A three-dimensional modeling and intelligent optimization design method considering uneven characteristics of road surface
By using 3D modeling and intelligent optimization design methods, combined with BIM, finite element analysis and deep learning models, the problem of defects caused by the uneven characteristics of asphalt pavement was solved, achieving a more scientific and accurate design, extending the service life of the pavement and reducing costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHANGAN UNIV
- Filing Date
- 2026-03-04
- Publication Date
- 2026-06-09
AI Technical Summary
Existing asphalt pavement design methods fail to effectively consider uneven surface settlement and uneven modulus distribution, leading to rutting, fatigue damage, and other defects that affect the durability and reliability of the pavement structure. Furthermore, they lack the combination of data simulation and automated algorithms for multi-objective optimization.
A three-dimensional modeling and intelligent optimization design method is adopted. BIM model, finite element analysis software Hypermesh, MATLAB and ABAQUS are used for mesh generation and simulation. A prediction model is established by combining convolutional neural network and temporal convolutional network. The NSGA-II optimization algorithm is used to determine the optimal design scheme.
It improves the scientific rigor and precision of asphalt pavement structure design, extends service life, reduces design costs, and provides technical support for the intelligent transformation of road engineering.
Smart Images
Figure CN122174546A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of asphalt pavement simulation modeling technology, specifically a three-dimensional modeling and intelligent optimization design method that considers the uneven characteristics of pavement. Background Technology
[0002] The development of asphalt pavement stems from the innovation of design methods, fundamentally driven by ever-increasing demands. In the early stages of construction, asphalt pavement design methods combined classical mechanics calculations with extensive experimental conclusions, employing a mechanics-empirical approach.
[0003] However, the mechanics-empirical method has certain regional and limitations. Gradually, a combination of mechanics-empirical methods and numerical simulations with long-term performance data has emerged to optimize layer thickness and material parameters. However, current asphalt pavement optimization design rarely considers the changes in the stress state of the pavement structure caused by uneven surface settlement and modulus distribution, which can lead to rutting, fatigue damage, and other defects, thus affecting the durability and reliability of the asphalt pavement structure. Current asphalt pavement design methods use a layered elastic system as the basic theory, treating the pavement surface as an infinitely large plane, and do not yet consider the impact of differences in asphalt pavement surface settlement on the pavement structure.
[0004] In actual road operation, due to a combination of factors, asphalt pavement surfaces often exhibit varying settlement distributions according to certain patterns. These factors include the quality of asphalt pavement materials and insufficient compaction caused by substandard construction quality. To accurately describe the regular settlement distribution of the pavement surface, the random field theory of probability statistics is introduced. The various possible outcomes of random experiments are quantified, and the random experiments can then be characterized by real-valued single-valued functions, i.e., random variables. Furthermore, existing research methods do not combine data simulation results with automated algorithms for multi-objective optimization of pavement structure combinations. Therefore, this paper proposes a three-dimensional modeling and intelligent optimization design method that considers the non-uniform characteristics of pavement. Summary of the Invention
[0005] To address the aforementioned problems in existing technologies, this invention provides a three-dimensional modeling and intelligent optimization design method that considers the uneven characteristics of road surfaces. This method improves the scientific rigor, accuracy, and economy of asphalt pavement structure design, extends the service life of asphalt pavements, reduces design costs, and provides important technical support for the intelligent transformation of road engineering.
[0006] The technical solution to achieve the above objectives is: A three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces includes: Step S1: Based on the road linearity data, drawings, and pavement structure layer information, establish an integrated BIM (Building Information Modeling) model of the asphalt pavement road linearity and pavement structure. Then, use Hypermesh (finite element preprocessing software) to generate and check the fine mesh of the established BIM model to obtain the mesh model, and export it as an inp file. Step S2: The center coordinates of the target element are extracted from the generated inp file using Python. The parameter random field is generated using the covariance decomposition method code in MATLAB. Based on the algorithm, the modulus values of different components are inserted into the mesh to generate multiple inp files. Then, rutting and fatigue numerical simulations are performed in ABAQUS (general finite element analysis software) to obtain the simulated values of rutting deformation and fatigue life of the pavement structure under different modulus values. Step S3: Using the Latin hypercube sampling method, select 1000 sets of asphalt pavement parameter combination samples and calculate the rutting and fatigue life results through the BIM-FEM framework. Combine convolutional neural network and temporal convolutional network to establish a prediction model CNN-TCN-Attention (convolutional neural network-temporal convolutional network-attention mechanism). Use the sample results to train the prediction model and obtain the relationship between rutting and fatigue life and each input variable. Step S4: Optimize the sample data using the NSGA-II (Non-dominated sorting genetic algorithm second generation) optimization algorithm, and use TOPSIS to find the best solution in the Pareto optimal solution to determine the best design scheme for the asphalt pavement structure.
[0007] Preferably, step S1 includes: Step S11: Use road data information to create road lines in Dyanmo (visual programming plugin), and develop Dynamo nodes to establish three-dimensional parametric design curves; Step S12: Read the road surface parameter data, calculate the angle information and corresponding point position information of each road segment according to the given road structure parameters, determine and place the adaptive family position of the road structure, establish a parametric BIM model of the road structure and assign road material and material properties. Step S13: Export the BIM model created in Revit (live modification) as a .sat file, and record data such as geometric properties, material properties, environmental factors and load conditions in a .txt file. Import the .sat file into Hypermesh (finite element preprocessing software) for mesh generation, and define the mesh size standard and control standard based on geometric features.
[0008] Preferably, step S2 includes: Step S21: Define the top layer element set Set-top in the finite element model. Use C3D8I type eight-node hexahedral elements. Solve the geometric center coordinates of each element through mathematical operations based on the element node coordinate data. Use Python code to export the mesh model as an inp file. The inp file includes the basic parameters of the pavement structure, the number of meshes, the number of nodes, and the node coordinates. Step S22: Import the output inp file into the MATLAB (Matrix Laboratory) database. Based on the MATLAB platform, construct a random field model of material parameters using the covariance decomposition method. By defining the covariance function and performing matrix decomposition, generate random field samples that conform to statistical characteristics, realize the numerical characterization of the spatial variability of material parameters, and generate multiple inp files with random moduli. Step S23: Import the inp file into ABAQUS, use Python scripts to automate the model processing, perform temperature field simulation and asphalt pavement rutting deformation process in sequence, use static analysis method, apply tire grounding load, calculate and output stress intensity factor at key locations. Step S24: Compare and analyze the numerical simulation results, i.e., the stress intensity factor, with the actual engineering case data to verify the calculation accuracy and reliability of the established analysis framework.
[0009] Preferably, step S3 includes: Step S31: Using the Latin hypercube sampling method, 1000 representative asphalt pavement parameter combinations are selected as input samples. The ABAQUS calculation results are extracted through an automated processing flow to obtain the corresponding 1000 sets of rutting deformation and fatigue life data. Step S32: Construct a deep learning model CNN-TCN-Attention that integrates convolutional neural networks, temporal convolutional networks, and attention mechanisms; Step S33: Divide the sample data into training set and test set in a ratio of 8:2. Evaluate the prediction accuracy of the model by comparing and analyzing the training error and test error. Finally, establish the prediction relationship between asphalt pavement performance parameters and material and structural parameters.
[0010] Preferably, step S4 includes: Step S41: The improved NSGA-II algorithm is used to perform multi-objective optimization design of asphalt pavement performance; Step S42: Based on the established CNN-TCN-Attention deep learning model, a regression prediction system for pavement performance is constructed. After obtaining the Pareto optimal solution set, the TOPSIS (Approximation Ideal Solution Ranking Method) multi-criteria decision-making method is used to select the best balance scheme from the non-dominated solution set, and finally determine the optimal structural design scheme for asphalt pavement.
[0011] Preferably, in step S41, the improved NSGA-II algorithm includes the following three key technological innovations: First, the fast non-dominated sorting algorithm is used to reduce the computational complexity from O(MN) of the traditional method. 3 ) decreased to O(MN) 2 ); Secondly, a crowding distance mechanism is adopted to replace the traditional shared function method; Finally, an elite retention strategy is used to merge the parent and offspring populations to compete.
[0012] Compared with the prior art, the beneficial effects of the present invention are: This invention considers the three-dimensional modeling and automated finite element simulation of the uneven characteristics of asphalt pavement for multi-objective optimization design of pavement. It uses Hypermesh software to re-divide and check the BIM model, increasing the accuracy of the mesh model. It also develops MATLAB code to generate a modulus random field from the generated mesh model, which can more accurately simulate the uneven characteristics of actual asphalt pavement and improve the accuracy of numerical simulation of asphalt pavement. This invention establishes a multi-objective prediction model by combining convolutional neural networks and temporal convolutional networks and introducing an attention mechanism. The model is trained using training and test sets, and the sample data is optimized using the NSGA-II optimization algorithm. Finally, TOPSIS is used to find the optimal solution in the Pareto optimal solution, thus determining the optimal design scheme for asphalt pavement structure. This improves the scientificity, accuracy, and economy of asphalt pavement structure design, extends the service life of asphalt pavement, reduces design costs, and provides important technical support for the intelligent transformation of road engineering. Attached Figure Description
[0013] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a three-dimensional modeling and intelligent optimization design method that takes into account the uneven characteristics of road surfaces according to the present invention; Figure 2 This invention establishes a BIM model and performs fine mesh generation and inspection in Hypermesh software to obtain a specific flowchart of the mesh model. Figure 3 This is another specific flowchart of the process of establishing a BIM model in this invention and performing fine mesh generation and inspection in Hypermesh software to obtain a mesh model. Figure 4 This is a flowchart illustrating the specific process of obtaining simulated values of rutting deformation and fatigue life of road structures under different modulus values in this invention. Figure 5 This is another specific flowchart for obtaining the simulated values of rutting deformation and fatigue life of road structure under different modulus values in this invention; Figure 6 This is a comparative analysis chart of rut depth between traditional empirical formulas and the numerical simulation of rut depth of this invention; Figure 7 This is a flowchart illustrating the specific process of training a prediction model using sample results to obtain the relationship between rutting and fatigue life and various input variables in this invention. Figure 8 This is a flowchart illustrating the process of using the NSGA-II optimization algorithm to optimize sample data and employing TOPSIS to find the optimal solution in the Pareto optimal solution to determine the best design scheme for asphalt pavement structure in this invention. Figure 9 This is a basic flowchart of the NSGA-II algorithm in this invention. Detailed Implementation
[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0015] like Figure 1 As shown, a three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces includes: Step S1: Based on the road linearity data, drawings, and pavement structure layer information, establish an integrated BIM model of asphalt pavement road linearity and pavement structure. Then, use Hypermesh software to generate and check the fine mesh of the established BIM model to obtain the mesh model, and export it as an inp file.
[0016] like Figure 2 , 3 As shown, step S1 includes: Step S11: Use road data information to create road linearity in Dyanmo and develop Dynamo nodes to establish three-dimensional parametric design curves.
[0017] Specifically, the coordinates of the road centerline are read from an Excel spreadsheet using the "File Path" node. This table contains the x and y coordinates, design elevation, and azimuth of each control point. Develop the “Control.CenterLineByPoints” node to read coordinate information from Excel and generate a spline curve as the road centerline; Develop the “Control.3DCurveChainagePoint” node to read the location information of the road route and obtain the coordinates of points at a fixed length along the curve; Create a circle based on a given plane and radius, generate a surface based on the circle, calculate the intersection points between the surface and the object, and create control points for alignment control. Develop the “Control.createThickenAndGetPerimeterCurve” node to offset the road centerline, and then use Loft to generate a surface to obtain the boundary curve of the surface; Then, in Dynamo, the "Control.JoinCurves" node is developed using Python Script to group the boundary curves and use them as control lines for the road surface structure model, providing a foundation for the subsequent road surface structure model.
[0018] Step S12: Read the road surface parameter data, calculate the angle information and corresponding point position information of each road segment based on the given road structure parameters, determine and place the adaptive family position of the road structure, establish a parametric BIM model of the road structure and assign road material and material properties.
[0019] Specifically, the "ReadExcel.ImportExcel" node is used to read pavement layer parameter data from an Excel spreadsheet, including the material, thickness, pavement width, slope angle, cross-sectional angle, pavement rotation angle, and surface slope of each structural layer. To achieve a parametric asphalt pavement model, adaptive points are created to indirectly generate an adaptive pavement structure. The Python script code uses given pavement structure parameters to calculate the angle information and corresponding point position information of each road segment and returns the results for modeling. In the required library, define a function to split the list into blocks of a specified size, and input the required parameters "aligns", "segments", and "lengths". Calculate the angle and point position for each segment and store the results in the corresponding list. Finally, the calculated angles and point positions are output to generate an adaptive pavement structure family with detailed parameter properties, so that a parameterized pavement structure model can be built by changing the relevant parameters. Creating a parametric design route using the "Solid" node in Dynamo: First, create a parametric design route using the "NurbsCurve" node; then, use "PolyCurve" at key control points to generate parametric road cross-section profiles for different structural layers; finally, use the Solid node to sweep along the path to establish a parametric pavement structure model, and use the Elment.SetPaeameterByname node to automatically import the generated model as a family instance into the Revit platform; Use the "Springs.FamilyInstance.ByGeometry" node to link geometric and material properties to BIM model properties, providing data support for subsequent design changes and numerical simulations.
[0020] Step S13: Export the BIM model created in Revit as a .sat file, and record data such as geometric properties, material properties, environmental factors and load conditions in a .txt file. Import the .sat file into Hypermesh for mesh generation, and define the mesh size standard and control standard based on geometric features.
[0021] Specifically, Revit outputs the BIM model as a .sat file, and outputs geometric properties, material properties, environmental factors, and load conditions to a .txt file; Hypermesh software is used to mesh .sat files. The mesh size and control criteria are defined based on the geometric features in the .criteria and .param files. The .tcl script calls these two files to achieve automated mesh partitioning, checking and optimization.
[0022] Common random field simulation methods include the moving average method, the discrete Fourier transform method, and the covariance matrix decomposition method. Among these, the covariance matrix decomposition method has a wide range of applications, high computational efficiency, and is relatively easy to unify with the model used in finite element analysis. Based on this, step S2 selects the covariance matrix decomposition method to calculate the random modulus distribution of asphalt pavement.
[0023] Step S2: The center coordinates of the target unit are extracted from the generated inp file using Python. A parametric random field is generated using MATLAB code based on the covariance decomposition method. The modulus values of different components are inserted into the grid according to the algorithm to generate multiple inp files. Then, rutting and fatigue numerical simulations are performed in ABAQUS to obtain the simulated values of rutting deformation and fatigue life of the pavement structure under different modulus values.
[0024] like Figure 4-6 As shown, step S2 includes: Step S21: Define the top layer element set Set-top in the finite element model. Use C3D8I type eight-node hexahedral elements. Solve the geometric center coordinates of each element through mathematical operations based on the element node coordinate data. Use Python code to export the mesh model as an inp file. The inp file includes the basic parameters of the pavement structure, the number of meshes, the number of nodes, and the node coordinates.
[0025] Step S22: Import the output inp file into the MATLAB database. Based on the MATLAB platform, construct a random field model of material parameters using the covariance decomposition method. By defining the covariance function and performing matrix decomposition, generate random field samples that conform to statistical characteristics, realize the numerical characterization of the spatial variability of material parameters, and generate multiple inp files with random moduli.
[0026] Specifically, the coordinates of the grid center nodes are calculated and generated using MATLAB code based on the covariance function, and the process is as follows: Step 1: Set the road surface material parameters, elastic modulus and Poisson's ratio, and set the material parameter variation space, design the coefficient of variation and mean, determine the type of autocorrelation function, and determine the horizontal and vertical autocorrelation distances.
[0027] Step 2: Input the road surface material parameters, select the center point interpolation calculation method, and input the element type and parameter values of the model.
[0028] Step 3: Calculate the random field, define the set of random fields to be assigned, extract all nodes, all elements and element node numbers in the component, save the element number to be assigned, calculate the center coordinates of each element using the interpolation method, and export the element center coordinates to the MATLAB software database.
[0029] Based on the element center coordinates calculated in the previous steps, different material parameter modulus values are assigned to different coordinates using MATLAB code, generating inp files with different parameter values. The process is as follows: Step 1: Parameter settings, determine the number of .inp file types, the number of random field simulations, the number of element types, and determine the save path for the newly generated .inp files.
[0030] Step 2: Determine the components that need to be assigned random field values, and determine the set that needs to be assigned random field values, which belongs to a part of the components.
[0031] Step 3: Generate the inp file, extract all nodes in the component, extract all elements and element node numbers in the component, determine the elements that need to be assigned random fields, and remove redundant sets and faces.
[0032] Step 4: Calculate material parameters and generate an inp file. Calculate the random field using the center point interpolation method and perform linear interpolation based on the calculated node file.
[0033] Step 5: Modify the material parameters, generate the inp file, determine the number of lines in the inp file for the material parameters, determine the number of material parameters, calculate the random parameter values, and write them in the inp file.
[0034] Step S23: Import the inp file into ABAQUS, use Python scripts to automate the model processing, perform temperature field simulation and asphalt pavement rutting deformation process in sequence, use static analysis method, apply tire grounding load, calculate and output stress intensity factor at key locations.
[0035] Specifically, the inp file generated in MATLAB is imported into the ABAQUS finite element software. A road structure model is built in the inp file. The assigned mesh model is combined with the original model using the finite element software assembly method. Bound contact is used to constrain the model, thereby merging the two models into one. Finally, a road finite element structure model with different parameter moduli is obtained. Then, Python is used to automatically encode the inp file, including geometric information, material properties, and boundary definitions. After that, FORTRAN is used to define the temperature and heat flow that change with time in the external environment to simulate the temperature field of the road structure under external conditions. First, the corresponding temperature field is imported by adding initial conditions using "type=TEMPERATURE,type=pave-tem.odb". Then, elastic analysis step (Static) and creep analysis step (Visco) are established in the Step module. The total analysis step time is set to the cumulative load application time. Then, tire grounding load is applied to simulate road rutting after 2 million standard axle cycles. Re-import the mesh model into the inp file, define the viscoelastic properties of the material, select the static analysis mode, apply tire ground load, output the stress intensity factor, and calculate the fatigue life of the asphalt pavement using the stress life method.
[0036] Step S24: Compare and analyze the numerical simulation results, i.e., the stress intensity factor, with the actual engineering case data to verify the calculation accuracy and reliability of the established analysis framework.
[0037] Step S3: Using the Latin hypercube sampling method, select 1000 sets of asphalt pavement parameter combination samples and calculate the rutting and fatigue life results through the BIM-FEM framework. Combine convolutional neural network (CNN) and temporal convolutional network (TCN) to establish a prediction model CNN-TCN-Attention. Use the sample results to train the prediction model and obtain the relationship between rutting and fatigue life and each input variable.
[0038] like Figure 7 As shown, step S3 includes: Step S31: Using the Latin hypercube sampling method, 1000 representative asphalt pavement parameter combinations are selected as input samples. The ABAQUS calculation results are extracted through an automated processing flow to obtain the corresponding 1000 sets of rutting deformation and fatigue life data. Step S32: Construct a deep learning model CNN-TCN-Attention that integrates convolutional neural networks, temporal convolutional networks, and attention mechanisms; Step S33: Divide the sample data into training set and test set in a ratio of 8:2. Evaluate the prediction accuracy of the model by comparing and analyzing the training error and test error. Finally, establish the prediction relationship between asphalt pavement performance parameters and material and structural parameters.
[0039] Specifically, the prediction model CNN-TCN-Attention structure mainly includes 6 convolutional layers, 6 pooling layers, 1 fully connected layer, 2 TCN layers, and 1 attention layer; the number of convolutional kernels in the convolutional layers are 16, 32, 32, 64, 64, and 64 respectively, while the region size of the pooling layer is set to 2×1; the dropout ratio of the model is set to 0.25, the learning rate is 0.008, and the number of training epochs is 1000.
[0040] To evaluate the model's performance, root mean square error (RMSE), mean absolute error (MAE), and goodness of fit were used as evaluation metrics for the prediction model. Goodness of fit represents the correlation between the model's features and the prediction target.
[0041] Step S4: Optimize the sample data using the NSGA-II optimization algorithm, and use TOPSIS to find the best solution in the Pareto optimal solution to determine the optimal design scheme for the asphalt pavement structure.
[0042] like Figure 8 , 9 As shown, step S4 includes: Step S41: The improved NSGA-II algorithm is used to perform multi-objective optimization design of asphalt pavement performance.
[0043] The improved NSGA-II algorithm includes the following three key technological innovations: First, the fast non-dominated sorting algorithm is used to reduce the computational complexity from O(MN) of the traditional method. 3 ) decreased to O(MN) 2 ); Secondly, a crowding distance mechanism is adopted to replace the traditional shared function method; Finally, an elite retention strategy is used to merge the parent and offspring populations to compete.
[0044] By analyzing the advantages of different optimization algorithms, it was found that the NSGA-II optimization algorithm is more effective in predicting rutting and fatigue life of asphalt pavements. The NSGA-II algorithm is used for multi-objective optimization, featuring elitism and rapid non-dominant ranking. For multi-objective optimization problems, the task is to find the optimal compromise solution in the design domain, known as the "Pareto front." The basic process of the NSGA-II algorithm is as follows: 1) Initialize the population Obtain the parent population; 2) Obtain the offspring population through fast non-dominant sorting and genetic operators; 3) Merge the parent and offspring populations, perform rapid non-dominant ranking, and calculate crowding distance; 4) New parental populations are generated by selecting suitable individuals; 5) Based on the individuals of the new parent population, use genetic operators to obtain a new offspring population, and continue for one generation until the maximum number of generations is reached.
[0045] It is important to note that performing a fast, non-explicit sort before selection, preserving good solutions in the next generation, and crossovering and mutation of bad solutions aligns with elitist principles.
[0046] Obtaining the Pareto optimal solution set is crucial for multi-objective optimization. When seeking the Pareto optimal solution set, the non-dominant front is formed by the solution set in the population, which is not dominated by any other solution; for a size of The population can be stratified using a fast non-explicit sorting algorithm; in order to distribute the solutions evenly on the Pareto front, the non-primary solutions in the same layer should be diverse; otherwise, individual solutions will be concentrated in a certain position, which will make it difficult to obtain the Pareto optimal solution set.
[0047] The NSGA-II algorithm employs the crowding distance method to find the density of other solutions surrounding a specific solution in the same non-dominant layer. To ensure a uniform distribution of the solution set across the target optimization region, the NSGA-II algorithm incorporates the calculation of crowding distance. It equals the sum of the distances between any two adjacent solutions in each optimization objective, that is: ; In the formula, For the first The individual in the first Function values on each target For the first The individual in the first Function values on each target The first in the current non-dominated layer The maximum value of each target. The first in the current non-dominated layer The minimum value of each objective; To enable competition between parents and offspring to produce a new generation of individuals, and to preserve the optimal solution obtained to promote algorithm convergence, the NSGA-II algorithm introduces elitism. The elitism process is as follows: 1) Combining offspring populations Japanese parent group Thus forming the whole Size is (Assume the initial population size is) ) 2) To Perform a fast, non-explicit sort and calculate the crowding distance to obtain the non-dominant set. .
[0048] Step S42: Based on the established CNN-TCN-Attention deep learning model, a regression prediction system for pavement performance is constructed. After obtaining the Pareto optimal solution set, the TOPSIS multi-criteria decision method is used to select the best balance scheme from the non-dominated solution set, and finally determine the optimal structural design scheme for asphalt pavement.
[0049] Based on the above reasons, the sample data was optimized using the NSGA-II optimization algorithm, and the optimal solution was found in the Pareto optimal solution using TOPSIS to determine the best design scheme for the asphalt pavement structure.
[0050] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces, characterized in that, include: Step S1: Based on the road linearity data, drawings, and pavement structure layer information, establish an integrated BIM model of asphalt pavement road linearity and pavement structure. Then, use Hypermesh software to generate and check the fine mesh of the established BIM model to obtain the mesh model and export it as an inp file. Step S2: The center coordinates of the target unit are extracted from the generated inp file using Python. The parameter random field is generated using the covariance decomposition method code in MATLAB. Based on the algorithm, the modulus values of different components are inserted into the grid to generate multiple inp files. Then, rutting and fatigue numerical simulations are performed in ABAQUS to obtain the simulated values of rutting deformation and fatigue life of the pavement structure under different modulus values. Step S3: Using the Latin hypercube sampling method, select 1000 sets of asphalt pavement parameter combination samples and calculate the rutting and fatigue life results through the BIM-FEM framework. Combine convolutional neural network and temporal convolutional network to establish a prediction model CNN-TCN-Attention. Use the sample results to train the prediction model and obtain the relationship between rutting and fatigue life and each input variable. Step S4: Optimize the sample data using the NSGA-II optimization algorithm, and use TOPSIS to find the best solution in the Pareto optimal solution to determine the optimal design scheme for the asphalt pavement structure.
2. The three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces according to claim 1, characterized in that, Step S1 includes: Step S11: Use road data information to create road linearity in Dyanmo and develop Dynamo nodes to establish three-dimensional parametric design curves; Step S12: Read the road surface parameter data, calculate the angle information and corresponding point position information of each road segment according to the given road structure parameters, determine and place the adaptive family position of the road structure, establish a parametric BIM model of the road structure and assign road material and material properties. Step S13: Export the BIM model created in Revit as a .sat file, and record data such as geometric properties, material properties, environmental factors and load conditions in a .txt file. Import the .sat file into Hypermesh for mesh generation, and define the mesh size standard and control standard based on geometric features.
3. The three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces according to claim 1, characterized in that, Step S2 includes: Step S21: Define the top layer element set Set-top in the finite element model. Use C3D8I type eight-node hexahedral elements. Solve the geometric center coordinates of each element through mathematical operations based on the element node coordinate data. Use Python code to export the mesh model as an inp file. The inp file includes the basic parameters of the pavement structure, the number of meshes, the number of nodes, and the node coordinates. Step S22: Import the output inp file into the MATLAB database. Based on the MATLAB platform, construct a random field model of material parameters using the covariance decomposition method. By defining the covariance function and performing matrix decomposition, generate random field samples that conform to statistical characteristics, realize the numerical characterization of the spatial variability of material parameters, and generate multiple inp files with random moduli. Step S23: Import the inp file into ABAQUS, use Python scripts to automate the model processing, perform temperature field simulation and asphalt pavement rutting deformation process in sequence, use static analysis method, apply tire grounding load, calculate and output stress intensity factor at key locations. Step S24: Compare and analyze the numerical simulation results, i.e., the stress intensity factor, with the actual engineering case data to verify the calculation accuracy and reliability of the established analysis framework.
4. The three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces according to claim 1, characterized in that, Step S3 includes: Step S31: Using the Latin hypercube sampling method, 1000 representative asphalt pavement parameter combinations are selected as input samples. The ABAQUS calculation results are extracted through an automated processing flow to obtain the corresponding 1000 sets of rutting deformation and fatigue life data. Step S32: Construct a deep learning model CNN-TCN-Attention that integrates convolutional neural networks, temporal convolutional networks, and attention mechanisms; Step S33: Divide the sample data into training set and test set in a ratio of 8:
2. Evaluate the prediction accuracy of the model by comparing and analyzing the training error and test error. Finally, establish the prediction relationship between asphalt pavement performance parameters and material and structural parameters.
5. The three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces according to claim 1, characterized in that, Step S4 includes: Step S41: The improved NSGA-II algorithm is used to perform multi-objective optimization design of asphalt pavement performance; Step S42: Based on the established CNN-TCN-Attention deep learning model, a regression prediction system for pavement performance is constructed. After obtaining the Pareto optimal solution set, the TOPSIS multi-criteria decision method is used to select the best balance scheme from the non-dominated solution set, and finally determine the optimal structural design scheme for asphalt pavement.
6. The three-dimensional modeling and intelligent optimization design method considering the uneven characteristics of road surfaces according to claim 5, characterized in that, In step S41, the improved NSGA-II algorithm includes the following three key technological innovations: First, the fast non-dominated sorting algorithm is used to reduce the computational complexity from O(MN) of the traditional method. 3 ) decreased to O(MN) 2 ); Secondly, a crowding distance mechanism is adopted to replace the traditional shared function method; Finally, an elite retention strategy is used to merge the parent and offspring populations to compete.