design institute.

[0039] 3) Finite element numerical simulation model construction:

[0040] Construct a finite element numerical simulation model of foundation pit excavation.

[0041] 4) Preliminary calculation and twin database establishment:

[0042] Perform preliminary calculations on the finite element model and obtain preliminary calculation results.

[0043] Perform orthogonal design parameter combination.

[0044] Multiple sets of finite element numerical simulation data are calculated.

[0045] Establish a twin database.

[0046] 5) Real-time monitoring data storage system:

[0047] Collect support structure horizontal displacement data, soil stress data, water level height data and environmental parameter data.

[0048] Perform data preprocessing, data processing and analysis.

[0049] Extract the data combination corresponding to the twin database.

[0050] 6) Parameter sensitivity analysis and prediction model construction:

[0051] Sensitivity analysis of input parameters was performed.

[0052] Build the initial prediction model, determine the input layer and output layer, train the model, and establish the corresponding relationship. 7) Model training and parameter update:

[0053] Using the new parameter combination values, the parameters in the initial calculation model are updated.

[0054] The target prediction model is obtained through training.

[0055] 8) Deformation prediction and construction adjustment:

[0056] The effective parameter combination is input into the target prediction model as input parameters to predict the deformation result of the next stage of foundation pit excavation.

[0057] Determine whether to excavate to the bottom of the foundation pit based on the deformation results.

[0058] If the base is not reached, return to the prediction for the next stage of foundation pit excavation until the base of the target building is reached.

[0059] In the above embodiment, by combining geological survey data and foundation pit support design drawings to construct a finite element numerical simulation model, and using the twin database to train the prediction model, the deformation during the foundation pit excavation process can be predicted more accurately, providing a reliable reference basis for construction.

[0060] By combining real-time monitoring data with the twin database to generate effective parameter combinations, and inputting the target prediction model for prediction, dynamic monitoring and real-time prediction of the foundation pit excavation process are achieved, which can timely detect potential problems and take corresponding measures to ensure construction safety.

[0061] The prediction results can be used to determine whether the construction plan needs to be adjusted. If the predicted deformation exceeds the safety threshold, an early warning can be triggered and support structure reinforcement measures can be recommended, thereby optimizing the construction plan, reducing construction risks, and improving construction efficiency.

[0062] Preferably, in S1, establishing a digital geological model based on pre-collected geological survey drilling data includes:

[0063] Assume that the collected borehole sample data set is ,in, is the i-th borehole sample data value, and the Kriging interpolation formula is used to perform spatial interpolation on the borehole sample data value to establish a digital geological model:

[0064] ,

[0065] in, Point to be estimated The drill hole sample data value at is the weight coefficient, By the variation function Calculated and must satisfy ,

[0066] Based on the variation function Weight coefficient Solve, the variation function is:

[0067] ,

[0068] Where h is the distance between the drilling sample data points, is the number of pairs of drilling sample data points with a spacing of h, and The spatial position in the foundation pit and The drill hole sample data value at ;

[0069] Based on the digital geological model, a stratigraphic distribution map is generated by importing it into a mapping tool, and the stratigraphic distribution map is divided into j cells according to different stratigraphic layers i, and the material properties corresponding to the i-th soil layer are extracted from the pre-collected geological survey drilling data. The material properties include elastic modulus , Poisson's ratio , internal friction angle and cohesion And calculate the elastic modulus corresponding to j cells respectively by weighted average method , Poisson's ratio , internal friction angle and cohesion ,

[0070] For the elastic modulus corresponding to the jth unit cell for:

[0071] ,

[0072] For the Poisson's ratio corresponding to the jth cell for:

[0073] ,

[0074] For the internal friction angle corresponding to the jth unit cell for:

[0075] ,

[0076] For the cohesion corresponding to the jth cell for:

[0077] ,

[0078] Where n is the number of holes adjacent to the jth cell, is the weight coefficient corresponding to the elastic modulus, is the weight coefficient corresponding to Poisson’s ratio, is the weight coefficient corresponding to the internal friction angle, is the weight coefficient corresponding to cohesion.

[0079] Kriging interpolation is a geostatistical method that estimates geological property values at unknown locations based on known borehole sample data. It accounts for the spatial autocorrelation of the data—that is, closer points tend to have more similar geological properties. This allows for the generation of a continuous geological property distribution map that more accurately reflects the spatial variation of geological conditions.

[0080] Variogram The variogram is the core of kriging interpolation, describing the spatial correlation of geological attribute values over distance. Using the variogram, we can calculate weight coefficients, which determine the contribution of each borehole sample to the estimated point. The variogram calculation directly influences the allocation of weight coefficients. The weight coefficients determine the importance of each borehole sample in the interpolation process. Appropriate weight coefficient allocation improves interpolation accuracy and avoids estimation bias caused by uneven data distribution.

[0081] When calculating the material properties of each cell, a weighted averaging method is used to fully consider the weight of the drillhole data adjacent to the cell. The weight coefficient can be adjusted based on factors such as the distance between the drillhole and the cell and the reliability of the data, making the calculation results more representative. The weighted averaging method can avoid the errors that may be caused by simple averaging. For example, if a cell is close to multiple drillholes with significantly different data from these drillholes, simple averaging may result in large errors. However, the weighted averaging method can reasonably calculate the material property values of the cell based on the weight of each drillhole data, thereby improving the accuracy of the entire digital geological model.

[0082] In order to convert digital geological models into visual stratigraphic distribution maps, it is necessary to select appropriate mapping tools. Common mapping tools include GIS (Geographic Information System) tools, geological modeling tools, and general 3D modeling tools.

[0083] For example, use ArcGIS tools to generate a stratigraphic distribution map:

[0084] Data preparation: Export the data of the digital geological model into a CSV file, which contains the coordinates (X, Y, Z) and geological attribute values (such as elastic modulus E) of each point.

[0085] Data import: In ArcGIS, use the "Add Data" function to load the CSV file.

[0086] Meshing: Use the Create Fishnet tool to divide the excavation area into a mesh.

[0087] Attribute assignment: Use Spatial Interpolation tools such as Kriging to assign geological attribute values to each grid cell.

[0088] Visualization settings: Select the "Symbol System" option, set the color mapping, and represent different elastic modulus values with different colors.

[0089] Generate maps: The final stratigraphic distribution map can be exported to PDF or other formats for further analysis or presentation.

[0090] Through this process, the digital geological model is effectively converted into an intuitive stratigraphic distribution map, providing important visual support for foundation pit excavation deformation prediction and construction management.

[0091] In the above embodiment, the Kriging interpolation formula is used to perform spatial interpolation on the borehole sample data, which can more accurately establish a digital geological model, fully consider the spatial variability and complexity of geological conditions, and provide a more accurate geological basis for subsequent finite element numerical simulation.

[0092] The stratum distribution map is divided into multiple cells, and the material properties of each cell are calculated using the weighted average method, so that the finite element model can more carefully reflect the mechanical characteristics of different strata, improve the accuracy of the finite element numerical simulation, and thus improve the reliability of the prediction results.

[0093] Preferably, in S1, a finite element numerical simulation model of foundation pit excavation is constructed based on the foundation pit support design drawings of the design institute and in combination with the digital geological model, including:

[0094] Draw the foundation pit geometric model using a drawing tool based on the foundation pit support design drawing of the design institute, map the foundation pit geometric model to the stratum distribution map, and divide the foundation pit geometric model into finite element grid cells corresponding to j cells in the stratum distribution map, and assign the material property of each finite element grid cell to the elastic modulus of the corresponding cell. , Poisson's ratio , internal friction angle and cohesion ;

[0095] Establish the material stiffness matrix D corresponding to the jth finite element mesh element and define the plastic behavior of the soil, specifically:

[0096] For linear elastic materials under plane strain conditions, the material stiffness matrix D is expressed as:

[0097] ,

[0098] The plastic behavior of soil is defined by the Mohr-Coulomb yield function, which is:

[0099] ,

[0100] in, is the first stress invariant, is the second deviatoric stress invariant,

[0101] The finite element numerical simulation model of foundation pit excavation is obtained.

[0102] It should be understood that the foundation pit support design drawings provided by the design institute contain detailed information such as the shape, size, location and type of the support structure, etc. This information is the basis for constructing the finite element geometry model.

[0103] Use professional drafting software (such as AutoCAD or GeoStudio) to draw the excavation geometry according to the design drawings. This includes the excavation boundaries and the location and shape of supporting structures (such as retaining walls and anchors). Align and map the drawn excavation geometry with the stratigraphic distribution map in the digital geological model. Ensure that every part of the excavation geometry accurately corresponds to the corresponding stratigraphic cell. This step is crucial for subsequent meshing and material property assignment.

[0104] The meshing method is the finite element method, which requires dividing complex geometric models into small units (grids) for numerical calculations. The quality of the meshing directly affects the accuracy and efficiency of the finite element analysis.

[0105] Select an appropriate mesh type (e.g., quadrilateral, triangular, etc.) based on the excavation geometry and stratum distribution. For complex geometries, a hybrid mesh may be necessary. A denser mesh is recommended for improved calculation accuracy in critical areas (e.g., near support structures and areas with dramatic stratum fluctuations). In relatively flat areas, a sparser mesh can be used to reduce the computational effort.

[0106] Assign the material properties of each finite element mesh element to the property values of the corresponding cell. This can be accomplished through interpolation or direct assignment. For example, if a mesh element is completely within a formation cell, the property value of that formation is directly assigned. If the mesh element spans multiple formation cells, the property value must be calculated using methods such as weighted averaging.

[0107] In the above embodiment, the foundation pit geometric model is drawn according to the foundation pit support design drawings of the design institute and mapped to the stratigraphic distribution map, which can accurately simulate the actual geometric shape and positional relationship of the foundation pit and provide an accurate geometric basis for finite element numerical simulation.

[0108] Establishing the material stiffness matrix of the finite element mesh unit and using the Mohr-Coulomb yield function to define the plastic behavior of the soil can more reasonably describe the mechanical response of the soil during the excavation process, making the finite element numerical simulation model closer to the actual engineering situation and improving the accuracy of the prediction results.

[0109] Preferably, in S2, orthogonal design parameter combination is performed on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and a twin database is established through the multiple sets of finite element numerical simulation data, specifically:

[0110] Based on the finite element mesh elements in the foundation pit excavation finite element numerical simulation model, the input parameter set to be optimized is determined. , and divide each parameter into k horizontal constraint ranges based on geological survey data:

[0111] , , ,

[0112] in, 、 、 and Determined by geological survey data respectively;

[0113] Use orthogonal experimental design method to select orthogonal table , generate N groups of parameter combinations ,in, is the number of parameters, k is the number of levels, ;

[0114] For each parameter combination , extracting key node displacement data based on the finite element numerical simulation model of foundation pit excavation and stress data , forming the output data set ;

[0115] Combining parameters and output dataset Associative storage, building a twin database .

[0116] It should be understood that, based on geological survey data, a reasonable horizontal constraint range is defined for each parameter. For example, the elastic modulus may vary within a certain range, and the Poisson's ratio also has a reasonable value range.

[0117] Orthogonal experimental design is used to select appropriate orthogonal tables and generate multiple sets of parameter combinations. Orthogonal experimental design is an efficient experimental design method that can comprehensively examine the interaction of multiple factors with a relatively small number of experiments.

[0118] The twin database contains simulation results under various parameter combinations, providing rich training samples for machine learning models and helping to improve the generalization ability of the model.

[0119] In the above embodiment, by determining the input parameter set to be optimized and dividing the horizontal constraint range, and then using the orthogonal experimental design method to generate multiple sets of parameter combinations, a large amount of finite element numerical simulation data can be efficiently generated, and a twin database covering various working conditions can be constructed, which provides rich data support for subsequent model training and helps to improve the generalization ability and prediction accuracy of the prediction model.

[0120] By associating and storing parameter combinations and output data sets to build a twin database, effective data organization and management are achieved, facilitating subsequent data query, analysis, and model training, and improving the operating efficiency of the entire prediction system.

[0121] Preferably, in S3, an initial prediction model is constructed, and the initial prediction model is trained using the established twin database to obtain a target prediction model, including:

[0122] Parameter combination Data in and displacement data of key nodes The data in are normalized as follows:

[0123] ,

[0124] ,

[0125] in, and Parameter combinations The mean and standard deviation of and The displacement data of key nodes are The mean and standard deviation of

[0126] An initial prediction model is constructed based on a PSO-BP neural network, wherein the initial prediction model includes an input layer, a hidden layer, and an output layer, and the number of nodes in the input layer and the output layer is set, as well as the activation function and the number of layers and the number of nodes in each layer of the hidden layer are set, and training parameters are defined. The initial weight of the initial prediction model is optimized by a particle swarm PSO algorithm;

[0127] After normalization, the training data set The initial prediction model is trained, and the weight and bias parameters of the initial prediction model are updated through a back-propagation algorithm to obtain a target prediction model.

[0128] It should be understood that the particle swarm optimization (PSO) algorithm is an optimization algorithm based on swarm intelligence, which simulates the foraging behavior of bird flocks to find the optimal solution. In neural networks, the PSO algorithm can be used to optimize the network's initial weights, avoiding the problem of traditional BP neural networks easily falling into local optimality.

[0129] The initial prediction model consists of an input layer, hidden layers, and an output layer. The number of nodes in the input layer typically matches the number of input parameters, while the number of nodes in the output layer matches the number of prediction targets. The number of hidden layer nodes and layers should be adjusted based on the specific problem to achieve optimal prediction results.

[0130] For example: input layer: the number of nodes is 4, corresponding to the parameter combination; hidden layer: contains 2 layers, each node is 10, the activation function is ReLU, which is used to capture the nonlinear relationship between input and output; output layer: the number of nodes is 3, corresponding to the displacement data of the key nodes, the activation function is linear, and the output is the normalized displacement prediction value.

[0131] The training parameters are defined as follows: the loss function is the mean square error (MSE), which is used to measure the deviation between the predicted value and the true value; the optimizer is the Adam algorithm, and the learning rate is set to 0.001 to accelerate model convergence; the training rounds are 500 times and the batch size is 32 to ensure sufficient data learning.

[0132] Apply the particle swarm optimization (PSO) to optimize the initial weights of the neural network, including:

[0133] Particle swarm parameters: population size 5, maximum number of generations 50;

[0134] Optimization goal: minimize the fitness value of the initial weight of the neural network, and the fitness function is the MSE on the training set;

[0135] Weight update rule: adjust the particle position by weighting the global optimum and the individual optimum.

[0136] In the above-mentioned embodiment, normalizing the parameter combinations and key node displacement data can eliminate the impact of data dimension and magnitude differences on model training, making model training more stable and efficient. Constructing an initial prediction model based on a PSO-BP neural network and optimizing the initial weights using a particle swarm optimization (PSO) algorithm can effectively avoid the local optimality problem of traditional BP neural networks, improve the model's convergence speed and prediction accuracy, and thus obtain a more accurate target prediction model.

[0137] Preferably, in S4, monitoring data of the foundation pit is obtained in real time through pre-deployed sensors, and data corresponding to the twin database is extracted from the monitoring data and combined with the data to obtain an effective parameter combination, including:

[0138] The monitoring data of the foundation pit is obtained in real time through pre-deployed sensors. The monitoring data includes horizontal displacement data of the support structure obtained by displacement sensors, soil stress data obtained by earth pressure cells, water level data obtained by water level sensors, and environmental parameters obtained by temperature and humidity sensors.

[0139] Extract the characteristic vectors of the horizontal displacement data of the support structure, the soil stress data and the water level height data corresponding to the twin database from the monitoring data to obtain the real-time characteristic vector ,in, is the displacement characteristic, is the stress characteristic, is the water level characteristic;

[0140] Calculate real-time feature vectors based on Euclidean distance formula The feature vector corresponding to each parameter combination in the twin database Similarity:

[0141] ,

[0142] in, is the similarity distance between the lth group of parameter combinations and the real-time feature vector. The smaller the value, the more similar the characteristics of the real-time data and the lth group of parameters in the twin database are. is the actual measured value of the i-th feature in the real-time feature vector, is the simulated value of the ith feature corresponding to the lth parameter combination in the database, is the value range of the i-th dimension feature in the twin database;

[0143] Select the top K parameter combinations with the highest similarity ;

[0144] Perform weighted fusion on the selected parameter combinations to generate effective parameter combinations :

[0145] ,

[0146] ,

[0147] Among them, the weight Similarity Inversely proportional to each other, ensuring that high similarity parameter combinations contribute more to the fusion results. is the similarity distance of the kth group among the first K groups of parameter combinations after screening, ,

[0148] Combine the fused effective parameters according to the k level constraint ranges The values in that exceed the k level constraints are truncated to the boundary values.

[0149] In the above-mentioned embodiment, by using a variety of pre-deployed sensors to obtain real-time monitoring data from the foundation pit and extracting feature vectors corresponding to the twin database, it is possible to fully utilize the real-time on-site monitoring data, provide the prediction model with the latest engineering information, and enhance the timeliness and accuracy of the prediction results. Based on the Euclidean distance formula, the similarity between the real-time feature vector and the parameter combination in the twin database is calculated, and the top K groups of parameter combinations with the highest similarity are selected for weighted fusion. This can screen out the parameter combination closest to the current working conditions, generate an effective parameter combination that is more consistent with the actual situation, and further improve the reliability of the prediction results.

[0150] Preferably, in step S5, the effective parameter combination is input as an input parameter into the target prediction model to predict the deformation result of the next stage of foundation pit excavation, including:

[0151] For valid parameter combinations ( ) for normalization;

[0152] The normalized effective parameter combination Input the target prediction model and calculate the normalized displacement prediction value through forward propagation , the calculation formula is:

[0153] ,

[0154] Among them, the training data set outputs the mean and standard deviation of the displacement;

[0155] Denormalize the standardized predicted values to restore them to actual engineering units:

[0156] ,

[0157] The predicted displacement Match the key node coordinates of the finite element model, generate a visual deformation cloud map, and output the maximum displacement value And displacement change trend curve;

[0158] According to the maximum displacement Determine whether the construction plan needs to be adjusted. , If the safety threshold is exceeded, an early warning is triggered and reinforcement measures for the supporting structure are recommended.

[0159] In the above embodiment, the effective parameter combination is normalized and input into the target prediction model, and the standardized displacement prediction value is obtained through forward propagation calculation, and then denormalized to restore it to the actual engineering unit. This can accurately predict the deformation result of the next stage of foundation pit excavation and provide strong support for construction decision-making.

[0160] The system generates a visual deformation cloud map, outputs the maximum displacement value, and outputs a displacement trend curve. This intuitively displays the deformation of the foundation pit, allowing construction personnel to quickly understand the project status. Furthermore, the system determines whether the construction plan needs to be adjusted based on the maximum displacement value, triggers an early warning, and recommends reinforcement measures when necessary, effectively ensuring construction safety.

[0161] Preferably, in S6, judging whether to excavate to the base of the foundation pit according to the deformation result includes:

[0162] Calculate the spatial relative displacement deviation between the base design elevation and the deformation result;

[0163] Whether the convergence condition is met is determined based on the spatial relative displacement deviation. If so, it is determined that the excavation has reached the base.

[0164] For example, the convergence conditions include:

[0165] Convergence condition 1: ,in, is the absolute value of the spatial relative displacement deviation, is the engineering allowable error;

[0166] Convergence condition 2: displacement change rate of z consecutive excavation stages ,in, is the rate threshold. For example, z=3, such as 0.1 mm / day.

[0167] In the above embodiment, the spatial relative displacement deviation from the design elevation of the foundation is calculated based on the predicted deformation results, and the convergence condition is used to determine whether excavation has reached the foundation. This accurately determines whether the foundation pit excavation has reached the foundation of the target building, avoiding over-excavation or under-excavation and ensuring construction quality. Setting multiple convergence conditions, including the absolute value of the spatial relative displacement deviation and the rate of change of displacement during successive excavation stages, allows for a more comprehensive assessment of the convergence of the foundation pit excavation, improving the accuracy and reliability of the judgment and providing a scientific basis for controlling the construction progress.

[0168] like Figure 3 As shown, another embodiment of the present invention provides a foundation pit excavation deformation prediction device based on finite element method, comprising:

[0169] The finite element model construction module is used to establish a digital geological model based on the pre-collected geological survey drilling data. According to the foundation pit support design drawings of the design institute and combined with the digital geological model, a finite element numerical simulation model of the foundation pit excavation is constructed;

[0170] A twin database establishment module is used to perform orthogonal design parameter combination on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and to establish a twin database based on the multiple sets of finite element numerical simulation data;

[0171] The target prediction model construction module is used to build an initial prediction model. The initial prediction model is trained through the established twin database to obtain the target prediction model.

[0172] An effective parameter combination module is used to obtain the monitoring data of the foundation pit in real time through pre-deployed sensors, extract the data corresponding to the twin database from the monitoring data, and combine it with the data to obtain an effective parameter combination;

[0173] A deformation prediction module is used to input the effective parameter combination as an input parameter into the target prediction model to predict the deformation result of the next stage of foundation pit excavation;

[0174] It is also used to determine whether the excavation has reached the base of the foundation pit based on the deformation result. If not, the next stage of foundation pit excavation is predicted until the base of the target building is reached.

[0175] Another embodiment of the present invention provides a finite element-based foundation pit excavation deformation prediction device, including a memory, a processor, and a computer program stored in the memory and runnable on the processor. When the processor executes the computer program, the finite element-based foundation pit excavation deformation prediction method as described above is implemented.

[0176] Another embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the above-mentioned finite element-based foundation pit excavation deformation prediction method is implemented.

[0177] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0178] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is merely a logical functional division. In actual implementation, other division methods may be used. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not implemented.

[0179] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected based on actual needs to achieve the objectives of the embodiments of the present invention.

[0180] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for predicting foundation pit excavation deformation based on finite element method, characterized in that: include: S1. Build a digital geological model based on pre-collected geological survey drilling data. Construct a finite element numerical simulation model for foundation pit excavation based on the foundation pit support design drawings of the design institute and in combination with the digital geological model. S2. Perform orthogonal design parameter combination on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and establish a twin database based on the multiple sets of finite element numerical simulation data; S3. Build an initial prediction model and train the initial prediction model using the established twin database to obtain a target prediction model. S4. Obtaining monitoring data of the foundation pit in real time through pre-deployed sensors, extracting data corresponding to the twin database from the monitoring data and combining it with the data to obtain an effective parameter combination; S5. Inputting the effective parameter combination as input parameters into the target prediction model to predict the deformation result of the next stage of foundation pit excavation; S6. Determine whether the excavation has reached the base of the foundation pit based on the deformation result. If not, return to S4 to predict the next stage of foundation pit excavation until the base of the target building is reached. In S2, orthogonal design parameter combination is performed on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and a twin database is established through the multiple sets of finite element numerical simulation data, specifically: Based on the finite element mesh elements in the foundation pit excavation finite element numerical simulation model, the input parameter set to be optimized is determined. And based on the geological survey data, each parameter is divided into k horizontal constraint ranges: E j ∈[E min ,E max ],n j ∈[ν min ,n max ], c j ∈[c min ,c max ], Among them, E j ,ν j , c j Represents the elastic modulus E corresponding to the jth cell j , Poisson's ratio ν j 、Internal friction angle φ j and cohesion c j ;E min / E max 、v min / v max 、 and c min / c max Determined by geological survey data respectively; Adopt orthogonal experimental design method and select orthogonal table L N (k m ), generate N groups of parameter combinations Where m = 4 is the number of parameters, k is the number of levels, and l = 1, 2, ..., N; For each parameter combination C l , extract key node displacement data u based on the foundation pit excavation finite element numerical simulation model l =(u x,l ,u y,l ,u z,l ) and stress data σ l =(σ xx,l ,σ yy,l ,σ zz,l ), forming the output data set D l =(u l ,σ l ); Combine the parameters C l And the output dataset D l Associated storage, building a twin database DB = {(C l , D l )|l=1,2,…,N}.

2. The method for predicting foundation pit excavation deformation according to claim 1, characterized in that: In S1, a digital geological model is established based on pre-collected geological survey drilling data, including: Assume that the collected borehole sample data set is Z(x i )={x1,x2,...,x i }, where x i is the i-th borehole sample data value, and the Kriging interpolation formula is used to perform spatial interpolation on the borehole sample data value to establish a digital geological model: in, is the drilling sample data value at the point x0 to be estimated, λ i is the weight coefficient, λ i Calculated by the variation function γ(h), and must satisfy Based on the variation function γ(h) the weight coefficient λ i Solve, the variation function is: Where h is the distance between the borehole sample data points, N(h) is the number of pairs of borehole sample data points with a spacing of h, and Z(x i ) and Z(x i +h) is the spatial position x in the foundation pit i and x i The drill hole sample data value at +h; Based on the digital geological model, a stratigraphic distribution map is generated by importing it into a mapping tool. The stratigraphic distribution map is divided into j cells according to different stratigraphic layers i, and the material properties corresponding to the i-th soil layer are extracted from the pre-collected geological survey drilling data. The material properties include the elastic modulus E i , Poisson's ratio ν i 、Internal friction angle φ i and cohesion c i And calculate the elastic modulus E corresponding to j cells by weighted average method j , Poisson's ratio ν j 、Internal friction angle φ j and cohesion c j , For the elastic modulus E corresponding to the jth unit cell j for: For the Poisson's ratio ν corresponding to the jth cell j for: For the internal friction angle φ corresponding to the jth unit cell j for: For the cohesion c corresponding to the jth cell j for: Where n is the number of holes adjacent to the jth cell, w Eij is the weight coefficient corresponding to the elastic modulus, w νij is the weight coefficient corresponding to Poisson’s ratio, w φ ij is the weight coefficient corresponding to the internal friction angle, w cij is the weight coefficient corresponding to cohesion.

3. The foundation pit excavation deformation prediction method according to claim 2, characterized in that: In S1, a finite element numerical simulation model of foundation pit excavation is constructed based on the foundation pit support design drawings of the design institute and combined with the digital geological model, including: According to the foundation pit support design drawings of the design institute, a foundation pit geometric model is drawn using a drawing tool. The foundation pit geometric model is mapped to the stratum distribution map, and the foundation pit geometric model is divided into finite element grid cells corresponding to j cells in the stratum distribution map. The material property of each finite element grid cell is assigned to the elastic modulus E of the corresponding cell. j , Poisson's ratio ν j , internal friction angle and cohesion c j ; Establish the material stiffness matrix D corresponding to the jth finite element mesh element and define the plastic behavior of the soil, specifically: For linear elastic materials under plane strain conditions, the material stiffness matrix D is expressed as: The plastic behavior of soil is defined by the Mohr-Coulomb yield function, which is: Where I1 is the first stress invariant, J2 is the second deviatoric stress invariant, The finite element numerical simulation model of foundation pit excavation is obtained.

4. The method for predicting foundation pit excavation deformation according to claim 1, characterized in that: In S3, an initial prediction model is constructed and trained using the established twin database to obtain a target prediction model, including: For parameter combination C l The data in and the key node displacement data u l The data in are normalized as follows: Among them, μ C and σ C They are parameter combination C l The mean and standard deviation, μ D and σ D are the key node displacement data u l The mean and standard deviation of An initial prediction model is constructed based on a PSO-BP neural network, wherein the initial prediction model includes an input layer, a hidden layer, and an output layer, and the number of nodes in the input layer and the output layer is set, as well as the activation function and the number of layers and the number of nodes in each layer of the hidden layer are set, and training parameters are defined. The initial weight of the initial prediction model is optimized by a particle swarm PSO algorithm; The initial prediction model is trained using the normalized training data set, and the weights and bias parameters of the initial prediction model are updated using a back-propagation algorithm to obtain a target prediction model.

5. The method for predicting foundation pit excavation deformation according to claim 4, characterized in that: In S4, monitoring data of the foundation pit is obtained in real time through pre-deployed sensors, and data corresponding to the twin database is extracted from the monitoring data and combined with the data to obtain an effective parameter combination, including: The monitoring data of the foundation pit is obtained in real time through pre-deployed sensors. The monitoring data includes horizontal displacement data of the support structure obtained by displacement sensors, soil stress data obtained by earth pressure cells, water level data obtained by water level sensors, and environmental parameters obtained by temperature and humidity sensors. Extract the characteristic vectors of the horizontal displacement data of the supporting structure, the soil stress data and the water level height data corresponding to the twin database from the monitoring data to obtain the real-time characteristic vector F real =(u key ,σ n ,ΔH w / Δt), where u key is the displacement characteristic, σ n is the stress characteristic, ΔH w / Δt is the water level characteristic; Calculate the real-time feature vector F based on the Euclidean distance formula real The feature vector F corresponding to each parameter combination in the twin database l Similarity: Among them, d l is the similarity distance between the lth group of parameter combinations and the real-time feature vector, F real,i is the actual measurement value of the i-th feature in the real-time feature vector, F l,i is the simulated value of the ith feature corresponding to the lth parameter combination in the database, range(F i ) is the value range of the i-th dimension feature in the twin database; Select the top K parameter combinations C with the highest similarity topK ={C1,C2,...,C K }; Perform weighted fusion on the selected parameter combinations to generate effective parameter combination C eff : Among them, the weight w k Similarity Inversely proportional, is the similarity distance of the kth group among the first K groups of parameter combinations after screening, k=1,2,...,K, Combine the fused effective parameters C according to the k level constraint ranges eff The values in that exceed the k level constraints are truncated to the boundary values.

6. The method for predicting foundation pit excavation deformation according to claim 4, characterized in that: In S5, the effective parameter combination is input as an input parameter into the target prediction model to predict the deformation result of the next stage of foundation pit excavation, including: For valid parameter combination C eff Perform normalization processing; Combining the normalized effective parameters Input the target prediction model and calculate the normalized displacement prediction value through forward propagation The calculation formula is: Where W is the standard deviation of the output displacement of the training data set, and B is the mean of the output displacement of the training data set; Denormalize the standardized predicted values to restore them to actual engineering units: The predicted displacement u pred Match the key node coordinates of the finite element model, generate a visual deformation cloud map, and output the maximum displacement value u max And displacement change trend curve; According to the maximum displacement value u max Determine whether the construction plan needs to be adjusted. If u max ≥u threshold ,u threshold If the safety threshold is exceeded, an early warning is triggered and reinforcement measures for the supporting structure are recommended.

7. The method for predicting foundation pit excavation deformation according to claim 6, characterized in that: In S6, judging whether to excavate to the base of the foundation pit according to the deformation result includes: Calculate the spatial relative displacement deviation between the base design elevation and the deformation result; Whether the convergence condition is met is determined based on the spatial relative displacement deviation. If so, it is determined that the excavation has reached the base.

8. The method for predicting foundation pit excavation deformation according to claim 7, characterized in that: The convergence conditions include: Convergence condition 1: |Δh|≤δ tol , where |Δh| is the absolute value of the spatial relative displacement deviation, δ tol is the engineering allowable error; Convergence condition 2: displacement change rate of z consecutive excavation stages Among them, η tol is the rate threshold.

9. A finite element-based foundation pit excavation deformation prediction device, characterized in that: include: The finite element model construction module is used to establish a digital geological model based on the pre-collected geological survey drilling data. According to the design institute's foundation pit support design drawings and combined with the digital geological model, a finite element numerical simulation model of the foundation pit excavation is constructed; The twin database establishment module is used to perform orthogonal design parameter combination on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data, and establish a twin database based on the multiple sets of finite element numerical simulation data; The target prediction model construction module is used to build an initial prediction model. The initial prediction model is trained through the established twin database to obtain the target prediction model. An effective parameter combination module is used to obtain the monitoring data of the foundation pit in real time through pre-deployed sensors, extract the data corresponding to the twin database from the monitoring data, and combine it with the data to obtain an effective parameter combination; A deformation prediction module is used to input the effective parameter combination as an input parameter into the target prediction model to predict the deformation result of the next stage of foundation pit excavation; It is also used to determine whether the excavation has reached the base of the foundation pit based on the deformation result. If not, it is used to predict the next stage of foundation pit excavation until the base of the target building is reached. In the twin database establishment module, orthogonal design parameter combination is performed on the constructed foundation pit excavation finite element numerical simulation model to obtain multiple sets of finite element numerical simulation data. The twin database is established based on the multiple sets of finite element numerical simulation data, specifically: Based on the finite element mesh elements in the foundation pit excavation finite element numerical simulation model, the input parameter set to be optimized is determined. And based on the geological survey data, each parameter is divided into k horizontal constraint ranges: E j ∈[E min ,E max ],n j ∈[ν min ,n max ], c j ∈[c min ,c max ], Among them, E j ,ν j , c j Represents the elastic modulus E corresponding to the jth cell j , Poisson's ratio ν j 、Internal friction angle φ j and cohesion c j ;E min / E max 、v min / v max 、 and c min / c max Determined by geological survey data respectively; Adopt orthogonal experimental design method and select orthogonal table L N (k m ), generate N groups of parameter combinations Where m = 4 is the number of parameters, k is the number of levels, and l = 1, 2, ..., N; For each parameter combination C l , extract key node displacement data u based on the foundation pit excavation finite element numerical simulation model l =(u x,l ,u y,l ,u z,l ) and stress data σ l =(σ xx,l ,σ yy,l ,σ zz,l ), forming the output data set D l =(u l ,σ l ); Combine the parameters C l And the output dataset D l Associated storage, building a twin database DB = {(C l , D l )|l=1,2,…,N}.

Images