Method, device, equipment and medium for regularizing irregular shape of road hidden disease

Abstract

The application discloses a road hidden disease irregular shape regularization method and device, equipment and medium, relates to road disease detection and structure performance evaluation technical field, including: through relying on three-dimensional ground penetrating radar multi-attribute data, disease is segmented, based on the geometric characteristics of the disease, an equivalent inertia ellipsoid is constructed as a geometric initial model, the damage factor is obtained by fusing radar multi-medium attribute mapping modulus constraint, a multi-constraint multi-physical field weighted error optimization model is built, and the optimal equivalent ellipsoid and modulus are obtained by means of global iteration optimization of the surrogate model, and then finite element mechanics evaluation is carried out. The application not only retains the macro mechanics geometry and stiffness characteristics of the real disease, improves the equivalent simulation accuracy, avoids the grid division problem of the irregular model, greatly reduces the iterative calculation amount relying on the surrogate model, and is suitable for batch structure bearing capacity evaluation of road diseases.

Description

Technical Field

[0001] This invention relates to the field of road defect detection and structural performance evaluation technology, and in particular to a method, device, equipment and medium for regularizing the irregular morphology of hidden road defects. Background Technology

[0002] Currently, finite element modeling of road defects based on radar data falls into two main categories. The first approach simplifies the irregular defects obtained from ground-penetrating radar (GPR) data segmentation into standard regular geometric shapes such as spheres and cubes, directly embedding them into the layered road structure model for mechanical simulation. The second approach fully preserves the original irregular three-dimensional contours of the defects output from GPR data segmentation, directly importing them into finite element software for modeling and calculation. However, existing simulation methods rely solely on a single radar attenuation parameter to roughly set the defect material modulus, using only single-point deflection of the pavement as the mechanical equivalence criterion. The optimization process involves repeatedly calling the finite element solver for iterative calculations, relying on a small number of field-measured defects for optimization samples, lacking standardized methods for generating batch samples, and separately debugging and optimizing the geometric model and defect stiffness parameters.

[0003] However, the simplified geometric modeling method completely ignores the anisotropic spatial extension characteristics brought about by the irregular shape of the actual disease, changes the stress concentration area and damage evolution path of the disease, and significantly reduces the accuracy of road bearing capacity evaluation. Secondly, the geometric cleanup workload of retaining the original irregular contour modeling method is huge, the mesh generation failure rate is high, the finite element solution convergence is difficult, and it cannot adapt to the needs of batch engineering inspection. The disease modulus relies only on the empirical assignment of a single radar parameter, lacks a multi-media attribute fusion quantitative mapping method, and the material stiffness parameter has a low matching degree with the loose and water-containing state of the actual disease. Only the pavement deflection single index is used to evaluate the accuracy of the equivalent model, without taking into account key mechanical indicators such as the tensile stress at the bottom of the layer and the shear stress at the interface, resulting in insufficient mechanical fit of the equivalent model. Direct iterative finite element parameter optimization calculation is extremely costly, and there is no offline sample library and proxy model dimensionality reduction acceleration method, resulting in low optimization efficiency; the geometric shape and disease modulus are optimized separately, and it is impossible to achieve simultaneous mechanical equivalence of morphology and stiffness in both dimensions; some optimization schemes add additional surface area constraints, but there is no corresponding disease surface area extraction and calculation process, and the constraint conditions cannot be implemented.

[0004] Therefore, how to automatically construct a regularized equivalent model of internal road defects that balances mechanical equivalence accuracy and finite element calculation efficiency based on multi-attribute data from 3D ground-penetrating radar, and simultaneously achieve synergistic optimization of defect geometry and damage stiffness to reduce the computational cost of batch road bearing capacity evaluation has become an urgent problem to be solved. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device, equipment and medium for regularizing the irregular shape of hidden road defects, aiming to solve the technical problem of how to achieve synergistic optimization of defect geometry and damage stiffness, and reduce the calculation cost of batch road bearing capacity evaluation.

[0006] To achieve the above objectives, this application proposes a method for regularizing the irregular morphology of hidden road defects, including: Three-dimensional ground-penetrating radar data of the target road area is collected and preprocessed. Amplitude attribute information and multi-dimensional medium attribute information are extracted from the preprocessed three-dimensional data volume. The disease area is segmented according to the amplitude attribute information to obtain an irregular disease binary volume and multi-attribute distribution data. The volume, centroid, and moment of inertia tensor are calculated based on the irregular disease binary body. The principal axis directions are obtained by eigenvalue decomposition of the moment of inertia tensor. An equivalent inertial ellipsoid is constructed based on the volume, centroid, and principal axis directions to obtain a geometrically equivalent initial solution. The multi-attribute distribution data is input into a preset nonlinear mapping model for multi-attribute fusion mapping to obtain the spatial distribution of damage factors, and material parameter prior constraints are generated based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials. An optimization model is constructed based on the prior constraints of geometric and material parameters in the geometrically equivalent initial solution. According to the optimization model, the geometric parameters and equivalent initial elastic modulus are globally optimized and iterated using a preset surrogate model to obtain the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus. The equivalent initial elastic modulus is determined by three-dimensional ground-penetrating radar data through multi-attribute extraction, damage mapping and volume weighted averaging. Based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, a finite element model of a road containing equivalent regularized defects is constructed and mechanical solutions are performed to obtain the evaluation results of the road structure's mechanical response.

[0007] Furthermore, to achieve the above objectives, this application also proposes a device for regularizing the irregular morphology of hidden road defects, the device comprising: The extraction module is used to collect three-dimensional ground-penetrating radar data of the target road area and perform preprocessing. It extracts amplitude attribute information and multi-dimensional medium attribute information from the preprocessed three-dimensional data volume, and segments the disease area according to the amplitude attribute information to obtain irregular disease binary volume and multi-attribute distribution data. The geometric equivalent module is used to calculate the volume, centroid, and moment of inertia tensor based on the irregular disease binary body, perform eigenvalue decomposition on the moment of inertia tensor to obtain the direction of the principal axis of inertia, construct an equivalent inertial ellipsoid based on the volume, centroid, and direction of the principal axis of inertia, and obtain the geometric equivalent initial solution. The material mapping module is used to input the multi-attribute distribution data into a preset nonlinear mapping model to perform multi-attribute fusion mapping, obtain the spatial distribution of damage factors, and generate prior constraints on material parameters based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials. The construction module is used to construct an optimization model based on the prior constraints of geometric parameters and material parameters in the geometrically equivalent initial solution; The optimization module is used to perform global optimization iteration on geometric parameters and equivalent initial elastic modulus according to the optimization model using a preset surrogate model to obtain the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus. The equivalent initial elastic modulus is determined by three-dimensional ground penetrating radar data through multi-attribute extraction, damage mapping and volume weighted averaging. The results module is used to construct a finite element model of a road containing equivalent regularized defects based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, and to perform mechanical solution to obtain the evaluation results of the mechanical response of the road structure.

[0008] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method for regularizing the irregular morphology of hidden road defects as described above.

[0009] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the method for regularizing irregular morphology of hidden road defects as described above.

[0010] This application utilizes multi-attribute data from 3D ground-penetrating radar to segment road defects. Based on the geometric features of the defects, an equivalent inertial ellipsoid is constructed as the initial geometric model. Modulus constraints are obtained by integrating radar multi-medium attribute mapping damage factors. A multi-constraint, multi-physics field weighted error optimization model is built. Using a surrogate model, the optimal equivalent ellipsoid and modulus are obtained through global iterative optimization, followed by finite element mechanical evaluation. This application retains the macroscopic mechanical geometry and stiffness characteristics of real road defects, improving the accuracy of equivalent simulation, while avoiding the challenges of meshing irregular models. The surrogate model significantly reduces the amount of iterative computation, making it suitable for batch road defect bearing capacity testing. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1This is a flowchart illustrating the first embodiment of the method for regularizing the irregular morphology of hidden road defects in this application. Figure 2 This is a flowchart illustrating the second embodiment of the method for regularizing the irregular morphology of hidden road defects in this application; Figure 3 This is a schematic diagram of the modular structure of the road hidden defects irregular shape regularization device of this application; Figure 4 This is a schematic diagram of the equipment structure of the hardware operating environment involved in the method for regularizing the irregular morphology of hidden road defects in the embodiments of this application.

[0013] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0014] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0015] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0016] Hidden road defects can easily expand and cause road surface collapse. The industry needs to combine 3D ground-penetrating radar and falling weight deflectometers to conduct joint evaluations of defect morphology and bearing capacity. In engineering, finite element simulation is often used to replace on-site measurements. Existing solutions either simplify defects into standard geometric shapes, resulting in distorted mechanical simulations, or directly use original irregular defect models, leading to difficulties in mesh processing and low solution efficiency. At the same time, the defect modulus is only assigned empirically based on a single radar parameter, and the equivalent effect is judged only by the single index of deflection. Direct iterative finite element optimization is computationally expensive, and geometric and stiffness parameters cannot be optimized in a coordinated manner.

[0017] Therefore, how to automate the entire process of foreign object image recognition, risk quantification and assessment, and intelligent regulatory retrieval and decision-making has become an urgent problem to be solved.

[0018] Based on the above, this application also provides a method for regularizing the irregular morphology of hidden road defects, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the method for regularizing the irregular morphology of hidden road defects in this application.

[0019] In the first embodiment, the method for regularizing the irregular morphology of hidden road defects includes steps S10 to S60: Step S10: Collect three-dimensional ground-penetrating radar data of the target road area and preprocess it. Extract amplitude attribute information and multi-dimensional medium attribute information from the preprocessed three-dimensional data volume. Segment the disease area according to the amplitude attribute information to obtain irregular disease binary volume and multi-attribute distribution data.

[0020] It should be noted that 3D ground-penetrating radar data refers to the raw electromagnetic wave reflection data obtained by scanning the underground structure of a road using a multi-channel ground-penetrating radar system; amplitude attribute information refers to the spatial distribution data of the intensity of electromagnetic wave reflection signals, used to characterize the wave impedance differences at the interface of the underground medium; multi-dimensional medium attribute information can be parameters such as center frequency attenuation, instantaneous phase continuity, amplitude statistical variance, or dielectric constant dispersion gradient, used to reflect the porosity or water content of the medium inside the disease; irregular disease binary volume refers to a 3D data volume that uses binary functions to characterize the spatial distribution of the disease, where voxels belonging to the disease area are marked as 1, and other areas are marked as 0; multi-attribute distribution data refers to the set of multi-dimensional attribute values ​​corresponding to the disease area extracted from the medium attribute volume.

[0021] Further, step S10 includes: First, using a multi-channel three-dimensional ground-penetrating radar system to scan along the target road area to obtain raw scan data, which records the reflection response information of the underground medium to electromagnetic waves.

[0022] Secondly, zero-bias correction and background denoising are performed sequentially on the original scan data. Zero-bias correction is used to eliminate the DC offset of the system, and background denoising is used to suppress direct wave and antenna coupling noise, so as to obtain the corrected data.

[0023] Next, bandpass filtering and migration imaging are performed on the corrected data. Bandpass filtering is used to filter out interference signals outside the frequency range, and migration imaging is used to reposition the tilted reflective interface to its true spatial location, resulting in a three-dimensional data volume reflecting the differences in underground dielectric properties. .

[0024] Next, the amplitude attribute volume is extracted from the three-dimensional data volume. The amplitude attribute volume is used to characterize the spatial distribution of electromagnetic wave reflection intensity, depict the interface morphology between the disease and the surrounding medium, and obtain amplitude attribute information.

[0025] Then, the center frequency attenuation attribute volume, instantaneous phase continuity attribute volume, and amplitude statistical variance attribute volume are extracted from the three-dimensional data volume. The center frequency attenuation attribute reflects the degree of dispersion loss of electromagnetic waves in the diseased medium, the instantaneous phase continuity attribute reflects the phase jump characteristics of the disease boundary, and the amplitude statistical variance attribute reflects the inhomogeneity of the medium inside the disease, thus obtaining multi-dimensional medium attribute information.

[0026] Then, based on the amplitude attribute information, threshold segmentation and 3D region growing are performed. Threshold segmentation is used to initially determine the boundary of the disease area, and 3D region growing is used to connect adjacent voxels of the same type and fill internal voids to obtain an irregular disease binary volume. The specific formula is expressed as follows: Finally, the data from the multi-dimensional media attribute information and the corresponding regions of the irregular disease binary body are associated and stored, i.e., extracted. The center frequency attenuation value, instantaneous phase continuity value, and amplitude statistical variance value corresponding to the region form a data set with one-to-one correspondence between spatial location and attribute value, resulting in multi-attribute distribution data.

[0027] Step S20: Calculate the volume, centroid, and moment of inertia tensor based on the binary volume of the irregular disease. Perform eigenvalue decomposition on the moment of inertia tensor to obtain the direction of the principal axis of inertia. Construct an equivalent inertial ellipsoid based on the volume, centroid, and direction of the principal axis of inertia to obtain the geometrically equivalent initial solution.

[0028] It should be noted that the moment of inertia tensor is a second-order symmetric tensor describing the moment of inertia and product of inertia of a rigid body about its axes, and its components are determined by the coordinate integral of the diseased region relative to the center of mass; the principal axis directions refer to the directions of the eigenvectors obtained after the eigenvalue decomposition of the moment of inertia tensor, representing the three orthogonal principal directions of the spatial extension of the disease; the equivalent inertial ellipsoid is a triaxial ellipsoid with the same volume, the same center of mass, and the same principal axis directions as the irregular disease, and its semi-axis length is determined by solving the eigenvalues ​​of the moment of inertia tensor simultaneously; the geometrically equivalent initial solution is a parameter vector containing the equivalent ellipsoid semi-axis length, the coordinates of the center of mass, and the attitude Euler angles, which serves as the starting point for subsequent mechanical optimization iterations.

[0029] Further, step S20 includes: First, performing voxel integration based on the coordinates and state values ​​of each voxel in the binary volume of the irregular disease; multiplying the coordinates of voxels belonging to the disease region (state value 1) by the voxel volume element and summing the results; then dividing by the product of the total number of voxels and the voxel volume element to obtain the volume, as shown in the specific formula: in Indicates the total volume of the affected area; Represents voxel coordinates in three-dimensional space; Represents the binary volume of irregular disease in coordinates The status value at this location is 1 if it belongs to a diseased area, and 0 otherwise. Indicates along The volume elements of voxels in three directions (i.e., the side length of a single voxel in the corresponding direction).

[0030] Simultaneously, the weighted integral of each voxel coordinate and state value is divided by the volume to obtain the centroid coordinates, as shown in the following formula: in This indicates the coordinates of the centroid of the diseased area.

[0031] Then, establish a relative coordinate system with the center of mass as the origin, calculate the moment of inertia tensor of the irregular disease binary volume relative to the center of mass, and let... The tensor components are in the following form: in The first tensor of the moment of inertia tensor Line 1 Column component; The Kroneck symbol is used when... hour ,when hour ; Represents the relative coordinates of a voxel with respect to its center of mass; These represent the relative coordinates at... direction and The directional component.

[0032] Next, the moment of inertia tensor is subjected to eigenvalue decomposition, and the expression of the eigenvalue decomposition is obtained. Solving this problem yields multiple eigenvalues ​​and their corresponding unit eigenvectors. In this example, three eigenvalues ​​are obtained. And the corresponding three unit eigenvectors.

[0033] Next, the directions of the principal axes of inertia are determined based on multiple unit eigenvectors, and the corresponding Euler angles are obtained. The directions of the principal axes of inertia are determined based on three unit eigenvectors; these three eigenvectors represent the three principal axes of inertia of the defect, and the corresponding Euler angles are denoted as... .

[0034] Then, based on the volume and the direction of the principal inertial axis, the lengths of multiple semi-axes of the equivalent inertial ellipsoid are solved simultaneously to obtain the set of semi-axe lengths. Specifically, a triaxial ellipsoid with the same volume as the lesion, the same center of mass, and principal axis directions consistent with the principal inertial axis directions is constructed. The moments of inertia of the homogeneous ellipsoid about the three principal axes are: in Indicates a homogeneous ellipsoid revolving around Moment of inertia of the shaft (major axis); Indicates a homogeneous ellipsoid revolving around Moment of inertia of the shaft (central shaft); Indicates a homogeneous ellipsoid revolving around Moment of inertia of the shaft (short shaft). This represents the length of the semi-major axis of the equivalent inertial ellipsoid; This represents the length of the mid-semi-axis of the equivalent inertial ellipsoid; Let represent the length of the minor semi-axis of the equivalent inertial ellipsoid. , , Solve simultaneously for the lengths of the three semi-axes of the equivalent inertial ellipsoid: Finally, a parameter vector is constructed based on the set of semi-axis lengths, the centroid, and the Euler angles to obtain the geometrically equivalent initial solution.

[0035] Step S30: Input the multi-attribute distribution data into the preset nonlinear mapping model to perform multi-attribute fusion mapping, obtain the spatial distribution of damage factors, and generate prior constraints on material parameters based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials.

[0036] It should be noted that the preset nonlinear mapping model can be a support vector regression model or a neural network model, used to map multi-dimensional radar attributes into damage factors; the spatial distribution of damage factors refers to the quantitative distribution of the degree of material damage at each point in the damaged area, with a value range of 0 to 1, where 0 indicates that the material is intact and 1 indicates that it has completely lost its load-bearing capacity; the preset elastic modulus of healthy materials refers to the benchmark value of the elastic modulus of the material in the undamaged area of ​​the road structure layer; the preset modulus attenuation coefficient is a proportional coefficient used to determine the ratio of the lower limit of the elastic modulus of the damaged area to the elastic modulus of the healthy material, usually taken as 0.1; the prior constraints of material parameters refer to the set of parameters including the equivalent initial elastic modulus and the modulus search boundary, providing the initial values ​​and feasible region of material parameters for subsequent optimization iterations.

[0037] Further, step S30 includes: first, extracting the center frequency attenuation attribute, instantaneous phase continuity attribute, amplitude statistical variance attribute, and dielectric constant dispersion gradient attribute from the multi-attribute distribution data, and combining the four attribute values ​​according to spatial location into a feature vector to obtain a multi-attribute vector.

[0038] The multi-attribute vector is then input into a preset nonlinear mapping model for multi-attribute fusion mapping. Taking the center frequency attenuation attribute as an example, the normalized attenuation rate is the percentage decrease in center frequency from the surface to that point. The spatial distribution of damage factors is obtained through the nonlinear mapping function. The specific formula is as follows: in This represents the normalized decay rate; This indicates the preset minimum attenuation threshold; This indicates the preset maximum attenuation threshold. The two thresholds are determined comprehensively based on the center frequency of the ground penetrating radar, the type of underground medium, and on-site calibration data.

[0039] Next, based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials, the equivalent elastic modulus of the damaged area is calculated using a volume-weighted average to obtain the equivalent initial elastic modulus. The specific formula is as follows: in Indicates the equivalent initial elastic modulus; This indicates that the volume of the diseased area is integrated; This indicates the preset elastic modulus of healthy materials, which is the benchmark value of the elastic modulus of materials in undamaged areas of the road structure layer; Represents a volume element. .

[0040] Then, based on the preset modulus attenuation coefficient and the preset elastic modulus of healthy materials, the lower limit of the modulus is calculated, and the preset elastic modulus of healthy materials is used as the upper limit of the modulus to obtain the modulus search boundary. ,in ， This is the preset modulus attenuation coefficient, which is a small positive coefficient, usually 0.1.

[0041] Finally, the equivalent initial elastic modulus and the modulus search boundary are associated and stored to obtain the prior constraints of the material parameters.

[0042] Step S40: Construct an optimization model based on the prior constraints of geometric and material parameters in the geometrically equivalent initial solution.

[0043] It should be noted that the multi-physics response weighted comprehensive error refers to the weighted sum of the normalized errors of various mechanical response quantities such as pavement deflection response, bottom layer horizontal tensile stress, interlayer shear stress, and pavement surface strain energy density; the volume identity constraint refers to the equality constraint condition that keeps the volume of the equivalent ellipsoid equal to the volume of the original defect during the optimization process; the surface area approximation constraint refers to the constraint condition that limits the deviation between the surface area of ​​the equivalent ellipsoid and the surface area of ​​the original defect to not exceed the preset range, which is used to prevent geometric distortion during the optimization process.

[0044] Further, step S40 includes: first, merging the multiple semi-axis lengths, centroid coordinates, and attitude Euler angles in the geometrically equivalent initial solution with the equivalent initial elastic modulus into an optimization variable vector. ; Secondly, establish volume identity constraints based on volume, the specific formula being: .

[0045] Represent one of the multiple semi-axis lengths as a function of the remaining semi-axis length and volume, for example... We obtain the independent optimization variables after dimensionality reduction.

[0046] Next, according to the preset measurement point layout plan, multiple deflection measurement point positions are determined on the road surface. The deflection measurement points are arranged longitudinally and laterally along the wheel load action point to cover the deflection basin range. At the key interfaces of each structural layer, multiple interlayer stress measurement point positions are determined to obtain a set of multi-physics field measurement point positions.

[0047] Then, based on the multiphysics measurement point location set, the normalized root mean square error of the deflection response at multiple deflection measurement points is calculated for the road finite element model containing the equivalent ellipsoid and the reference finite element model containing the original irregular defects. The specific calculation formula is as follows: in This represents the normalized root mean square error of the deflection response. This indicates the number of deflection measurement points, which is usually 15 to 25, depending on the spatial resolution requirements of the deflection basin. The finite element model of a road containing an equivalent ellipsoid is shown in the first... Deflection values ​​at each deflection measuring point; The reference finite element model containing the original irregular defects is shown in the first... Deflection values ​​at each deflection measuring point; Indicates the deflection measuring point number. .

[0048] The normalized root mean square error of interlayer shear stress at multiple interlayer stress measurement points is specifically formulated as follows: in This represents the normalized root mean square error of interlayer shear stress. This indicates the number of interlayer stress measurement points, which is usually taken as 8 to 12, depending on the number of structural layers and interfaces. The finite element model of a road containing an equivalent ellipsoid is shown in the first... Interlayer shear stress at each interlayer stress measurement point; The reference finite element model containing the original irregular defects is shown in the first... Interlayer shear stress at each interlayer stress measurement point; Indicates the sequence number of the interlayer stress measurement point. .

[0049] Normalized root mean square error of horizontal tensile stress at the bottom of each structural layer: in This represents the normalized root mean square error of horizontal tensile stress. This indicates the number of tensile stress measurement points at the bottom of the layer, usually 6 to 10, depending on the number of bottom surfaces of the structural layers and the area of ​​stress concern. The finite element model of a road containing an equivalent ellipsoid is shown in the first... Horizontal tensile stress at each bottom tensile stress measuring point; The reference finite element model containing the original irregular defects is shown in the first... The horizontal tensile stress at each of the bottom tensile stress measurement points; k represents the sequence number of the bottom tensile stress measurement point. K.

[0050] The three normalization errors mentioned above are weighted and summed according to preset weighting coefficients to obtain the weighted comprehensive error of the multiphysics objective function calculation. The specific formula for the multiphysics objective function is as follows: in This indicates that the goal is to minimize the weighted synthesis error of the multiphysics response. The objective function is a multiphysics objective function. , , For preset weighting coefficients, The value is determined based on the importance of each response to the structural bearing capacity evaluation.

[0051] Finally, based on the modulus search boundary, the preset center position boundary, and the preset surface area approximation constraint, modulus constraints, center position constraints, and geometric shape constraints are established. The specific constraints are as follows: in , The preset boundary range representing the x-coordinate of the ellipsoid center is determined based on the lateral spatial range of the road structure; , The preset boundary range representing the y-coordinate of the ellipsoid center is determined based on the longitudinal spatial range of the road structure; , The preset boundary range of the z-coordinate of the ellipsoid center is determined based on the spatial range of the road structure depth direction. This represents the coordinates of the equivalent ellipsoid center in the global coordinate system; Represents the surface area of ​​the equivalent ellipsoid; This represents the surface area of ​​the original irregular disease. This represents the preset surface area deviation threshold, which is determined based on the acceptable degree of morphological approximation in engineering experience.

[0052] Finally, based on the dimensionality-reduced independent optimization variables, the multiphysics objective function, and the aforementioned constraints, an optimization problem is constructed, and an optimization model is obtained.

[0053] Step S50: Based on the optimization model, a preset surrogate model is used to perform global optimization iteration on the geometric parameters and the equivalent initial elastic modulus to obtain the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus.

[0054] It should be noted that the preset surrogate model can be the Kriging model, which assumes that the objective function is a Gaussian process and constructs the predicted value and prediction variance through the correlation matrix between sample points. Global optimization iteration refers to the iterative process of selecting sampling points within the design space according to the expected improvement criterion, alternately executing surrogate model prediction and actual finite element calculation, and gradually approaching the global optimal solution. The equivalent initial elastic modulus is determined by ground penetrating radar data through three steps: multi-attribute extraction, damage mapping, and volume-weighted averaging.

[0055] Further, step S50 includes: First, based on an equivalent inertial ellipsoid, synthetic irregular defects are generated in batches using a parametric random perturbation method to obtain a preset synthetic irregular defect database. Specifically, based on an equivalent inertial ellipsoid, synthetic irregular defects are generated in batches using a parametric random perturbation method. Multi-frequency three-dimensional Perlin noise is superimposed on the radial distance of each point on the surface of the reference ellipsoid. By controlling the noise amplitude parameter (usually taken as 0.05 to 0.30 times the length of the reference semi-axis) and the frequency parameter (usually taken as 1 to 8 cycles per unit length), irregular shapes with different volumes, elongations, and surface roughness are generated. The number of generated geometric feature spaces covering common road defects is saved as a three-dimensional binary volume or a closed triangular mesh to obtain a preset synthetic irregular defect database.

[0056] Secondly, the geometric parameters and the equivalent initial elastic modulus are combined into an optimization variable vector, which includes the semi-axis length, centroid coordinates, attitude Euler angles and the equivalent initial elastic modulus, to obtain the initial optimization parameters.

[0057] Next, based on the initial optimization parameters, sampling is performed according to the optimization model to obtain multiple sets of parameter samples. For each set of parameter samples, a reference finite element model containing the synthetic irregularity and an equivalent finite element model containing the equivalent ellipsoid are established. Batch deflection response calculations are performed to obtain multiple sets of objective function sample values. Specifically, based on the initial optimization parameters in the optimization model, multiple sets of parameter samples are extracted within the design space using Latin hypercube sampling or Sobol sequence sampling methods. The number of samples is 10 to 15 times the dimension of the variables, typically 110 to 165 sets. For each set of parameter samples, a synthetic irregularity is selected from the database, and two finite element models are established: Model A contains the synthetic irregularity and calculates the reference deflection value, reference inter-layer shear stress, and reference horizontal tensile stress; Model B contains the equivalent ellipsoid with parameters of the sample set and calculates the equivalent deflection value, equivalent inter-layer shear stress, and equivalent horizontal tensile stress. The corresponding objective function value is calculated according to the multiphysics objective function. This process is independent of each sample and can be executed in parallel to obtain multiple sets of objective function sample values.

[0058] Next, multiple sets of parameter samples and objective function sample values ​​are input into a pre-defined surrogate model to obtain the surrogate model's predicted values ​​and prediction variance. Specifically, a Kriging surrogate model is trained using multiple sets of parameter samples and objective function sample values. The Kriging model assumes the objective function is a Gaussian process. The predicted values ​​and prediction variance are constructed using the correlation matrix between sample points and the correlation vector between predicted points and sample points. The specific formulas for the predicted values ​​and prediction variance are as follows: in This represents the Kriging prediction value at prediction point x; The parameter representing the mean estimate of the Gaussian process; This represents the correlation vector between the predicted point x and all sample points; Represents the correlation matrix between sample points; This represents the inverse of the correlation matrix; This represents the vector of objective function values ​​at all sample points; This represents the vector of optimization variables at the prediction point; This represents the Kriging prediction variance at prediction point x; The variance estimation parameters of the Gaussian process are used to obtain the trained surrogate model.

[0059] Then, based on the surrogate model's predicted value and prediction variance, the next sampling point is determined using a preset expected improvement criterion, resulting in newly added sampling parameters. Specifically, efficient global optimization is performed on the trained surrogate model. In each iteration, the next sampling point requiring real finite element calculation is selected by maximizing the expected improvement. The expected improvement is determined jointly by the minimum objective function value among all current real calculation samples, the surrogate model's predicted value, and the prediction variance. The specific formula is as follows: in This represents the expected improvement at prediction point x; This represents the minimum objective function value among all current real computational samples; This represents the predicted value of the Kriging surrogate model at prediction point x; This represents the standard deviation of the Kriging surrogate model's predictions at prediction point x; The cumulative distribution function represents the standard normal distribution; Let represent the probability density function of the standard normal distribution. Then, select the parameter point that maximizes the desired improvement as the new sampling parameter.

[0060] The true objective function value is then calculated based on the newly added sampling parameters, and the preset surrogate model is updated. This process is repeated until the preset stopping condition is met, resulting in the target surrogate model. Specifically, the true objective function value is calculated based on the newly added sampling parameters, the new samples are added to the training set, and the surrogate model is updated. This process is repeated until the expected improvement is lower than a preset threshold or the maximum number of iterations is reached (usually 50 to 100 times). After the iteration stops, the target surrogate model is obtained.

[0061] Finally, parameter vectors and corresponding objective function values ​​of all real computational samples are extracted from the target surrogate model to obtain a real sample set. The objective function values ​​in the real sample set are sorted in ascending order to obtain a sorted sample sequence. The parameter vector with the smallest objective function value is selected from the sorted sample sequence to obtain the optimal parameter candidate. Real finite element verification calculations are performed on the optimal parameter candidate to obtain the verification objective function value. When the deviation between the verification objective function value and the surrogate model prediction value does not exceed a preset deviation threshold, the optimal parameter candidate is determined as the optimal equivalent ellipsoid parameter and the optimal equivalent elastic modulus. When the deviation between the verification objective function value and the surrogate model prediction value exceeds the preset deviation threshold, the sample corresponding to the verification objective function value is added to the real sample set and the surrogate model is retrained. Specifically, firstly, all real computational samples accumulated during the iteration process are extracted from the target surrogate model. These samples include the parameter vectors and corresponding objective function values ​​obtained by real finite element calculations in each iteration, and are summarized into a real sample set. Next, the objective function values ​​in the real sample set are sorted in ascending order to obtain a sorted sample sequence, with samples of objective function values ​​arranged from smallest to largest. The parameter vector at the beginning of the sorted sample sequence, i.e., the parameter vector with the smallest objective function value, is selected as the optimal parameter candidate. Then, a finite element model with an equivalent ellipsoid is rebuilt for the optimal parameter candidate, and a real finite element calculation is performed to obtain the verification objective function value. This verification calculation uses the same finite element solver and boundary conditions as the surrogate model training phase. Following this, the verification objective function value is compared with the surrogate model's prediction value for the optimal parameter candidate, and the relative deviation between the two is calculated. When the relative deviation does not exceed a preset deviation threshold (usually 0.05 to 0.15), it indicates that the surrogate model's prediction accuracy near the optimal parameter candidate meets the requirements, and the optimal parameter candidate is determined as the optimal equivalent ellipsoid parameter and the optimal equivalent elastic modulus. Finally, when the relative deviation exceeds the preset deviation threshold, it indicates that the surrogate model has a large approximation error in this region. The samples corresponding to the verification objective function value are added to the real sample set and the Kriging surrogate model is retrained. The parameter vector with the smallest objective function value is selected from the updated real sample set as the new optimal parameter candidate. The verification is repeated until the relative deviation meets the preset deviation threshold.

[0062] Step S60: Construct a finite element model of a road containing equivalent regularized defects based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, and perform mechanical solution to obtain the evaluation results of the road structure mechanical response.

[0063] It should be noted that the road structure mechanical response evaluation results refer to a mechanical response dataset that includes the pavement deflection basin curve, the distribution of horizontal tensile stress at the bottom of the layer, the distribution of interlayer shear stress, and the distribution of pavement surface strain energy density, which is used for road structure bearing capacity assessment or remaining life prediction.

[0064] Specifically, a regular ellipsoid geometry is constructed based on the optimal equivalent ellipsoid parameters and embedded into a layered road structure finite element model containing the surface layer, base layer, subbase layer, and subgrade. The model is meshed using structured hexahedral or tetrahedral meshes, with appropriate densification in the ellipsoidal region to meet computational accuracy. The optimal equivalent elastic modulus is assigned to the internal mesh elements of the ellipsoid, and the healthy material parameters of each structural layer are assigned to the corresponding layer mesh elements. Standard vehicle loads are applied to the model surface at the specified loading positions according to evaluation requirements. The finite element equations are solved, outputting the pavement surface displacement field, the bottom layer horizontal tensile stress field, the interlayer shear stress field, and the pavement surface strain energy density field. The pavement deflection basin curve is extracted from the displacement field, and the bottom layer horizontal tensile stress distribution, key interface shear stress distribution, and pavement surface strain energy density distribution are extracted from the stress field and strain energy density field, forming the road structure mechanical response evaluation results. By embedding the optimized optimal equivalent regularized defects into the standard finite element model, a direct conversion from ground-penetrating radar data to road structure mechanical response evaluation is achieved, providing a quantitative basis for road structure bearing capacity assessment and remaining life prediction.

[0065] This embodiment segmentes road defects using multi-attribute data from 3D ground-penetrating radar. Based on the geometric features of the defects, an equivalent inertial ellipsoid is constructed as the initial geometric model. Modulus constraints are obtained by integrating radar multi-medium attribute mapping damage factors. A multi-constraint, multi-physics field weighted error optimization model is built. The optimal equivalent ellipsoid and modulus are obtained through global iterative optimization using a surrogate model, followed by finite element mechanical evaluation. This approach preserves the macroscopic mechanical geometry and stiffness characteristics of the actual defects, improving the accuracy of the equivalent simulation, while avoiding the challenges of meshing irregular models. The surrogate model significantly reduces the amount of iterative computation, making it suitable for batch road defect bearing capacity testing.

[0066] In the second embodiment, based on the first embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 The method for regularizing the irregular morphology of hidden road defects, step S60, further includes steps S201 to S206: Step S201: Construct an equivalent regularized ellipsoid geometry based on the optimal equivalent ellipsoid parameters, and assign internal material properties to the equivalent regularized ellipsoid geometry based on the optimal equivalent elastic modulus to obtain an equivalent regularized disease body.

[0067] Specifically, the optimal equivalent ellipsoid parameters, including the major semi-axis, middle semi-axis, minor semi-axis, centroid coordinates, and attitude Euler angles, are read. A triaxial ellipsoid geometry is constructed in the finite element preprocessing software, for example, an ellipsoid with a major axis of 1.2m and a minor axis of 0.6m, to simulate the main extension direction of irregular cavities. This constructed regular geometry can be directly recognized by the finite element software without complex geometry cleanup operations. Then, the optimal equivalent elastic modulus is assigned to the entire internal region of the ellipsoid. For example, an equivalent defect with a modulus of 800MPa represents a region where the material stiffness has severely degraded, while the structural layers outside the ellipsoid retain their original healthy material parameters, forming a clear material interface. Finally, the constructed equivalent regularized defect is output as a geometric and material property file recognizable by the finite element model, resulting in an equivalent regularized defect, which serves as the basic unit for subsequent embedding into the overall road structure model. This is done because the regular ellipsoid shape can be directly meshed by the finite element software, avoiding mesh distortion or meshing failure caused by irregular curved surfaces.

[0068] Step S202: Embed the equivalent regularized disease body into the layered road structure finite element model, and assign corresponding elastic modulus, Poisson's ratio and density parameters to each structural layer to obtain a road finite element model containing equivalent regularized disease.

[0069] It should be noted that the geometric framework for establishing a layered road structure is constructed by building the boundaries of each layer from bottom to top or from top to bottom, following the order of surface layer, base layer, subbase layer, and subgrade. For example, the surface layer thickness is 0.18m, the base layer thickness is 0.36m, the subbase layer thickness is 0.20m, and the subgrade depth is 5m to eliminate boundary effects. The thickness of each layer is determined based on actual road design parameters or test data. Then, an equivalent regularized defect is embedded into the target layer. For example, an ellipsoidal defect with a modulus of 800MPa is embedded in the middle of the base layer. The centroid coordinates are located according to the optimization results to ensure that its spatial position is consistent with the original irregular defect. This is because the spatial position of the defect directly affects the stress concentration area and settlement distribution, and it must be kept the same as the actual defect to ensure the comparability of the mechanical response. Secondly, corresponding material parameters were assigned to each structural layer. The surface layer had an elastic modulus of 12000 MPa, Poisson's ratio of 0.30, and density of 2400 kg / m³; the base layer had an elastic modulus of 5000 MPa, Poisson's ratio of 0.25, and density of 2200 kg / m³; the subbase layer had an elastic modulus of 2500 MPa, Poisson's ratio of 0.30, and density of 2000 kg / m³; and the subgrade had an elastic modulus of 100 MPa, Poisson's ratio of 0.35, and density of 1800 kg / m³. These parameter values ​​were based on field sampling test data. For damaged areas, the optimal equivalent elastic modulus was used to replace the healthy material parameters of the original layer at that location. Finally, the integrity of the model after assigning material parameters was checked to ensure that the interfaces between each layer were continuous and that there were no missing or conflicting parameters, resulting in a finite element model of the road containing equivalent regularized defects. This was done because the stiffness differences between the layers of the layered structure determine the load transfer path and stress distribution characteristics, and accurate material parameter assignment is a prerequisite for obtaining reliable mechanical response results.

[0070] Step S203: Mesh the finite element model of the road containing equivalent regularized defects, and refine the mesh in the equivalent regularized defect area to obtain the finite element mesh model.

[0071] It should be noted that, firstly, the finite element model of the road containing equivalent regularized defects is meshed globally using structured hexahedral or tetrahedral meshes. A coarser mesh with a side length of 0.40m is used for the subgrade and base course, which are far from the defect area. In these areas, the stress gradient is small, and the coarse mesh is sufficient to meet the accuracy requirements and reduce computational workload. Then, local mesh refinement is performed in the equivalent regularized defect area. The mesh size near the ellipsoid boundary is refined to 0.08m, and the mesh size inside the ellipsoid is further refined to 0.05m. Stress concentration and strain gradient are significant near the defect boundary, requiring a denser mesh to accurately capture stress singularities and displacement abrupt changes. Secondly, a mesh transition zone is set up, establishing a gradual transition in mesh size between the refined and unrefined areas, from 0.08m to 0.40m. The length of the transition zone is typically 2 to 3 times the side length of the coarse mesh to avoid abrupt changes in mesh size that could lead to loss of computational accuracy or convergence difficulties. Finally, check the mesh quality indicators, including aspect ratio, distortion, and Jacobian determinant. Remove substandard elements and re-disassemble until the preset quality threshold is met to obtain the finite element mesh model. Low-quality meshes will introduce numerical errors and may even cause the finite element solution to diverge. A high-quality discretization model is the basic guarantee for obtaining reliable mechanical response results.

[0072] Step S204: Apply standard vehicle load to the surface of the finite element mesh model and set the loading position according to the evaluation requirements to obtain the loaded finite element model.

[0073] Specifically, firstly, a standard vehicle load is applied to the surface of the finite element mesh model. The standard vehicle load adopts a double-circle uniformly distributed load, with an axle load typically of 100kN, a wheel pressure typically of 0.7MPa, and a ground contact radius typically of 0.1065m. These parameters are taken according to the standard axle load BZZ-100 specified in the "Specifications for Design of Highway Asphalt Pavement". The distance between the two centers of the double-circle load is typically 1.5 times the ground contact diameter, i.e., 319.5mm. The reason for this is that the double-circle uniformly distributed load can better simulate the contact pressure distribution between the actual vehicle tire and the road surface, which is more in line with engineering practice than the single-circle load. Secondly, the loading position is set according to the evaluation requirements. The loading position is usually located directly above the equivalent regularized disease body to simulate the most unfavorable load condition, or offset by a certain distance to simulate the scenario where the load acts on the edge of the disease or far away from the disease area. Typical offset distances are usually 0m, 0.5m, 1.0m, and 1.5m. This is because the stress concentration and deflection distribution characteristics of the pavement differ significantly under different loading positions. By loading at multiple positions, the impact range of the disease on the structural performance can be comprehensively evaluated. Finally, the load boundary conditions and model constraint conditions (usually fixed constraints on the soil base surface and horizontal constraints on the sides) are combined to obtain the finite element model after loading.

[0074] Step S205: Perform mechanical solution on the loaded finite element model to obtain multiphysics mechanical response data.

[0075] It should be noted that the multiphysics mechanical response data includes the deflection basin curve of the pavement surface, the horizontal tensile stress at the bottom of each structural layer, and the shear stress distribution at key interfaces.

[0076] Specifically, firstly, a mechanical solution is performed on the loaded finite element model. The structural equilibrium equations are solved using the finite element method to obtain the surface displacement field, the bottom layer horizontal tensile stress field, and the interlayer shear stress field. During the solution process, the Newton-Raphson iteration method is used to handle material nonlinearity, and the convergence tolerance is typically taken as... to The reason for this is that there is a significant stiffness difference between the equivalent damaged material and the healthy material, which is a heterogeneous problem requiring iterative solution to ensure equilibrium accuracy. Then, deflection basin curves are extracted from the pavement surface displacement field. Measuring points are arranged longitudinally and laterally along the center of wheel load action, and the vertical displacement values ​​of each measuring point are recorded to form deflection basin curves. These curves reflect the overall deformation characteristics of the pavement under load and are a direct indicator for evaluating the structural bearing capacity. Next, the distribution of horizontal tensile stress at the bottom of each structural layer is extracted from the horizontal tensile stress field at the bottom of each layer. The location and value of the maximum tensile stress are of particular interest. The horizontal tensile stress at the bottom of each layer is a key mechanical indicator controlling fatigue cracking in asphalt surface layers and semi-rigid base layers. When the stress level exceeds the material's fatigue strength, the structure will develop bottom-up cracks. Finally, the distribution of shear stress at key interfaces is extracted from the interlayer shear stress field. Key interfaces include the surface layer-base layer interface and the base layer-subbase layer interface. Excessive interlayer shear stress will lead to interlayer slippage or voids, affecting the effective transfer of load. Finally, the deflection basin curve, the horizontal tensile stress distribution at the bottom of the layer, and the shear stress distribution at the key interface are correlated and stored to obtain multi-physics mechanical response data.

[0077] Step S206: Based on the multi-physics mechanical response data, assess the bearing capacity of the road structure or predict its remaining life to obtain the road structure mechanical response evaluation results.

[0078] Specifically, firstly, the deflection basin curve of the pavement surface is compared with the preset allowable deflection value. The preset allowable deflection value is determined according to the road grade and design specifications. For example, it is usually 0.5mm to 1.0mm for asphalt pavement of highways. If the measured maximum deflection value at the center of the deflection basin exceeds the allowable value, the structural bearing capacity is deemed insufficient. This is because the deflection value directly reflects the overall stiffness of the pavement under load, and excessive deflection indicates that the overall bearing capacity of the structure has been reduced. Then, the horizontal tensile stress at the bottom of each structural layer is compared with the preset fatigue strength. The fatigue strength of the asphalt surface layer is usually 1.5MPa to 3.0MPa, and the fatigue strength of the semi-rigid base layer is usually 0.5MPa to 1.0MPa. If the maximum horizontal tensile stress at the bottom of the layer exceeds the corresponding fatigue strength, the layer is deemed to have a risk of fatigue cracking. The horizontal tensile stress at the bottom of the layer is a key indicator for controlling the bottom-up crack propagation of asphalt pavement and semi-rigid base layer. Secondly, the shear stress at key interfaces is compared with the preset shear strength. The interlayer shear strength is typically taken as 0.3 MPa to 0.6 MPa. If the interlayer shear stress exceeds the shear strength, it is determined that there is a risk of interlayer slippage or voiding. Excessive interlayer shear stress will lead to interruption of the load transfer path and accelerate pavement damage. Based on this, the current load-bearing state of the structure is graded by comparing the results of three indicators: deflection, tensile stress, and shear stress. For example, it is classified into four levels: good, moderate, poor, and dangerous. Finally, the remaining life is calculated based on the stress level and the material fatigue equation. The fatigue equation usually adopts Miner's linear cumulative damage theory or Paris crack propagation formula, which correlates the number of cyclic loads with the material fatigue life to obtain the evaluation results of the road structure's mechanical response.

[0079] This embodiment achieves a direct transformation from ground-penetrating radar data to road structure mechanical response evaluation by embedding the optimal equivalent regularized defects into a standard finite element model. The equivalent regularized shape facilitates mesh generation and solution, avoiding computational difficulties caused by irregular geometry. By locally refining the mesh in the defect area, the overall computational scale is controlled while ensuring computational accuracy. Through the comprehensive extraction of multi-physics mechanical response data, a comprehensive quantitative basis is provided for road structure bearing capacity assessment and remaining life prediction, improving the reliability and engineering applicability of road structure performance evaluation results. The overall technical solution achieves equivalent approximation of the mechanical response from irregular defects to a regularized model, reduces the geometric cleanup workload of finite element modeling, and improves mesh generation efficiency and solution convergence.

[0080] Based on the first embodiment of this application, this application also provides a device for regularizing the irregular shape of hidden road defects. Please refer to... Figure 3 The device includes: The extraction module is used to collect three-dimensional ground-penetrating radar data of the target road area and perform preprocessing. It extracts amplitude attribute information and multi-dimensional medium attribute information from the preprocessed three-dimensional data volume, segments the disease area according to the amplitude attribute information, and obtains irregular disease binary volume and multi-attribute distribution data.

[0081] The geometric equivalent module is used to calculate the volume, centroid, and moment of inertia tensor based on the binary volume of the irregular disease. It performs eigenvalue decomposition on the moment of inertia tensor to obtain the direction of the principal axis of inertia. Based on the volume, centroid, and direction of the principal axis of inertia, it constructs an equivalent inertial ellipsoid to obtain the geometric equivalent initial solution.

[0082] The material mapping module is used to input multi-attribute distribution data into a preset nonlinear mapping model for multi-attribute fusion mapping to obtain the spatial distribution of damage factors, and generate prior constraints for material parameters based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials.

[0083] The module is used to construct an optimization model based on the prior constraints of geometric and material parameters in the geometrically equivalent initial solution.

[0084] The optimization module is used to perform global optimization iteration on geometric parameters and equivalent initial elastic modulus based on the optimization model and a preset surrogate model, so as to obtain the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus.

[0085] The results module is used to construct a finite element model of a road containing equivalent regularized defects based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, and to perform mechanical solutions to obtain the evaluation results of the road structure's mechanical response.

[0086] The road hidden defect irregularity regularization device provided in this application, employing the road hidden defect irregularity regularization method in the above embodiments, can solve the technical problem of how to achieve synergistic optimization of defect geometry and damage stiffness, thereby reducing the computational cost of batch road bearing capacity evaluation. Compared with the prior art, the beneficial effects of the road hidden defect irregularity regularization device provided in this application are the same as those of the road hidden defect irregularity regularization method provided in the above embodiments, and other technical features in the road hidden defect irregularity regularization device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0087] This application provides a road hidden defect irregular shape regularization device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the road hidden defect irregular shape regularization method in the above embodiment 1.

[0088] The following is for reference. Figure 4 This document illustrates a structural schematic diagram of a road concealed defects irregularity regularization device suitable for implementing embodiments of this application. The road concealed defects irregularity regularization device in embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), and vehicle terminals (e.g., vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 4 The road hidden defects irregular shape regularization device shown is merely an example and should not impose any limitation on the function and scope of use of the embodiments of this application.

[0089] like Figure 4 As shown, a road hidden defect irregularity regularization device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to a program stored in read-only memory (ROM) or a program loaded from a storage device into random access memory (RAM). The RAM also stores various programs and data required for the operation of the road hidden defect irregularity regularization device. The processing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus. Typically, the following can be connected to the I / O interface: input devices including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices including, for example, magnetic tapes, hard disks, etc.; and communication devices. The communication device allows the road hidden defect irregularity regularization device to communicate wirelessly or wiredly with other devices to exchange data. Although the diagrams show various road hidden defects and irregular morphological regularization devices, it should be understood that implementation or possession of all of them is not required. More or fewer may be implemented alternatively.

[0090] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0091] The road hidden defect irregularity regularization device provided in this application, employing the road hidden defect irregularity regularization method described in the above embodiments, can solve the technical problem of how to achieve synergistic optimization of defect geometry and damage stiffness, thereby reducing the computational cost of batch road bearing capacity evaluation. Compared with the prior art, the beneficial effects of the road hidden defect irregularity regularization device provided in this application are the same as those of the road hidden defect irregularity regularization method provided in the above embodiments, and other technical features in this road hidden defect irregularity regularization device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0092] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0093] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0094] This application provides a computer-readable medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the method for regularizing the irregular morphology of hidden road defects in the above embodiments.

[0095] The computer-readable medium provided in this application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor devices, or any combination thereof. More specific examples of computer-readable media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable medium may be any tangible medium containing or storing a program that can be executed by instructions, used by a device, or used in conjunction with it. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0096] The aforementioned computer-readable medium may be included in the road hidden defects irregular shape regularization device; or it may exist independently and not be installed in the road hidden defects irregular shape regularization device.

[0097] The aforementioned computer-readable medium carries one or more programs that, when executed by the road hidden disease irregularity regularization device, enable the road hidden disease irregularity regularization device to write computer program code for performing the operations of this application in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—and conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0098] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, all blocks in the flowcharts or block diagrams may represent a module, segment, or portion of code containing one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that all blocks in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using dedicated hardware-based implementations that perform the specified functions or operations, or using a combination of dedicated hardware and computer instructions.

[0099] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0100] The readable medium provided in this application is a computer-readable medium, which stores computer-readable program instructions (i.e., a computer program) for executing the above-described method for regularizing the irregular morphology of hidden road defects. This solves the technical problem of how to achieve synergistic optimization of defect geometry and damage stiffness, thereby reducing the computational cost of batch road bearing capacity evaluation. Compared with the prior art, the beneficial effects of the computer-readable medium provided in this application are the same as those of the method for regularizing the irregular morphology of hidden road defects provided in the above embodiments, and will not be elaborated upon here.

[0101] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method for regularizing the irregular morphology of hidden road defects.

[0102] The computer program product provided in this application can solve the technical problem of how to achieve synergistic optimization of road defect geometry and damage stiffness, thereby reducing the computational cost of batch road bearing capacity evaluation. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the road hidden defect irregularity regularization method provided in the above embodiments, and will not be repeated here.

[0103] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for regularizing the irregular morphology of hidden road defects, characterized in that, The method includes: Three-dimensional ground-penetrating radar data of the target road area is collected and preprocessed. Amplitude attribute information and multi-dimensional medium attribute information are extracted from the preprocessed three-dimensional data volume. The disease area is segmented according to the amplitude attribute information to obtain an irregular disease binary volume and multi-attribute distribution data. The volume, centroid, and moment of inertia tensor are calculated based on the irregular disease binary body. The principal axis directions are obtained by eigenvalue decomposition of the moment of inertia tensor. An equivalent inertial ellipsoid is constructed based on the volume, centroid, and principal axis directions to obtain a geometrically equivalent initial solution. The multi-attribute distribution data is input into a preset nonlinear mapping model for multi-attribute fusion mapping to obtain the spatial distribution of damage factors, and material parameter prior constraints are generated based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials. An optimization model is constructed based on the prior constraints of geometric and material parameters in the geometrically equivalent initial solution. Based on the optimization model, a preset surrogate model is used to perform global optimization iteration on the geometric parameters and the equivalent initial elastic modulus to obtain the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus. Based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, a finite element model of a road containing equivalent regularized defects is constructed and mechanical solutions are performed to obtain the evaluation results of the mechanical response of the road structure. The step of constructing an optimization model based on the prior constraints of geometric and material parameters in the geometrically equivalent initial solution includes: The multiple semi-axis lengths, centroid coordinates, and attitude Euler angles in the geometrically equivalent initial solution are combined with the equivalent initial elastic modulus to form an optimization variable vector; Based on the volume, establish the volume identity constraint relationship, and express one of the multiple semi-axis lengths as a function of the remaining semi-axis length and volume to obtain the independent optimization variables after dimensionality reduction; Based on the preset measurement point layout scheme, multiple deflection measurement point locations are determined on the road surface, and multiple interlayer stress measurement point locations are determined at the key interfaces of each structural layer, thus obtaining a set of multi-physics field measurement point locations. Based on the set of multiphysics measurement points, the weighted comprehensive error of the normalized root mean square error of the deflection response at multiple deflection measurement points, the normalized root mean square error of the interlayer shear stress at multiple interlayer stress measurement points, and the normalized root mean square error of the horizontal tensile stress at the bottom of each structural layer are defined for the road finite element model containing the equivalent ellipsoid and the reference finite element model containing the original irregular defects. The objective function of the multiphysics field is obtained. Based on the modulus search boundary, the preset center position boundary, and the preset surface area approximation constraint, modulus constraints, center position constraints, and geometric shape constraints are established. Based on the reduced-dimensional independent optimization variables, the multiphysics objective function, and the constraints, an optimization problem is constructed to obtain an optimization model.

2. The method as described in claim 1, characterized in that, The steps of collecting and preprocessing 3D ground-penetrating radar data of the target road area, extracting amplitude attribute information and multi-dimensional medium attribute information from the preprocessed 3D data volume, segmenting the disease area based on the amplitude attribute information, and obtaining irregular disease binary volume and multi-attribute distribution data include: The target road area was scanned using a multi-channel three-dimensional ground-penetrating radar system to obtain raw scan data; Zero-bias correction and background denoising are performed sequentially on the original scan data to obtain corrected data; The corrected data is subjected to bandpass filtering and offset imaging processing to obtain a three-dimensional data volume; The amplitude attribute volume is extracted from the three-dimensional data volume to obtain the amplitude attribute information; The center frequency attenuation attribute volume, instantaneous phase continuity attribute volume, and amplitude statistical variance attribute volume are extracted from the three-dimensional data volume to obtain multi-dimensional medium attribute information; Based on the amplitude attribute information, threshold segmentation and three-dimensional region growth are performed to obtain an irregular disease binary body; The data in the multi-dimensional medium attribute information and the corresponding regions of the irregular disease binary body are associated and stored to obtain multi-attribute distribution data.

3. The method as described in claim 1, characterized in that, The steps of calculating the volume, centroid, and moment of inertia tensor based on the irregular disease binary body, performing eigenvalue decomposition on the moment of inertia tensor to obtain the principal axis directions, and constructing an equivalent inertial ellipsoid based on the volume, centroid, and principal axis directions to obtain a geometrically equivalent initial solution include: Based on the coordinates and state values ​​of each voxel in the irregular disease binary volume, voxel integration is performed to obtain the volume and centroid. Establish a relative coordinate system based on the centroid, calculate the moment of inertia tensor of the irregular disease binary body relative to the centroid, and obtain the moment of inertia tensor. The moment of inertia tensor is subjected to eigenvalue decomposition to obtain multiple eigenvalues ​​and corresponding multiple unit eigenvectors; The direction of the principal axis of inertia is determined based on the multiple unit eigenvectors, and the corresponding Euler angles are obtained; Based on the volume and the direction of the principal inertial axis, the lengths of multiple semi-axes of the equivalent inertial ellipsoid are solved simultaneously to obtain the set of semi-axe lengths; Based on the set of semi-axis lengths, the centroid, and the Euler angles, a parameter vector is constructed to obtain the geometrically equivalent initial solution.

4. The method as described in claim 1, characterized in that, The step of inputting the multi-attribute distribution data into a preset nonlinear mapping model for multi-attribute fusion mapping to obtain the spatial distribution of damage factors, and generating prior constraints on material parameters based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials, includes: The center frequency decay attribute, instantaneous phase continuity attribute, amplitude statistical variance attribute, and dielectric constant dispersion gradient attribute are extracted from the multi-attribute distribution data to obtain a multi-attribute vector; The multi-attribute vector is input into a preset nonlinear mapping model to perform multi-attribute fusion mapping, thereby obtaining the spatial distribution of damage factors. Based on the spatial distribution of the damage factors and the preset elastic modulus of healthy materials, the equivalent elastic modulus of the diseased area is calculated by volume weighted average to obtain the equivalent initial elastic modulus. The lower limit of the modulus is calculated based on the preset modulus attenuation coefficient and the preset elastic modulus of the healthy material, and the upper limit of the modulus is taken as the preset elastic modulus of the healthy material to obtain the modulus search boundary. The equivalent initial elastic modulus and the modulus search boundary are associated and stored to obtain the prior constraints of material parameters.

5. The method as described in claim 1, characterized in that, The step of performing global optimization iterations on the geometric parameters and equivalent initial elastic modulus using a preset surrogate model based on the optimization model to obtain the optimal equivalent ellipsoidal parameters and optimal equivalent elastic modulus includes: Based on the equivalent inertial ellipsoid, synthetic irregular diseases are generated in batches through a parameterized random perturbation method, resulting in a pre-defined synthetic irregular disease database. The geometric parameters and the equivalent initial elastic modulus are combined into an optimization variable vector to obtain the initial optimization parameters; Based on the initial optimization parameters, sampling is performed according to the optimization model to obtain multiple sets of parameter samples. For each set of parameter samples, a reference finite element model containing synthetic irregular defects and an equivalent finite element model containing an equivalent ellipsoid are established respectively. Batch deflection response calculation is performed to obtain multiple sets of objective function sample values. The multiple sets of parameter samples and the multiple sets of objective function sample values ​​are input into the preset proxy model to obtain the proxy model's predicted value and prediction variance. Based on the predicted value and prediction variance of the surrogate model, the next sampling point is determined by a preset expected improvement criterion, and the new sampling parameters are obtained. The true objective function value is calculated based on the newly added sampling parameters and the preset proxy model is updated. This process is repeated until the preset stopping condition is met to obtain the target proxy model. The parameter vectors and corresponding objective function values ​​of all real computational samples are extracted from the target proxy model to obtain the real sample set; The objective function values ​​in the real sample set are sorted in ascending order to obtain the sorted sample sequence. The parameter vector with the smallest objective function value is selected from the sorted sample sequence to obtain the optimal parameter candidates; Perform real finite element verification calculations on the optimal parameter candidates to obtain the verification objective function values; When the deviation between the verification objective function value and the surrogate model prediction value does not exceed the preset deviation threshold, the optimal parameter candidate is determined as the optimal equivalent ellipsoid parameter and the optimal equivalent elastic modulus; When the deviation between the verification objective function value and the surrogate model prediction value exceeds a preset deviation threshold, the sample corresponding to the verification objective function value is added to the real sample set and the surrogate model is retrained.

6. The method as described in claim 1, characterized in that, The steps of constructing a road finite element model containing equivalent regularized defects based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, and performing mechanical solutions to obtain the road structural mechanical response evaluation results include: Based on the optimal equivalent ellipsoid parameters, an equivalent regularized ellipsoid geometry is constructed, and based on the optimal equivalent elastic modulus, the internal material properties of the equivalent regularized ellipsoid geometry are assigned to obtain an equivalent regularized disease body. The equivalent regularized disease body is embedded into the layered road structure finite element model, and corresponding elastic modulus, Poisson's ratio and density parameters are assigned to each structural layer to obtain a road finite element model containing equivalent regularized disease. The road finite element model containing equivalent regularized defects is meshed, and the mesh is refined in the equivalent regularized defect area to obtain a finite element mesh model. A standard vehicle load is applied to the surface of the finite element mesh model, and the loading position is set according to the evaluation requirements to obtain the loaded finite element model. The finite element model after loading is subjected to mechanical solution to obtain multi-physics mechanical response data, which includes the deflection basin curve of the road surface, the horizontal tensile stress at the bottom of each structural layer, and the shear stress distribution at key interfaces. Based on the multi-physics mechanical response data, the bearing capacity of the road structure is assessed or the remaining life is predicted, and the mechanical response evaluation results of the road structure are obtained.

7. A device for regularizing the irregular shapes of hidden road defects, characterized in that, The device is applied to the method for regularizing the irregular morphology of hidden road defects as described in any one of claims 1-6, and the device comprises: The extraction module is used to collect three-dimensional ground-penetrating radar data of the target road area and perform preprocessing. It extracts amplitude attribute information and multi-dimensional medium attribute information from the preprocessed three-dimensional data volume, and segments the disease area according to the amplitude attribute information to obtain irregular disease binary volume and multi-attribute distribution data. The geometric equivalent module is used to calculate the volume, centroid, and moment of inertia tensor based on the irregular disease binary body, perform eigenvalue decomposition on the moment of inertia tensor to obtain the direction of the principal axis of inertia, construct an equivalent inertial ellipsoid based on the volume, centroid, and direction of the principal axis of inertia, and obtain the geometric equivalent initial solution. The material mapping module is used to input the multi-attribute distribution data into a preset nonlinear mapping model to perform multi-attribute fusion mapping, obtain the spatial distribution of damage factors, and generate prior constraints on material parameters based on the spatial distribution of damage factors and the preset elastic modulus of healthy materials. The construction module is used to construct an optimization model based on the prior constraints of geometric parameters and material parameters in the geometrically equivalent initial solution; The optimization module is used to perform global optimization iterations based on the optimization model through a preset proxy model to obtain the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus. The results module is used to construct a finite element model of a road containing equivalent regularized defects based on the optimal equivalent ellipsoid parameters and the optimal equivalent elastic modulus, and to perform mechanical solution to obtain the evaluation results of the mechanical response of the road structure.

8. A device for regularizing the irregular shapes of hidden road defects, characterized in that, The device includes: a memory, a processor, and a road hidden disease irregularity regularization program stored in the memory and running on the processor, the road hidden disease irregularity regularization program being configured to implement the steps of the road hidden disease irregularity regularization method as described in any one of claims 1-6.

9. A storage medium, characterized in that, The storage medium stores a road hidden defects irregular morphology regularization program, which, when executed by a processor, implements the steps of the road hidden defects irregular morphology regularization method as described in any one of claims 1-6.

Images