A highway pavement quality detection method and system

Abstract

The application discloses a highway pavement quality detection method and system, relates to the technical field of highway material quality detection, and obtains original detection data and environmental parameters of a to-be-detected pavement; a deep learning model is used to analyze the original detection data, pavement disease characteristics are obtained and quantitatively calculated, structural state parameters are obtained, the original detection data are calculated based on a modified variable separation method and a temperature dielectric hybrid model, and pavement asphalt state data are obtained; a cracking probability model based on thermodynamic response is used to dynamically correct the structural state parameters and the environmental parameters, and fatigue damage indexes are calculated; a decision feature vector is constructed, feature mapping of the decision feature vector is performed through a nonlinear fusion decision model, a road health index is output, and a road health grade is determined, so that the efficiency and objectivity of road detection are improved, and accurate maintenance of the highway pavement is realized.

Description

Technical Field

[0001] This invention relates to the field of highway material testing technology, and in particular to a method and system for testing highway pavement quality. Background Technology

[0002] As a core component of transportation infrastructure, the health of highway pavement directly determines driving safety, traffic efficiency, and the total life-cycle cost of the road network, creating an urgent need for more refined, intelligent, and integrated detection technologies.

[0003] Traditional pavement quality inspection mainly relies on manual inspections and core sampling, which suffers from drawbacks such as low efficiency, strong subjectivity, and difficulty in continuous coverage of destructive testing. Furthermore, it is easily affected by the experience level of the inspectors, resulting in high rates of missed detections and data deviations. Although non-destructive testing equipment such as falling weight deflectometers and ground-penetrating radar are gradually being used, the industry generally faces the dilemma of data fragmentation: ground-penetrating radar excels at identifying internal defects but struggles to quantify structural mechanical properties; deflectometers can characterize structural strength but cannot pinpoint the spatial distribution of defects; and while new dielectric testing technologies such as capacitance sensing can reflect the state of asphalt mixtures, they lack integration with mechanical and electrical properties. The lack of systematic fusion methods for thermodynamic parameters and the absence of correlation models between apparent defects, internal structure, material state, and environmental factors make it difficult for test results to comprehensively characterize the true health status of pavements. Defect assessments rely heavily on static indicators, failing to consider the coupled effects of dynamic temperature field changes on asphalt modulus and fatigue damage. The prediction accuracy of fatigue life and cracking probability is insufficient, making it unable to support preventive maintenance decisions. The evaluation system has a low degree of standardization, making it difficult to interoperate indicators from different testing methods. Furthermore, traditional fusion models struggle to capture nonlinear correlations between features, and the quantification and grading of health indices lack precise technical support. Summary of the Invention

[0004] The technical problem solved by this invention is that existing technologies are difficult to accurately invert the state of asphalt mixtures and do not consider the dynamic impact of dynamic changes in the temperature field on pavement fatigue damage.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] Firstly, a method for testing the quality of highway pavement includes the following steps:

[0007] Step S1: Obtain the original detection data and environmental parameters of the road surface to be inspected;

[0008] Step S2: Analyze the original detection data using a deep learning model to obtain pavement distress characteristics and perform quantitative calculations to obtain structural state parameters. Calculate the original detection data based on the modified variable separation method and the temperature-dielectric hybrid model to obtain pavement asphalt state data.

[0009] Step S3: Based on the cracking probability model of thermodynamic response, the structural state parameters and environmental parameters are dynamically corrected to calculate the fatigue damage index.

[0010] Step S4: Standardize the structural state parameters, fatigue damage index, pavement asphalt state data and pavement distress characteristics, construct a decision feature vector, perform feature mapping on the decision feature vector through a nonlinear fusion decision model, and output the road health index.

[0011] Step S5: Determine the road health level based on the road health index.

[0012] Preferably, step S1 includes the following sub-steps:

[0013] Step S11: Use three-dimensional ground-penetrating radar to continuously scan the road section to be detected to obtain the original radar echo image data.

[0014] Step S12: Set key points for each unit road segment of the road section to be tested, deploy displacement sensors at each key point, use a falling weight deflectometer for fixed-point detection, and synchronously collect the road surface deflection response through the displacement sensors to obtain the original sequence data of the deflection basin. The original sequence data of the deflection basin includes sensor number, sequence data and load parameters. Use a coplanar capacitance sensor to directly fit and detect the road surface at the key points of the road section to be tested, and use an LCR measuring instrument to record the capacitance value to obtain the coplanar capacitance density detection data.

[0015] Step S13: Obtain pavement surface defect data through the pavement historical inspection record database. The pavement surface defect data includes defect type, location, measured area, and severity.

[0016] Step S14: Obtain the environmental parameters of the area where the road section to be detected is located. The environmental parameters include hourly air temperature, daily maximum temperature, daily minimum temperature, daily total solar radiation and average wind speed. Save the pavement surface defect data, original radar echo image data, original deflection basin sequence data and pavement capacitance data as original detection data.

[0017] Preferably, step S2 includes the following sub-steps:

[0018] Step S21: Extract the original radar echo image data from the original detection data, input the original radar echo image data into the deep learning target detection model, and obtain the disease results. The disease results include disease categories and bounding box coordinates. The disease categories include cracks, interlayer defects, and looseness.

[0019] Step S22: Based on the results of the disease, the distribution area of ​​various diseases within the unit road segment is statistically analyzed. According to the pixel resolution of the ground penetrating radar image, the total actual area of ​​the disease is calculated. The road segment to be detected is cut into several sub-segments according to key points. The percentage of the total actual area of ​​the disease to the area of ​​the unit sub-segment is calculated to obtain the internal fragmentation rate.

[0020] Mechanical property parameters were extracted from the original sequence data of the deflection basin to obtain structural state parameters:

[0021] Extract the central deflection value and radial distance deflection value at the load center. The radial distance deflection value includes a first radial deflection value and a second radial deflection value. Subtract the first radial deflection value from the central deflection value to obtain a first damage index. Subtract the second radial deflection value from the central deflection value to obtain a second damage index. Save the first damage index and the second damage index as structural state parameters.

[0022] Step S23: Based on the modified variable separation method, the preprocessed coplanar capacitance density detection data is inverted to obtain the dielectric constant of the asphalt mixture;

[0023] Step S24: The dielectric constant of the asphalt mixture is converted into pavement asphalt state data using a temperature-dielectric mixing model.

[0024] Preferably, step S23 specifically includes:

[0025] The coplanar capacitance density detection data is preprocessed to obtain the measured capacitance value, and the geometric parameters of the coplanar capacitance sensor are acquired. These geometric parameters include electrode width w, electrode length L, electrode spacing g, electrode thickness t, and the number of electrodes. The dielectric constant of asphalt mixtures was calculated using the bisection method iteratively.

[0026] Randomly set the initial search lower threshold and initial search upper threshold. The initial search lower threshold is less than the initial search upper threshold and both are positive. Calculate the average of the initial search lower threshold and initial search upper threshold as the midpoint value of the search interval. Call the modified variable separation method to calculate the total capacitance value based on the current midpoint value of the search interval. The processing logic for the total capacitance value is as follows:

[0027] Using the point-matching method of the Laplace equation, the potential expansion coefficient is obtained based on the boundary conditions by searching for the midpoint value of the interval. The internal unit capacitance is then calculated by integrating the electric field over the electrode surface using Gauss's law. and external unit capacitors The total capacitance value is calculated using the total capacitance correction model, and the calculation expression for the total capacitance correction model is as follows:

[0028] ;

[0029] in, This is an electrode thickness correction term. The length is extended to be equivalent to the edge effect.

[0030] Calculate the difference between the total capacitance value and the measured capacitance value. When the difference is greater than 0, replace the current midpoint value of the search interval with the upper limit threshold. When the difference is less than 0, replace the current midpoint value of the search interval with the lower limit threshold. Iteratively calculate and output the midpoint value of the search interval for the next round, and repeat the binary search method.

[0031] When the absolute value of the difference is less than the preset accuracy threshold, the iteration stops and the midpoint value of the current search interval is output as the dielectric constant of the asphalt mixture.

[0032] When the number of iterations exceeds the preset maximum number of iterations, the midpoint value of the current search interval is used as the dielectric constant of the asphalt mixture.

[0033] Preferably, step S24 specifically includes:

[0034] Based on the modified temperature dielectric mixing model, the dielectric constant of asphalt mixture is calculated, and the density of asphalt mixture is obtained.

[0035] Obtain the maximum density of asphalt on the road to be tested, calculate the ratio of the density of the asphalt mixture to the maximum density of asphalt, and obtain the compaction degree. Use the density of the asphalt mixture and the compaction degree as the pavement asphalt state data.

[0036] Preferably, step S3 includes the following sub-steps:

[0037] Step S31: Based on environmental parameters, establish an asphalt pavement temperature field estimation model based on the heat conduction equation, calculate the temperature distribution inside the pavement as a function of depth and time, and calculate the average temperature of the road section to be tested at the time of testing and the daily temperature difference extreme value of the pavement within a unit period.

[0038] Step S32: Based on the pavement asphalt state data and average temperature, the dynamic modulus of the asphalt surface layer is corrected by temperature-density coupling using the WLF equation and density correction pump coefficient to obtain the effective dynamic modulus.

[0039] Step S33: Take the effective dynamic modulus of the current road segment and the surface layer thickness of the current road segment as a sample, and use the Kriging surrogate model to calculate the mechanical response of each sample under standard axle load; calculate the fatigue life of each sample by combining the fatigue equation; and count the proportion of samples whose fatigue life is less than the design axle load number to obtain the longitudinal cracking probability.

[0040] Step S34: Based on the Miner linear cumulative damage criterion, calculate the fatigue damage due to periodic temperature changes, obtain the temperature damage index of cumulative temperature, and save the temperature damage index and longitudinal cracking probability as fatigue damage index.

[0041] Preferably, step S33 specifically includes:

[0042] The effective dynamic modulus and surface layer thickness of the current road segment are used as a set of samples. The effective dynamic modulus and surface layer thickness of all road segments are statistically analyzed to generate several sets of samples. The Latin hypercube sampling method is used to randomly generate N sets of samples. The N sets of samples are input into the Kriging surrogate model, and the mechanical response value corresponding to each set of samples is output.

[0043] A fatigue equation is constructed based on the transfer function, and the fatigue life estimate is obtained by calculating the effective dynamic modulus and mechanical response value. The calculation expression of the fatigue equation is as follows:

[0044] ;

[0045] in, This is an estimate of fatigue life. This is the mechanical response value. For effective dynamic modulus, , and These are the regression coefficients;

[0046] Obtain the total traffic load of the road section to be tested, compare the fatigue life estimate of each sample group with the total traffic load, count the number of failure samples in N samples whose fatigue life estimate is less than the total traffic load, calculate the proportion of the number of failure samples to N samples, and obtain the longitudinal cracking probability.

[0047] Preferably, step S4 specifically includes:

[0048] The structural state parameters, fatigue damage indexes, pavement asphalt state data, and internal breakage rate are standardized and spliced ​​to obtain a decision feature vector.

[0049] A nonlinear fusion decision model based on AFG-KAN is constructed, and feature mapping regression prediction is performed on the feature vector to output the road health index.

[0050] An AFG-KAN (Adaptive Feature Gated KAN) model is established, comprising an input layer, a feature gating layer, several intermediate hidden layers, and an output layer. The standardized decision feature vector is input into the feature gating layer and subjected to feature gating enhancement processing to obtain a weighted feature vector. The weighted feature vector is then input into the KAN hidden layer, and multi-feature fusion based on B-spline nonlinear transformation is performed to output the road health index.

[0051] Preferably, step S5 specifically includes:

[0052] A threshold for classifying road health levels is set, and the road health index is compared with the threshold to determine the road level. The road health level classification threshold includes a first threshold, a second threshold, and a third threshold.

[0053] When the road health index exceeds the first threshold, the road health level is determined to be Level 1.

[0054] When the second threshold is less than the road health index and less than the first threshold, the road health level is determined to be level two.

[0055] When the third threshold is less than the road health index and less than the second threshold, the road health level is determined to be level three.

[0056] When the road health index is less than the third threshold, the road health level is determined to be level four.

[0057] Secondly, a highway pavement quality detection system includes a data acquisition module, a status extraction module, a fatigue damage assessment module, a health index quantification module, and a grade determination module.

[0058] The acquisition module is used to acquire the original detection data and environmental parameters of the road surface to be inspected;

[0059] The state extraction module is used to analyze the original detection data to obtain pavement distress characteristics, and to quantify and calculate structural state parameters. Based on the modified variable separation method and the temperature-dielectric hybrid model, the original detection data is processed to obtain pavement asphalt state data.

[0060] The fatigue damage assessment module is used to dynamically correct the structural state parameters and environmental parameters through a cracking probability model of thermodynamic response, and calculate fatigue damage index.

[0061] The health index quantification module is used to construct a decision feature vector, perform feature mapping on the decision feature vector through a nonlinear fusion decision model, and output the road health index.

[0062] The level determination module is used to determine the road health level based on the road health index.

[0063] The beneficial effects of this invention are as follows: This invention integrates three-dimensional ground-penetrating radar, falling-weight deflectometer, coplanar capacitance sensor, and surface image data. Through deep data fusion, it can comprehensively capture the entire process of pavement information from microscopic material aging and mesoscopic structural damage to macroscopic mechanical attenuation, effectively avoiding missed diagnoses and misjudgments, and significantly improving the reliability of pavement quality inspection. It automatically identifies hidden defects in radar images through a deep learning target detection model, overcoming the problems of low efficiency and strong subjectivity in manual interpretation. The internal breakage rate index proposed in this invention innovatively transforms unstructured image features into structured numerical indicators, achieving precise quantification of the degree of internal pavement damage and providing a unified standard for lateral quality comparison of different road sections. Utilizing the difference principle of deflection basin geometry, by calculating the surface layer damage index and the overall damage index, it compares the mechanical response of the asphalt surface layer with the response of the base layer. This invention achieves precise decoupling and layered evaluation of the mechanical properties of pavement structural layers. It employs a modified variable separation method, introducing electrode thickness and edge effect correction terms to eliminate measurement errors caused by sensor non-ideal factors. By introducing a temperature-dielectric hybrid model, it reveals the law of dielectric constant change with temperature from a microscopic thermodynamic perspective, achieving dynamic decoupling of temperature and density. This allows the detection system to maintain highly consistent evaluation results under different seasonal and time-of-day temperature conditions, solving the problem of traditional capacitance methods being greatly affected by ambient temperature. In the decision-making stage, this invention uses adaptive feature-gated AFG-KAN, utilizing B-spline basis functions to more accurately fit the complex nonlinear relationship between pavement performance indicators and health indices. The feature gating mechanism automatically suppresses noise interference in the detection data, intuitively displaying the contribution of each feature to the final health score. Attached Figure Description

[0064] Figure 1 This is a flowchart of the steps of a highway pavement quality testing method provided in Embodiment 1 of the present invention;

[0065] Figure 2 This is a basic flowchart of a highway pavement quality testing system provided in Embodiment 2 of the present invention. Detailed Implementation

[0066] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0067] Example 1, referring to Figure 1 A method for testing the quality of highway pavement is provided, comprising the following steps:

[0068] Step S1: Obtain the original detection data and environmental parameters of the road surface to be inspected;

[0069] Step S2: Analyze the original detection data using a deep learning model to obtain pavement distress characteristics and perform quantitative calculations to obtain structural state parameters. Calculate the original detection data based on the modified variable separation method and the temperature-dielectric hybrid model to obtain pavement asphalt state data.

[0070] Step S3: Based on the cracking probability model of thermodynamic response, the structural state parameters and environmental parameters are dynamically corrected, and the fatigue damage index is calculated.

[0071] Step S4: Standardize the structural state parameters, fatigue damage index, pavement asphalt state data and pavement distress characteristics, construct a decision feature vector, perform feature mapping on the decision feature vector through a nonlinear fusion decision model, and output the road health index.

[0072] Step S5: Determine the road health level based on the road health index.

[0073] In this embodiment, a deep learning network is used to analyze the original detection data to automatically identify latent defects. Then, a modified variable separation method is used to invert the dielectric constant of the asphalt mixture from the capacitance signal. Finally, a temperature-dielectric mixing model is used to convert this into the density and compaction of the asphalt mixture. This overcomes the limitations of single detection methods and can simultaneously identify structural damage and material defects. The modified variable separation method ensures the reliability of the inversion results. For the large-scale detection needs of asphalt pavements, the traditional variable separation method is improved by introducing an electrode thickness correction term and an equivalent extension length for edge effects. This solves the problem of large calculation errors in traditional theories at low coupling heights. In step S24, not only is the relationship between dielectric constant and density established using the ALL model, but the influence of temperature on the dielectric constant of asphalt and aggregates is also innovatively introduced, eliminating the interference of ambient temperature fluctuations on density detection and achieving seamless connection. The unified theoretical framework of contact and non-contact detection provides a theoretical basis for vehicle-mounted high-speed mobile detection. The temperature compensation mechanism significantly improves the applicability and robustness of the method in different seasons and regions. Step S33 treats the road structure parameters as random variables and combines Monte Carlo simulation and Kriging surrogate model to upgrade the traditional deterministic fatigue life prediction to probabilistic risk assessment. Step S34 separately quantifies the cumulative damage caused by periodic temperature difference, filling the gap in the existing specifications that only consider traffic load damage. It comprehensively considers the dual fatigue effects of traffic load and temperature, making the life prediction more realistic. In addition, an AFG-KAN model is constructed to capture the complex interaction between decision feature vectors, which facilitates subsequent attribution analysis and grade determination, realizes precise maintenance of highway pavement, improves the efficiency and objectivity of road detection, and reduces reliance on human experience.

[0074] Step S1 includes the following sub-steps:

[0075] Step S11: Use three-dimensional ground-penetrating radar to continuously scan the road section to be detected to obtain the original radar echo image data.

[0076] Step S12: Set key points for each unit road segment of the road section to be tested, deploy displacement sensors at each key point, use a falling weight deflectometer for fixed-point detection, and synchronously collect the road surface deflection response through the displacement sensors to obtain the original sequence data of the deflection basin. The original sequence data of the deflection basin includes sensor number, sequence data and load parameters. Use a coplanar capacitance sensor to directly fit and detect the road surface at the key points of the road section to be tested, and use an LCR measuring instrument to record the capacitance value to obtain the coplanar capacitance density detection data.

[0077] Step S13: Obtain pavement surface distress data through the pavement historical inspection record database. The pavement surface distress data includes distress type, location, measured area, and severity.

[0078] Step S14: Obtain the environmental parameters of the area where the road section to be detected is located. The environmental parameters include hourly air temperature, daily maximum temperature, daily minimum temperature, daily total solar radiation and average wind speed. Save the pavement surface defect data, original radar echo image data, original deflection basin sequence data and pavement capacitance data as the original detection data.

[0079] In this embodiment, a 3km long section of highway was selected for testing. A three-dimensional ground-penetrating radar system was used for continuous scanning. The testing vehicle traveled at a normal speed of 60km / h, collecting radar data every 0.02m to obtain continuous three-dimensional radar echo images. The raw radar echo image data included timestamps, GPS coordinates, and reflected wave amplitude information for each channel. A key testing point was set every 100m within the testing section, and a falling weight deflectometer was used for testing. Nine displacement sensors were arranged according to a standard layout at the load center and radial distance. The falling weight mass was set to 300kg, and the lifting height was adjusted to generate a 50kN impact load. The load pulse duration was 25ms. Each measuring point was repeatedly tested three times, and the average value was taken as the raw sequence data of the deflection basin at that point. The data included the sensor number, deflection value, and load peak value.

[0080] At the same key point, an interdigitated coplanar capacitance sensor was used for contact density detection. The sensor's structural parameters were: electrode width w = 6 cm, electrode spacing g = 1 cm, electrode thickness t = 0.2 cm, electrode length L = 27 cm, and number of electrodes N = 4. The substrate was a 5 cm thick glass fiber board (relative permittivity = 4.5). The sensor was directly attached to the road surface, and the capacitance value after stabilization was recorded. Each measuring point was measured 5 times, and the average value was taken as the coplanar capacitance density detection data for that point.

[0081] The most recent periodic road condition inspection report was retrieved from the road maintenance unit's historical inspection record database. This report records detailed information on all apparent defects within the road section, including transverse cracks, longitudinal cracks, alligator cracks, and rutting. Hourly meteorological data for the area where the road section is located was then obtained. Finally, the pavement apparent defect data, raw radar echo image data, original deflection basin sequence data, and pavement capacitance data were saved as raw inspection data for subsequent analysis.

[0082] Step S2 includes the following sub-steps:

[0083] Step S21: Extract the original radar echo image data from the original detection data, input the original radar echo image data into the deep learning target detection model, and obtain the disease results. The disease results include disease categories and bounding box coordinates. Disease categories include cracks, poor interlayer structure, and looseness.

[0084] Step S22: Based on the results of the disease, the distribution area of ​​various diseases within the unit road segment is statistically analyzed. According to the pixel resolution of the ground penetrating radar image, the total actual area of ​​the disease is calculated. The road segment to be detected is cut into several sub-segments according to key points. The percentage of the total actual area of ​​the disease to the area of ​​the unit sub-segment is calculated to obtain the internal fragmentation rate.

[0085] Mechanical property parameters were extracted from the original sequence data of the deflection basin to obtain structural state parameters:

[0086] Extract the central deflection value and radial distance deflection value at the load center. The radial distance deflection value includes the first radial deflection value and the second radial deflection value. Subtract the first radial deflection value from the central deflection value to obtain the first damage index. Subtract the second radial deflection value from the central deflection value to obtain the second damage index. Save the first damage index and the second damage index as structural state parameters.

[0087] Step S23: Based on the modified variable separation method, the preprocessed coplanar capacitance density detection data is inverted to obtain the dielectric constant of the asphalt mixture;

[0088] Step S24: The dielectric constant of the asphalt mixture is converted into pavement asphalt state data using a temperature-dielectric mixing model.

[0089] In this embodiment, a deep learning target detection model (YOLOv8) is used to automatically analyze ground-penetrating radar images, overcoming the shortcomings of traditional manual interpretation, such as low efficiency, strong subjectivity, and difficulty in identifying subtle features. It can quickly and accurately locate hidden defects such as cracks, interlayer defects, and looseness. The internal breakage rate proposed in this embodiment innovatively transforms unstructured image defect features into structured numerical indicators. By calculating the proportion of defect area, it achieves a quantitative characterization of the degree of internal damage to the pavement, providing unified data for quality comparison of different road sections. The first damage index represents the surface layer damage index, and the second damage index represents the overall damage index. This embodiment changes the traditional single mode of evaluating pavement strength by relying on the center deflection value.

[0090] By utilizing the difference between the central deflection and the deflection at different radial distances, the mechanical response of the asphalt surface layer and the overall pavement structure can be effectively separated, clearly distinguishing whether the insufficient structural bearing capacity stems from the fatigue decay of the surface layer material or the failure of the deep structure. In step S23, the coplanar capacitance data is processed using the modified variable separation method to obtain the dielectric constant of the asphalt mixture, effectively eliminating measurement errors. In response to the nonlinear characteristics of the dielectric properties of the asphalt mixture changing with temperature, step S24 analyzes the dielectric response law under the thermodynamic field through the temperature-dielectric mixing model, uniformly correcting the test data collected at different temperatures to the standard reference state, realizing the dynamic decoupling of temperature and density, eliminating the deviation of test results caused by seasonal changes or solar temperature differences, ensuring the uniqueness and comparability of the pavement compaction evaluation results under different environmental conditions, and significantly improving the environmental adaptability and robustness of the test.

[0091] Step S23 specifically includes:

[0092] The coplanar capacitance density detection data is preprocessed to obtain the measured capacitance value, and the geometric parameters of the coplanar capacitance sensor are acquired. These geometric parameters include electrode width w, electrode length L, electrode spacing g, electrode thickness t, and number of electrodes. The dielectric constant of asphalt mixtures was calculated using the bisection method iteratively.

[0093] Randomly set the initial search lower threshold and initial search upper threshold. The initial search lower threshold is less than the initial search upper threshold and both are positive. Calculate the average of the initial search lower threshold and initial search upper threshold as the midpoint value of the search interval. Call the modified variable separation method to calculate the total capacitance value based on the current midpoint value of the search interval. The processing logic for the total capacitance value is as follows:

[0094] Using the point-matching method of the Laplace equation, the potential expansion coefficient is obtained based on the boundary conditions by searching for the midpoint value of the interval. The internal unit capacitance is then calculated by integrating the electric field over the electrode surface using Gauss's law. and external unit capacitors Wherein, the internal unit capacitance corresponds to the capacitance contribution between adjacent electrodes, and the external unit capacitance corresponds to the capacitance contribution between the edge electrode and adjacent electrodes. The number of terms in the series expansion is set to k=750 for internal units and k=2500 for external units. The total capacitance value is calculated using the total capacitance correction model, and the calculation expression of the total capacitance correction model is:

[0095] ;

[0096] in, This is an electrode thickness correction term. The length is extended to be equivalent to the edge effect.

[0097] Calculate the difference between the total capacitance and the measured capacitance. When the difference is greater than 0, replace the current midpoint value of the search interval with the upper limit threshold. When the difference is less than 0, replace the current midpoint value of the search interval with the lower limit threshold. Iterate and output the midpoint value of the search interval for the next round. Repeat the binary search method.

[0098] When the absolute value of the difference is less than the preset accuracy threshold, the iteration stops and the midpoint value of the current search interval is output as the dielectric constant of the asphalt mixture.

[0099] When the number of iterations exceeds the preset maximum number of iterations, the midpoint value of the current search interval is used as the dielectric constant of the asphalt mixture.

[0100] In this embodiment, the original coplanar capacitance density detection data is preprocessed to obtain the measured capacitance value used for inversion, and the geometric parameters of the currently used coplanar capacitance sensor are acquired. A bisection method is used as the inversion solver, with an initial lower search threshold of 1 and an initial upper search threshold of 10 randomly set. The midpoint value of the current search interval is calculated to be 5.5. Since the total capacitance value is a monotonically increasing function of the dielectric constant of the asphalt mixture, and because a larger dielectric constant means more energy stored in the measured material, resulting in a larger capacitance, the convergence of the bisection method is guaranteed. The modified variable separation method is then called, using the current midpoint value of the search interval as the assumed dielectric constant of the asphalt mixture, to calculate the total capacitance value.

[0101] Using the point-matching method of the Laplace equation, the potential expansion coefficient is solved by the midpoint value and boundary conditions of the current search interval. The electric field on the electrode surface is integrated according to Gauss's law to calculate the internal and external unit capacitances, respectively. The internal unit capacitance corresponds to the capacitance contribution between adjacent electrodes in the middle of the sensor, and the external unit capacitance corresponds to the capacitance contribution between the edge electrodes and their adjacent electrodes. A preset accuracy threshold is set. The maximum number of iterations is preset to 100. Electrode thickness and edge effect corrections are introduced. The total capacitance value is calculated through the total capacitance correction model. The search interval is adjusted according to the positive or negative difference between the total capacitance value and the measured capacitance value. When the difference is greater than 0, it indicates that the current assumed dielectric constant is too large, and the midpoint value of the current search interval is replaced with the upper limit threshold. When the difference is less than 0, it indicates that the current assumed dielectric constant is too small, and the midpoint value of the current search interval is replaced with the lower limit threshold. After the update is completed, the new midpoint value of the search interval is recalculated, and the next iteration is started. The bisection method is repeated to obtain the dielectric constant of the asphalt mixture.

[0102] Step S24 specifically includes:

[0103] Based on the modified temperature dielectric mixing model, the dielectric constant of asphalt mixture is calculated, and the density of asphalt mixture is obtained.

[0104] Obtain the maximum density of asphalt on the road to be tested, calculate the ratio of the density of the asphalt mixture to the maximum density of asphalt, and obtain the compaction degree. Use the density of the asphalt mixture and the compaction degree as the pavement asphalt state data.

[0105] In this embodiment, the dielectric constant of the asphalt mixture obtained from the inversion in step S23, the on-site testing temperature of the road section to be tested, and the mix design parameters of the asphalt mixture are obtained. The mix design parameters include the effective asphalt content, aggregate density, and asphalt binder density. Based on the pre-established dielectric-temperature relationship, the dielectric constant of the asphalt binder at the current testing temperature is calculated using a quadratic polynomial, and the dielectric constant of the aggregate is calculated using a linear model. A modified temperature-dielectric mixing model is constructed. The dielectric constants of the asphalt binder, aggregate, and asphalt mixture are input into the temperature-dielectric mixing model to obtain the asphalt mixture density. The calculation expression of the temperature-dielectric mixing model is as follows:

[0106] ;

[0107] in, The density of the asphalt mixture. Let be the dielectric constant of the asphalt mixture. The dielectric constant of the asphalt binder is . Let be the dielectric constant of the aggregate. For effective asphalt content, For aggregate density, This represents the maximum density of asphalt.

[0108] Calculate the ratio of asphalt mixture density to maximum asphalt density to obtain compaction degree. Use asphalt mixture density and compaction degree as pavement asphalt condition data. Package the calculated asphalt mixture density and compaction degree as pavement asphalt condition data for the current measuring point.

[0109] Step S3 includes the following sub-steps:

[0110] Step S31: Based on environmental parameters, establish an asphalt pavement temperature field estimation model based on the heat conduction equation, calculate the temperature distribution inside the pavement as a function of depth and time, and calculate the average temperature of the road section to be tested at the time of testing and the daily temperature difference extreme value of the pavement within a unit period.

[0111] Step S32: Based on the pavement asphalt state data and average temperature, the dynamic modulus of the asphalt surface layer is corrected by temperature-density coupling using the WLF equation and density correction pump coefficient to obtain the effective dynamic modulus.

[0112] Step S33: Take the effective dynamic modulus of the current road segment and the surface layer thickness of the current road segment as a sample, use the Kriging surrogate model to calculate the mechanical response of each sample under standard axle load; calculate the fatigue life of each sample by combining the fatigue equation; and count the proportion of samples whose fatigue life is less than the design axle load number to obtain the longitudinal cracking probability.

[0113] Step S34: Based on Miner's linear cumulative damage criterion, calculate the fatigue damage due to periodic temperature changes, obtain the temperature damage index of cumulative temperature, and save the temperature damage index and longitudinal cracking probability as fatigue damage index.

[0114] In this embodiment, the temperature of the asphalt pavement is established by using a one-dimensional heat conduction equation based on environmental parameters. The heat conduction equation is then discretized and solved using the finite difference method to calculate the temperature distribution inside the pavement structure as a function of depth and time. Based on the calculated temperature distribution, the average temperature of the pavement structure of the section to be tested at the time of testing (the time when the testing vehicle passes through) is extracted, and the daily extreme temperature difference of the pavement structure within the testing period (half a year in this embodiment) is calculated.

[0115] Based on pavement asphalt condition data and average temperature, the standard dynamic modulus of the asphalt surface layer is corrected to obtain the effective dynamic modulus:

[0116] The temperature shift factor is calculated using the WLF equation. The frequency at the detection temperature is shifted to the equivalent reduced frequency at the reference temperature. Combined with the master curve equation, the temperature-corrected modulus is obtained. The calculation expression of the WLF equation is as follows:

[0117] ;

[0118] in, , For material constants, For reference temperature, The average temperature. The temperature shift factor is used; the ratio of measured compaction to standard compaction is calculated and used as the density correction coefficient; the effective dynamic modulus equals the corrected modulus. Density correction factor;

[0119] In this embodiment, compaction degree is incorporated into the dynamic modulus correction system. Through the coupled calculation of the WLF equation and density correction coefficient, the dynamic modulus of the asphalt pavement is refined. Traditional methods only consider the influence of temperature on the modulus, ignoring the contribution of compaction degree, a key construction quality indicator, to the material stiffness. This embodiment introduces a density correction coefficient, so that the corrected effective dynamic modulus reflects both the ambient temperature and the compaction state during the service period, which is more consistent with the actual stress behavior of the pavement structure. The corrected effective dynamic modulus serves as the core random variable for subsequent cracking probability calculation. Traditional methods only evaluate the current state of the pavement, while this method can predict the cracking risk and fatigue damage accumulation during the future service period. The cracking probability and accumulated temperature fatigue damage output in step S3 are used as key feature vectors input into the subsequent AFG-KAN model, improving the prediction accuracy of the road health index.

[0120] Step S33 specifically includes:

[0121] The effective dynamic modulus and surface layer thickness of the current road segment are used as a sample. The effective dynamic modulus and surface layer thickness of all road segments are statistically analyzed to generate several sample groups. These sample groups are then randomly generated using the Latin hypercube sampling method (note that the number of N sample groups is much larger than the number of sample groups). A Kriging surrogate model is constructed. The processing logic includes: generating several design points based on the range of effective dynamic modulus and surface layer thickness using the Latin hypercube method; calculating the tensile strain at the bottom of the asphalt surface layer under standard axle load for each design point using a layered elastic system program; training the Kriging surrogate model with the design points as input and the tensile strain as output; then performing Monte Carlo reliability analysis; fitting the probability distribution type and parameters of each variable based on the measured effective dynamic modulus and surface layer thickness data of all sub-road segments; generating N random samples from this distribution using Latin hypercube sampling; inputting the N samples into the trained Kriging surrogate model; and outputting the mechanical response value corresponding to each sample.

[0122] Input N sets of samples into the Kriging surrogate model and output the mechanical response value corresponding to each set of samples;

[0123] The fatigue equation is constructed based on the transfer function. The fatigue life estimate is obtained by calculating the effective dynamic modulus and mechanical response value. The calculation expression of the fatigue equation is as follows:

[0124] ;

[0125] in, This is an estimate of fatigue life. This is the mechanical response value. For effective dynamic modulus, , and These are the regression coefficients;

[0126] Obtain the total traffic load of the road section to be tested, compare the fatigue life estimate of each sample group with the total traffic load, count the number of failure samples in N samples whose fatigue life estimate is less than the total traffic load, calculate the proportion of the number of failure samples to N samples, and obtain the longitudinal cracking probability.

[0127] In this embodiment, the fatigue equation adopts the modified Asphalt Institute model, with regression coefficients... , and The steps to determine this are as follows:

[0128] and The value of is determined based on the typical fatigue characteristics of asphalt mixtures, and is taken as an empirical constant. Take empirical constants Regression coefficient The calculation is based on the volumetric properties of the asphalt mixture in the road section to be tested, and the formula is:

[0129] ;

[0130] in, The porosity of the asphalt mixture is obtained by converting the compaction data obtained in step S2. The effective asphalt volume percentage is taken from the pavement construction mix design report; This is the field calibration coefficient, with a value of 18.4.

[0131] Step S4 specifically includes:

[0132] The structural state parameters, fatigue damage indexes, pavement asphalt state data, and internal breakage rate are standardized and spliced ​​to obtain a decision feature vector.

[0133] A nonlinear fusion decision model based on adaptive feature-gated KAN is constructed to perform feature mapping regression prediction on feature vectors and output the road health index.

[0134] An AFG-KAN model is established, comprising an input layer, a feature gating layer, several intermediate hidden layers, and an output layer. The standardized decision feature vector is input into the feature gating layer, which uses a sigmoid gating mechanism to strengthen the standardized decision feature vector. Let the standardized decision feature vector be x∈R. The gating layer first calculates the gating weight vector:

[0135] ;

[0136] in, This refers to the Sigmoid function (S-shaped function). The weight matrix is ​​a learnable matrix. The bias vector is used as the weight vector. Each element of the gated weight vector takes a value between 0 and 1. The weighted feature vector is obtained by element-wise multiplication. The weighted feature vector is input into the KAN hidden layer. Based on the multi-feature fusion of B-spline nonlinear transformation, the road health index is output.

[0137] Nonlinear activation functions are defined on the connection edges of neurons in each layer of the network. These nonlinear activation functions are represented by B-spline basis functions.

[0138] ;

[0139] in, It is a non-linear activation function. For k-th order B-spline basis functions defined on the mesh, These are the spline coefficients optimized through training;

[0140] Based on Kolmogorov's superposition theorem, the output of each layer of the network is calculated; the output value of each node in the layer is the superposition of the outputs of all nodes in the previous layer after being transformed by a nonlinear activation function, and the calculation formula is as follows:

[0141] ;

[0142] in, This is the output of the i-th node in the l-th layer. This represents the number of nodes in this layer; high-dimensional nonlinear mapping of multi-source heterogeneous features of the road surface is achieved through multi-layer stacking. This is the output of the j-th node in the (l+1)-th layer;

[0143] AFG-KAN adds a feature gating layer between the input layer and the first hidden layer to adaptively adjust the weights of the input features, based on the standard Kolmogorov-Arnold network (KAN).

[0144] Traditional linear weighting methods cannot handle the interactions between indicators, while AFG-KAN automatically captures the complex relationships between features through multi-layer nonlinear transformations, making the evaluation results more in line with actual engineering laws. The weight output of the feature gating layer and the B-spline basis function form of the KAN network give the model decision-making process clear physical meaning. This model can be transferred and applied to different regions and different road structures, thus maintaining high prediction accuracy.

[0145] Step S5 specifically includes:

[0146] A threshold for classifying road health levels is set, and the road health index is compared with the threshold to determine the road level. The road health level classification thresholds include a first threshold, a second threshold, and a third threshold.

[0147] When the road health index exceeds the first threshold, the road health level is determined to be Level 1.

[0148] When the second threshold is less than the road health index and less than the first threshold, the road health level is determined to be level two.

[0149] When the third threshold is less than the road health index and less than the second threshold, the road health level is determined to be level three.

[0150] When the road health index is less than the third threshold, the road health level is determined to be level four.

[0151] In this embodiment, the first threshold is set to 0.9, the second threshold to 0.8, and the third threshold to 0.6, mapping continuous road health knowledge into four levels: good, fair, and poor. Different health levels correspond to different maintenance needs: Level 1 indicates good road surface condition, and routine maintenance is recommended; Level 2 indicates minor damage, and preventive maintenance can be carried out; Level 3 indicates moderate damage, which requires special attention and treatment; Level 4 warns of severe deterioration, and structural repair must be carried out immediately.

[0152] Example 2, refer to Figure 2 A highway pavement quality inspection system is provided, including a data acquisition module, a state extraction module, a fatigue damage assessment module, a health index quantification module, and a grade determination module.

[0153] The data acquisition module is used to collect raw detection data and environmental parameters of the road surface to be inspected.

[0154] The state extraction module is used to analyze the original detection data to obtain the characteristics of pavement distress and to quantify and calculate the structural state parameters. Based on the modified variable separation method and the temperature-dielectric hybrid model, the original detection data is processed to obtain pavement asphalt state data.

[0155] The fatigue damage assessment module is used to dynamically correct structural state parameters and environmental parameters through a cracking probability model of thermodynamic response, and calculate fatigue damage index.

[0156] The health index quantification module is used to construct decision feature vectors, and to perform feature mapping on the decision feature vectors through a nonlinear fusion decision model to output the road health index.

[0157] The rating module is used to determine the road health level based on the road health index.

[0158] This embodiment integrates 3D ground-penetrating radar, falling-weight deflectometer, coplanar capacitance sensor, and surface image data. Through deep data fusion, it can comprehensively capture the entire process of pavement information from microscopic material aging and mesoscopic structural damage to macroscopic mechanical attenuation, effectively avoiding missed diagnoses and misjudgments, and significantly improving the reliability of pavement quality inspection. A deep learning target detection model automatically identifies hidden defects in radar images, overcoming the problems of low efficiency and strong subjectivity in manual interpretation. The internal breakage rate index proposed in this invention innovatively transforms unstructured image features into structured numerical indicators, achieving precise quantification of the degree of internal pavement damage and providing a unified standard for lateral quality comparison of different road sections. Utilizing the difference principle of deflection basin geometry, the mechanical response of the asphalt surface layer is separated from the response of the base layer by calculating the surface layer damage index and the overall damage index. This invention achieves precise decoupling and layered evaluation of the mechanical properties of pavement structural layers. It employs a modified variable separation method, introducing electrode thickness and edge effect correction terms to eliminate measurement errors caused by sensor non-ideal factors. By introducing a temperature-dielectric hybrid model, it reveals the law of dielectric constant change with temperature from a microscopic thermodynamic perspective, achieving dynamic decoupling of temperature and density. This allows the detection system to maintain highly consistent evaluation results under different seasonal and time-of-day temperature conditions, solving the problem of traditional capacitance methods being greatly affected by ambient temperature. In the decision-making stage, this invention uses adaptive feature-gated AFG-KAN, utilizing B-spline basis functions to more accurately fit the complex nonlinear relationship between pavement performance indicators and health indices. The feature gating mechanism automatically suppresses noise interference in the detection data, intuitively displaying the contribution of each feature to the final health score.

[0159] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media containing computer-usable program code. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions can also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means implemented in a flow... Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the protection scope of the present invention.

Claims

1. A method for testing the quality of highway pavement, characterized in that, Includes the following steps: Step S1: Obtain the original detection data and environmental parameters of the road surface to be tested, and at the same time use an LCR measuring instrument to record the capacitance value to obtain the coplanar capacitance density detection data. Step S2: Analyze the original detection data using a deep learning model to obtain pavement distress characteristics and perform quantitative calculations to obtain structural state parameters. Calculate the original detection data based on the modified variable separation method and the temperature-dielectric hybrid model to obtain pavement asphalt state data. Step S2 includes the following sub-steps: Step S21: Extract the original radar echo image data from the original detection data, input the original radar echo image data into the deep learning target detection model, and obtain the disease results. The disease results include disease categories and bounding box coordinates. The disease categories include cracks, interlayer defects, and looseness. Step S22: Based on the results of the disease, the distribution area of ​​various diseases within the unit road segment is statistically analyzed. According to the pixel resolution of the ground penetrating radar image, the total actual area of ​​the disease is calculated. The road segment to be detected is cut into several sub-segments according to key points. The percentage of the total actual area of ​​the disease to the area of ​​the unit sub-segment is calculated to obtain the internal fragmentation rate. Mechanical property parameters were extracted from the original sequence data of the deflection basin to obtain structural state parameters: Extract the central deflection value and radial distance deflection value at the load center. The radial distance deflection value includes a first radial deflection value and a second radial deflection value. Subtract the first radial deflection value from the central deflection value to obtain a first damage index. Subtract the second radial deflection value from the central deflection value to obtain a second damage index. Save the first damage index and the second damage index as structural state parameters. Step S23: Based on the modified variable separation method, the preprocessed coplanar capacitance density detection data is inverted to obtain the dielectric constant of the asphalt mixture. The coplanar capacitance density detection data is preprocessed to obtain the measured capacitance value, and the geometric parameters of the coplanar capacitance sensor are acquired. These geometric parameters include electrode width w, electrode length L, electrode spacing g, electrode thickness t, and the number of electrodes. The dielectric constant of asphalt mixtures was calculated iteratively using the bisection method. Step S24: The dielectric constant of the asphalt mixture is converted into pavement asphalt state data using a temperature-dielectric mixing model. Step S24 specifically includes: Based on the modified temperature dielectric mixing model, the dielectric constant of asphalt mixture is calculated, and the density of asphalt mixture is obtained. Obtain the maximum density of asphalt on the road to be tested, calculate the ratio of the density of the asphalt mixture to the maximum density of asphalt, and obtain the compaction degree. Use the density of the asphalt mixture and the compaction degree as the pavement asphalt state data. Step S3: Based on the cracking probability model of thermodynamic response, the pavement asphalt state data and environmental parameters are dynamically corrected to calculate the fatigue damage index. Step S31: Based on environmental parameters, establish an asphalt pavement temperature field estimation model based on the heat conduction equation, calculate the temperature distribution inside the pavement as a function of depth and time, and calculate the average temperature of the road section to be tested at the time of testing and the daily temperature difference extreme value of the pavement within a unit period. Step S32: Based on the pavement asphalt state data and average temperature, the dynamic modulus of the asphalt surface layer is corrected by temperature-density coupling using the WLF equation and density correction pump coefficient to obtain the effective dynamic modulus. Step S33: Take the effective dynamic modulus of the current road segment and the surface layer thickness of the current road segment as a sample, use the Kriging surrogate model to calculate the mechanical response of each sample under standard axle load; calculate the fatigue life of each sample by combining the fatigue equation; and count the proportion of samples whose fatigue life is less than the design axle load number to obtain the longitudinal cracking probability. Step S34: Based on Miner's linear cumulative damage criterion, calculate the fatigue damage due to periodic temperature changes, obtain the temperature damage index of cumulative temperature, and save the temperature damage index and longitudinal cracking probability as fatigue damage index. Step S4: Standardize the structural state parameters, fatigue damage index, pavement asphalt state data and internal breakage rate, construct a decision feature vector, perform feature mapping on the decision feature vector through a nonlinear fusion decision model, and output the road health index. Step S5: Determine the road health level based on the road health index.

2. The method for detecting the quality of highway pavement as described in claim 1, characterized in that, Step S1 includes the following sub-steps: Step S11: Use three-dimensional ground-penetrating radar to continuously scan the road section to be detected to obtain the original radar echo image data. Step S12: Set key points for each unit road segment of the road section to be tested, deploy displacement sensors at each key point, use a falling weight deflectometer for fixed-point detection, and synchronously collect the road surface deflection response through the displacement sensors to obtain the original sequence data of the deflection basin. The original sequence data of the deflection basin includes sensor number, sequence data and load parameters. Directly fit the road surface at the key points of the road section to be tested using a coplanar capacitance sensor. Step S13: Obtain pavement surface defect data through the pavement historical inspection record database. The pavement surface defect data includes defect type, location, measured area, and severity. Step S14: Obtain the environmental parameters of the area where the road section to be detected is located. The environmental parameters include hourly air temperature, daily maximum temperature, daily minimum temperature, daily total solar radiation and average wind speed. Save the pavement surface defect data, original radar echo image data, original deflection basin sequence data and pavement capacitance data as original detection data.

3. The method for detecting the quality of highway pavement as described in claim 1, characterized in that, Step S23 specifically includes: The coplanar capacitance density detection data is preprocessed to obtain the measured capacitance value, and the geometric parameters of the coplanar capacitance sensor are acquired. These geometric parameters include electrode width w, electrode length L, electrode spacing g, electrode thickness t, and the number of electrodes. The dielectric constant of asphalt mixtures was calculated using the bisection method iteratively. Randomly set the initial search lower threshold and initial search upper threshold. The initial search lower threshold is less than the initial search upper threshold and both are positive. Calculate the average of the initial search lower threshold and initial search upper threshold as the midpoint value of the search interval. Call the modified variable separation method to calculate the total capacitance value based on the current midpoint value of the search interval. The processing logic for the total capacitance value is as follows: Using the point-matching method of the Laplace equation, the potential expansion coefficient is obtained based on the boundary conditions by searching for the midpoint value of the interval. The internal unit capacitance is then calculated by integrating the electric field over the electrode surface using Gauss's law. and external unit capacitors The total capacitance value is calculated using the total capacitance correction model, and the calculation expression for the total capacitance correction model is as follows: ； in, This is an electrode thickness correction term. To extend the length equivalent to the edge effect, This represents the total capacitance value. Calculate the difference between the total capacitance value and the measured capacitance value. When the difference is greater than 0, replace the current midpoint value of the search interval with the upper limit threshold. When the difference is less than 0, replace the current midpoint value of the search interval with the lower limit threshold. Iteratively calculate and output the midpoint value of the search interval for the next round, and repeat the binary search method. When the absolute value of the difference is less than the preset accuracy threshold, the iteration stops and the midpoint value of the current search interval is output as the dielectric constant of the asphalt mixture. When the number of iterations exceeds the preset maximum number of iterations, the midpoint value of the current search interval is used as the dielectric constant of the asphalt mixture.

4. The method for detecting the quality of highway pavement as described in claim 1, characterized in that, Step S33 specifically includes: The effective dynamic modulus and surface layer thickness of the current road segment are used as a set of samples. The effective dynamic modulus and surface layer thickness of all road segments are statistically analyzed to generate several sets of samples. The Latin hypercube sampling method is used to randomly generate N sets of samples. The N sets of samples are input into the Kriging surrogate model, and the mechanical response value corresponding to each set of samples is output. A fatigue equation is constructed based on the transfer function, and the fatigue life estimate is obtained by calculating the effective dynamic modulus and mechanical response value. The calculation expression of the fatigue equation is as follows: ； in, This is an estimate of fatigue life. This is the mechanical response value. For effective dynamic modulus, , and These are the regression coefficients; Obtain the total traffic load of the road section to be tested, compare the fatigue life estimate of each sample group with the total traffic load, count the number of failure samples in N samples whose fatigue life estimate is less than the total traffic load, calculate the proportion of the number of failure samples to N samples, and obtain the longitudinal cracking probability.

5. The method for detecting the quality of highway pavement as described in claim 1, characterized in that, Step S4 specifically includes: The structural state parameters, fatigue damage indexes, pavement asphalt state data, and internal breakage rate are standardized and spliced ​​to obtain a decision feature vector. A nonlinear fusion decision model based on adaptive feature-gated KAN is constructed, and feature mapping regression prediction is performed on the feature vector to output the road health index. An AFG-KAN model is established, comprising an input layer, a feature gating layer, several intermediate hidden layers, and an output layer. The standardized decision feature vector is input into the feature gating layer and subjected to feature gating enhancement processing to obtain a weighted feature vector. The weighted feature vector is then input into the KAN hidden layer, and multi-feature fusion based on B-spline nonlinear transformation is performed to output the road health index.

6. The method for detecting the quality of highway pavement as described in claim 1, characterized in that, Step S5 specifically includes: A threshold for classifying road health levels is set, and the road health index is compared with the threshold to determine the road level. The road health level classification threshold includes a first threshold, a second threshold, and a third threshold. When the road health index exceeds the first threshold, the road health level is determined to be Level 1. When the second threshold is less than the road health index and less than the first threshold, the road health level is determined to be level two. When the third threshold is less than the road health index and less than the second threshold, the road health level is determined to be level three. When the road health index is less than the third threshold, the road health level is determined to be level four.

7. A highway pavement quality testing system, applied in a highway pavement quality testing method as described in any one of claims 1-6, characterized in that, It includes a data acquisition module, a status extraction module, a fatigue damage assessment module, a health index quantification module, and a grade determination module; The acquisition module is used to acquire the original detection data and environmental parameters of the road surface to be inspected; The state extraction module is used to analyze the original detection data to obtain pavement distress characteristics, and to quantify and calculate structural state parameters. Based on the modified variable separation method and the temperature-dielectric hybrid model, the original detection data is processed to obtain pavement asphalt state data. The fatigue damage assessment module is used to dynamically correct the structural state parameters and environmental parameters through a cracking probability model of thermodynamic response, and calculate fatigue damage index. The health index quantification module is used to construct a decision feature vector, perform feature mapping on the decision feature vector through a nonlinear fusion decision model, and output the road health index. The level determination module is used to determine the road health level based on the road health index.

Images