Three-dimensional GPR-FWD integrated rapid diagnosis method and system for concealed diseases of frozen soil road

Abstract

The invention relates to the technical field of road engineering nondestructive testing and state evaluation, in particular to a three-dimensional GPR-FWD integrated rapid diagnosis method and system for concealed diseases of a frozen soil road. The method comprises the following steps: synchronously acquiring pavement-base course-roadbed body data and a multi-base-distance deflection sequence of a road to be measured; an electromagnetic forward modeling library is called to generate a three-dimensional slice template, template matching is carried out, and disease categories, geometric scale parameters of diseases and similarity scores are output; carrying out power inversion of the drop weight deflectometer, and outputting the layer thickness and the equivalent modulus; performing calculation based on the multi-base-distance deflection sequence, and outputting a plurality of condition indexes; performing composite end point grading, and outputting disease grades and corresponding treatment suggestions; and generating a diagnosis report. The GPR map interpretation ambiguity is reduced, the layered modulus inversion accuracy is improved, a diagnosis-grading-disposal engineering closed loop is realized, the hidden disease discovery and disposal efficiency and accuracy can be remarkably improved, and the method has wide industrial application value.

Description

Technical Field

[0001] This invention relates to the field of non-destructive testing and condition assessment technology for road engineering, specifically to a rapid diagnostic method and system for hidden defects in permafrost roads using a three-dimensional GPR-FWD integrated approach. Background Technology

[0002] With the rapid development of transportation infrastructure in plateau permafrost regions, the road service environment exhibits complex coupled characteristics of low-temperature cycles, freeze-thaw cycles, strong radiation, and large diurnal temperature variations. Road structural layers (surface layer, base layer, subgrade) and subgrade are prone to hidden damage such as loosening, voids, water accumulation, delamination, and stiffness reduction. Current technologies often employ falling weight deflectometers (FWD) for pavement structure inversion assessment or ground penetrating radar (GPR) for layered structure, water content, and void detection. However, these single methods have the following limitations:

[0003] (1) Static / quasi-static inversion bias of falling weight deflectometer: Most inversion models ignore transverse anisotropy and dynamic characteristics, which can easily lead to upward bias of modulus and unclear layer identification; (2) Ambiguity in the interpretation of ground-penetrating radar images: Different diseases (such as loose, cavitary, water-rich, etc.) may have similar characteristics in terms of time window, travel time and amplitude, which are difficult to distinguish reliably based on experience alone; (3) Lack of cross-method integration: The on-site operation process and data coordinate system are not connected, making it difficult to spatially register and comprehensively judge the deflection basin index of FWD and the GPR body data; (4) Insufficient decision-making loop: Most results remain at the level of "discovering problems" and lack executable rules that are linked to graded disposal actions (such as recording, governance, and closure).

[0004] Therefore, there is an urgent need for a rapid diagnostic method and system integrating three-dimensional GPR-FWD for hidden defects in permafrost roads: Under the same workflow and coordinate system, it utilizes a GPR three-dimensional physical forward modeling template library to reduce ambiguity in map interpretation, employs dynamic inversion using a falling weight deflectometer considering transverse anisotropy to obtain layered equivalent moduli, and introduces SCI, BDI, and other parameters. The composite endpoint classification rules, etc., enable an engineered closed loop of "discovery-location-quantification-classification-disposal".

[0005] Existing patent CN116362059B discloses a method and system for assessing the deterioration of road soil performance. The method includes the following steps: identifying the road to be analyzed and obtaining its state information; determining the factors contributing to the deterioration of the road soil performance based on the state information; constructing a model to assess the impact of road soil performance deterioration; and evaluating the degree of deterioration of the road soil performance by combining the model with the factors. This method can assess the deterioration of road soil performance based on real-time acquired state information, promptly reflecting the deterioration status of road soil performance and providing accurate assessment results, which helps to take timely maintenance and management measures, improving road safety and reliability. However, this solution does not consider the dynamic inversion bias of the falling weight deflectometer and the ambiguity of the interpretation of ground radar maps, nor does it achieve engineering closed-loop processing, resulting in shortcomings in adaptability to complex environments, diagnostic accuracy, and engineering practicality.

[0006] Existing patent CN116341296B discloses a method for assessing the probability of road collapse, which mainly includes the following steps: collecting road data information of the road to be analyzed, and obtaining road soil layer information and underground pipeline information through the road data information; judging the road soil layer disease factors of the road to be analyzed based on the road soil layer information; building a disease factor influence intensity model, and obtaining the influence probability of road soil layer disease factors in combination with road soil layer information; building an underground pipeline deterioration influence intensity model, and obtaining the underground pipeline deterioration influence probability by combining the underground pipeline deterioration influence intensity model with underground pipeline information; summarizing the influence probabilities of road soil layer disease factors and underground pipeline deterioration to assess the collapse probability of the road to be analyzed. The road collapse probability assessment method provided by this invention improves the scientificity and accuracy of the road collapse assessment process and provides a reference for many important projects such as road maintenance and condition assessment. However, this scheme does not consider the dynamic inversion deviation of the falling weight deflectometer and the ambiguity of the interpretation of ground radar map, nor does it achieve engineering closed loop, and has shortcomings in adaptability to complex environments, diagnostic accuracy, and engineering practicality. Summary of the Invention

[0007] The purpose of this invention is to overcome the above-mentioned shortcomings in the existing technology and provide a three-dimensional GPR-FWD integrated rapid diagnosis method and system for hidden diseases of frozen soil roads.

[0008] In a first aspect, the present invention provides a rapid three-dimensional GPR-FWD integrated diagnostic method for hidden defects in permafrost roads, comprising the following steps: S1. Synchronously collect pavement-base-subgrade data and multi-base distance deflection sequence of the road to be tested; the multi-base distance deflection sequence is as follows: ,in Deflection at the point of application of the load. Deflection at the distant measuring point; S2. Call the electromagnetic forward modeling library, generate a three-dimensional slice template with the candidate defects as the simulation object, and then perform template matching between the three-dimensional slice template and the pavement-base-subgrade data, and output the defect category, geometric scale parameters of the defects and similarity score of the pavement-base-subgrade data. Based on the multi-base deflection sequence, a falling weight deflectometer dynamic inversion is performed on the pavement-base-subgrade of the road under test, and the layer thickness and equivalent modulus are output. Based on the multi-base deflection sequence, the surface layer condition index, base layer condition index, and subgrade condition index are calculated and output. The surface condition index is: , The The deflection values ​​of the near-distance measuring points in the multi-base deflection sequence; the base condition index is... , The , The deflection value at the mid-distance measuring point in the multi-base deflection sequence; the subgrade condition index is... ; S3. Based on the disease category, geometric scale parameters and similarity score of the disease; layer thickness and equivalent modulus; surface layer condition index, base layer condition index, and subgrade condition index, perform composite endpoint classification, and output disease level and corresponding treatment suggestions; S4. Based on the output results of S2 and S3, generate a diagnostic report.

[0009] The pavement-base-subgrade data is a three-dimensional integrated detection data covering the physical and geometric characteristics of the entire structure of the pavement, base and subgrade of frozen soil roads, including dielectric constant, geometric shape, electromagnetic reflection signals and other physical and geometric features.

[0010] Preferably, in step S1, the road surface-base-subgrade data is acquired by a three-dimensional ground-penetrating radar, the radar excitation of the three-dimensional ground-penetrating radar adopts Ricker wavelet or Gaussian modulated sine wave, and the transmitting antenna and receiving antenna of the three-dimensional ground-penetrating radar are equidistantly arranged along the survey line.

[0011] Preferably, the electromagnetic forward modeling library called in S2 is constructed based on the finite-difference time-domain method of a single-axis anisotropic fully matched layer, and meets the following requirements when called: The spatial grid step size ranges from 1 / 20 to 1 / 10 of the wavelength; the wavelength is the wavelength of the electromagnetic wave corresponding to the radar center frequency within the material with the lowest dielectric constant in the road structure under test. The time step satisfies the Courant stability condition, and the Courant number of the Courant stability condition ranges from 0.5 to 0.95. The boundary thickness of the uniaxial anisotropic fully matched layer is not less than 10 of the spatial grids, so that the false reflection energy at the boundary of the uniaxial anisotropic fully matched layer is suppressed to below -80dB.

[0012] Preferably, in step S2, the candidate diseases include spherical cavities, rectangular cavities, loose areas, and water-rich layers. When generating the three-dimensional slice template, the dielectric constant, conductivity, geometric parameters, porosity, and water content of the candidate diseases are discretized and sampled to construct a template library. The geometric parameters include three-dimensional morphological parameters, the spatial distribution range of the loose areas, and the vertical thickness of the water-rich layers. During the template matching process, a model including an attention mechanism and a feature extraction module is constructed to analyze the pavement-base-subgrade data and the multi-base deflection sequence, and the template matching similarity score is calculated by combining the weighted combination of normalized cross-correlation and structural similarity.

[0013] Preferably, in step S2, the dynamic inversion of the falling weight deflectometer considers the transverse anisotropy of the permafrost road structure, specifically including: A dynamic spectrum unit model of a layered structure is established, and the impact load time history and measuring point spacing of the falling weight deflectometer are input. The pavement structure layer of the road to be tested is defined as a transversely isotropic medium. A priori range is set for the parameters to be determined by the dynamic inversion of the falling weight deflectometer; the parameters to be determined include the layer thickness, equivalent modulus and Poisson's ratio of the pavement-base-subgrade of the road to be tested; The objective function for the dynamic inversion of the falling weight deflectometer is the time-domain displacement residual, the frequency-domain displacement residual, and... The weighted sum of the regularization terms is solved using a combination of global search and local optimization.

[0014] Here, Poisson's ratio is a mechanical parameter to be solved in the dynamic inversion of the falling weight deflectometer, along with parameters such as layer thickness and equivalent modulus. It serves as the core input of the transversely isotropic medium model, ensuring the accuracy of the deflection response simulation and thus supporting the accuracy of the inversion of core parameters such as layer thickness and equivalent modulus.

[0015] Preferably, in step S1, while simultaneously collecting pavement-base-subgrade data and multi-base deflection sequences of the road under test, pavement temperature is also collected simultaneously. In step S2, the road surface temperature is combined with the average temperature of the road under test over 24 hours, and the temperature of the asphalt layer mid-surface depth is predicted using an empirical model. The predicted temperature of the asphalt layer mid-surface depth is used as a standard reference temperature, and the equivalent modulus is corrected using a reference temperature correction formula to obtain the corrected equivalent modulus. Generate a sensitivity report of the corrected equivalent modulus to changes in the equivalent modulus.

[0016] Preferably, in step S3, the specific rules for the composite endpoint grading are as follows: A similarity threshold is set based on historical engineering case database data; when the similarity score exceeds the similarity threshold, it is determined that there are corresponding candidate defects in the pavement-base-subgrade of the road to be tested; the defect categories are divided into moderate looseness, severe looseness, voids, and severe waterlogging; The judgment thresholds for the surface layer condition index, base layer condition index, and subgrade condition index are set respectively. The judgment thresholds are calibrated based on the regional specifications and historical data of the road under test. The surface layer condition index, base layer condition index, and subgrade condition index are compared with their respective judgment thresholds. When the condition index of any structural layer is greater than the judgment threshold of that structural layer, the structural layer is judged to be damaged. When the disease category is at least one of severe loosening, voiding, or severe water accumulation, and the base course or subgrade is damaged, the disease level is determined to be Level III, and the corresponding treatment recommendation is: immediately close the road and prioritize treatment. When any of the following conditions are met: moderate looseness with damage to the base layer, damage to both the road surface and the base layer, or severe looseness with damage to the road surface, the disease level is determined to be Level II. The corresponding treatment recommendation is: treatment within a specified period to prevent the spread of the disease. In other cases, the disease level is determined to be Level I, and the corresponding treatment recommendation is: include it in routine monitoring and conduct regular re-examinations.

[0017] Preferably, in step S4, the generation of the diagnostic report specifically includes: Data collection: The data includes the disease categories, geometric scale parameters of the diseases, similarity scores, layer thickness, equivalent modulus, and calculation results of the multi-base deflection sequence output by S2, as well as the disease levels and corresponding treatment suggestions output by S3. Data visualization generation: Fill the numerical data in the data into a preset parameter table; associate the geometric scale parameters with the positioning data collected synchronously in S1, render a three-dimensional voxel positioning map, and draw the calculation results of the multi-base deflection sequence as a radar chart; Data integration: The parameter tables, three-dimensional voxel positioning maps, radar charts, disease levels, and treatment recommendations are integrated according to a preset report template to generate a diagnostic report file in PDF or HTML format; Data presentation: The diagnostic report file is combined with a geographic information system to visualize the spatial distribution and location of the defects in the road, and the diagnostic report is archived in the historical engineering case library.

[0018] In a second aspect, the present invention provides an integrated rapid diagnostic system for hidden defects in permafrost roads using three-dimensional GPR-FWD, comprising: The data acquisition subsystem is used to perform the following steps: S1. Synchronously collect pavement-base-subgrade data and multi-base distance deflection sequence of the road to be tested; the multi-base distance deflection sequence is as follows: ,in Deflection at the point of application of the load. Deflection at the distant measuring point; The computational processing subsystem is used to perform the following steps: S2. Call the electromagnetic forward modeling library, generate a three-dimensional slice template with the candidate defects as the simulation object, and then perform template matching between the three-dimensional slice template and the pavement-base-subgrade data, and output the defect category, geometric scale parameters of the defects and similarity score of the pavement-base-subgrade data. S3. Based on the multi-base deflection sequence, perform dynamic inversion of the pavement-base-subgrade body of the road under test using a falling weight deflectometer, and output the layer thickness and equivalent modulus. Based on the multi-base deflection sequence, the surface layer condition index, base layer condition index, and subgrade condition index are calculated and output. The surface condition index is: , The The deflection values ​​of the near-distance measuring points in the multi-base deflection sequence; the base condition index is... , The , The deflection value at the mid-distance measuring point in the multi-base deflection sequence; the subgrade condition index is... ; Based on the disease category, geometric scale parameters and similarity score of the disease; layer thickness and equivalent modulus; surface layer condition index, base layer condition index, and subgrade condition index, a composite endpoint classification is performed, and the disease level and corresponding treatment suggestions are output.

[0019] S4. Based on the output results of S2 and S3, generate a diagnostic report; A data storage subsystem is used to store the data required for the operation of the computing and processing subsystem.

[0020] Preferably, the data acquisition subsystem includes a three-dimensional ground-penetrating radar array, a falling-weight deflectometer loading device, a synchronous positioning module, and a temperature sensor module; the synchronous positioning module includes a global navigation satellite system and an inertial measurement unit module, supporting second pulse triggering and a unified time scale; The computing subsystem also includes a graphics processor, on which the step of generating the three-dimensional slice template is executed in parallel. The data acquisition subsystem is integrated into a vehicle-mounted or semi-vehicle-mounted platform, and the platform is equipped with an engineering support module, which includes: The power supply module for the work vehicle supplies power to each of the subsystems. A safety control auxiliary module, which deploys traffic safety commands on the road to be tested.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. This invention provides a rapid three-dimensional GPR-FWD integrated diagnostic method for hidden defects in permafrost roads. This method increases the accuracy of data composite by simultaneously acquiring pavement-base-subgrade data and multi-baseline deflection sequences of the road under test; selecting... Nine points were used for detection, enabling comprehensive and efficient acquisition of deflection data within the effective radius corresponding to the load diffusion range of the falling weight deflectometer. This reasonable number of measurement points effectively characterizes the stress and defects of the road structure, while avoiding redundancy and efficiency losses caused by excessive measurement points. A typical three-dimensional slice template for defects was generated using an electromagnetic forward modeling library to reduce ambiguity in the interpretation of the graphs. The falling weight deflectometer was combined with the transverse anisotropy characteristics of frozen soil roads to provide more accurate grading indicators for subsequent composite endpoint grading. A composite endpoint grading rule linking 3D ground-penetrating radar defect types with falling weight deflectometer deflection indicators was constructed, achieving a fully integrated process from accurate identification, layered location, quantitative assessment to grading and treatment recommendations for hidden road defects. This significantly improves the efficiency and accuracy of hidden defect discovery and treatment. Simultaneously, a closed-loop engineering system of "diagnosis-grading-treatment" is formed, effectively adapting to the road maintenance needs in complex environments such as frozen soil areas, significantly reducing the risk of road collapse, extending road service life, and reducing maintenance costs. 2. This invention provides a three-dimensional GPR-FWD integrated rapid diagnostic system for hidden defects in permafrost roads. The data acquisition subsystem integrates a three-dimensional ground-penetrating radar array, a falling-weight deflectometer loading device, a synchronous positioning module, and a temperature sensor module onto a vehicle-mounted or semi-vehicle-mounted platform. Through PPS triggering and unified time-scale technology, it can simultaneously acquire pavement-base-subgrade data and multi-baseline deflection sequences during mobile detection. This eliminates problems such as deviations in defect location and deflection measurement points, and environmental interference caused by acquisition time differences, providing a reliable foundation for subsequent data fusion of "physical defects and mechanical properties" and avoiding misjudgments due to data misalignment. Furthermore, the computational processing subsystem automates the process from data acquisition to diagnostic conclusion generation, reducing reliance on manual experience and improving diagnostic consistency and efficiency. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the operation of the three-dimensional ground-penetrating radar in Example 1.

[0023] Figure 2 This is a schematic diagram of the working principle of the three-dimensional ground-penetrating radar in Example 1.

[0024] Figure 3 This is a schematic diagram of the three-dimensional ground-penetrating radar system in Example 1.

[0025] Figure 4 This is a schematic diagram of the working of the falling weight deflectometer in Example 1.

[0026] Figure 5 This is a schematic diagram of the deflection basin curve of the falling weight deflectometer in Example 1.

[0027] Figure 6 This is a schematic diagram of a two-dimensional slice of the pavement-base-subgrade data in Example 1.

[0028] Figure 7 This is a schematic diagram of a longitudinal section of a three-dimensional section template for a typical disease in Example 1.

[0029] Figure 8 This is a schematic diagram of a transverse slice of a three-dimensional slice template for a typical disease in Example 1.

[0030] Figure 9 This is a schematic diagram of a horizontal slice at the spherical cavity in Example 1.

[0031] Figure 10 This is a schematic diagram of a horizontal slice at the rectangular cavity in Example 1.

[0032] Marked in the image: 1. Target body to be detected; 2. Transmitting antenna; 3. Receiving antenna; 4. Antenna array; 5. Monitoring point of falling weight deflectometer; 6. Load disk; 7. Falling weight deflectometer; 8. Deflection basin; 9. Displacement sensor; 10. Load center; 11. Longitudinal slice; 12. Transverse slice; 13. Horizontal slice. Detailed Implementation

[0033] The present invention will now be described in further detail with reference to specific embodiments. However, this should not be construed as limiting the scope of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0034] Unless otherwise specified, the terms "upper," "lower," "left," "right," "center," "inner," and "outer," etc., used in the description of specific embodiments of the present invention to indicate orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings, or the orientation or positional relationship in which the product / equipment / device is usually placed during use. These terms are merely for the purpose of facilitating the description of the present invention or simplifying the description in specific embodiments, and for enabling those skilled in the art to quickly understand the solution, and do not indicate or imply that a particular device / component / element must have a specific orientation, or be constructed and operated in a specific positional relationship. Therefore, they should not be construed as limitations on the present invention.

[0035] Furthermore, the use of terms such as "horizontal," "vertical," "suspended," "parallel," and "coaxial" does not imply that the corresponding device / component / element must be absolutely horizontal, vertical, suspended, parallel, or coaxial. Slight tilt or deviation is permissible, as long as it does not affect the normal function of the relevant component. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," not that the structure must be perfectly horizontal; a slight tilt is acceptable. "Coaxial" means that two components are arranged as coaxially as possible, allowing them to move coaxially or approximately coaxially when their relative positions change. Alternatively, it can be simplified to mean that the corresponding device / component / element, when arranged in "horizontal," "vertical," "suspended," "parallel," or "coaxial" directions, can have an error / deviation of ±10% relative to the corresponding direction, more preferably within ±8%, more preferably within ±6%, more preferably within ±5%, and more preferably within ±4%. For example, the deviation in the "coaxial" direction is controlled within 0.2-1mm, preferably within 0.2-0.5mm. As long as the corresponding device / component / element is within the error / deviation range, it can still achieve its function in the solution of the present invention.

[0036] Furthermore, the use of terms such as "first," "second," and "third" in terminology is merely for distinguishing descriptions of identical or similar components and should not be interpreted as emphasizing or implying the relative importance of a particular component.

[0037] Furthermore, in the description of the embodiments of the present invention, "several", "more than", and "a number of" represent at least two. The number can be any number, such as two, three, four, five, six, seven, eight, or nine, and can even exceed nine.

[0038] Furthermore, in the description of the technical solution of this invention, unless otherwise explicitly specified / limited / restricted, the terms "set up," "install," "connect," "link," "provided with," "laid out," and "arranged" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to connection methods commonly used in the art, such as welding, riveting, bolting, and threaded connections. Such connections can be mechanical, electrical, or communication connections; they can be direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components.

[0039] Example 1 This embodiment uses a frozen soil road as the test road. This road exhibits potential hidden defects such as loose base layer and water-rich subgrade caused by freeze-thaw cycles. The three-dimensional GPR-FWD integrated rapid diagnostic method and system are used to complete defect detection, assessment, and report generation. The specific implementation process is as follows: I. Equipment Selection (1) Data acquisition subsystem, integrated on the vehicle platform, specifically includes: (a) Three-dimensional ground-penetrating radar: Stepped frequency system equipment (such as GeoScope) is selected. TM The system operates in a frequency band covering 100MHz to 3GHz, with the center frequency selected based on target depth and resolution requirements (e.g., 400 to 900MHz). The sampling step size and number of stacks are set according to signal-to-noise ratio requirements. Radar excitation uses Ricker wavelet or Gaussian-modulated sine waves. Transmitting and receiving antennas are equidistantly positioned along the survey line to ensure lateral data continuity. It collects a data cube of pavement-base layer-subgrade structure, acquiring data from multiple survey lines simultaneously to form a three-dimensional image of the underground structure. Its operational overview is as follows: Figure 1 As shown, Figure 1 The diagram schematically illustrates the basic working principle of ground-penetrating radar, including a transmitting antenna 2 and a receiving antenna 3. The transmitting antenna emits high-frequency electromagnetic waves into the ground. During propagation, the electromagnetic waves are reflected when they encounter interfaces with different dielectric properties (such as the interface between the road surface, base layer, and subgrade) or the target object 1 (such as a cavity). The reflected signal is captured by the receiving antenna, and the underground structure is interpreted based on the signal's travel time and amplitude.

[0040] Figure 2 The image illustrates a scenario where the GeoScope™ 3D ground-penetrating radar system, preferred in this invention, is collecting data on a road. The figure shows an antenna array 4 mounted behind a vehicle. This array contains multiple antenna elements and can scan a wide area of ​​the road simultaneously while the vehicle is moving, thereby acquiring the 3D data volume beneath the road (i.e., the coordinate map shown on the right side of the image), achieving a comprehensive imaging of subsurface defects. Figure 3 This is a functional block diagram of the three-dimensional ground-penetrating radar system used in this invention. The system is centered around a host controller, which is responsible for issuing commands and processing data. The antenna array module is responsible for transmitting and receiving electromagnetic waves. The data acquisition and storage unit is responsible for digitizing and storing the analog signals received by the antenna.

[0041] (b) Falling weight deflectometer: Set standard impact energy level and collect multi-base deflection sequences. (Sensors are arranged radially from the loading point) to acquire instantaneous vertical displacement data from the load application point to the far-end measuring point, forming a deflection basin curve for subsequent deflection basin index calculation, thus reflecting the bearing capacity of the pavement structure. Its working principle is as follows: Figure 4 As shown, Figure 4 The working principle of the falling weight deflectometer 7 is illustrated schematically. The figure shows a weight falling from a certain height and impacting a load disk 6 placed on the road surface, thus simulating vehicle load. A series of displacement sensors 9 arranged along the diameter of the load-bearing plate of the falling weight deflectometer monitoring point 5 are used to measure the instantaneous vertical displacement of the road surface at different distances, i.e., the deflection value.

[0042] Figure 5 This is a schematic diagram of the deflection basin curve of a typical falling weight deflectometer, illustrating the variation of road surface displacement with distance from the load center. The horizontal axis represents the distance from the load center (10), corresponding to the position of displacement sensor 9 mentioned in the text. , ,..., The longitudinal direction represents the road surface deflection value, corresponding to the values ​​described in the text. Deflection value measured at the location The curve formed by connecting the deflection values ​​of each measuring point is called the deflection basin 8, and its shape reflects the overall bearing capacity of the pavement structure.

[0043] (c) Temperature sensor: Real-time acquisition of road surface temperature provides data for temperature correction of modulus in dynamic inversion of falling weight deflectometer, solves the problem of temperature sensitivity of asphalt pavement layer stiffness, and avoids interference of temperature fluctuations on bearing capacity assessment.

[0044] (d) Synchronous Positioning Module: Includes Global Navigation Satellite System and Inertial Measurement Unit modules. It adopts fusion positioning technology, supports pulse-per-second (PPS) triggering and unified time scale, and realizes unified coordinate registration between ground penetrating radar volume data and falling weight deflectometer measurement points, ensuring spatiotemporal data alignment and eliminating misalignment problems in cross-method data fusion.

[0045] The vehicle platform is also equipped with an engineering support module, which specifically includes: The power supply module of the work vehicle is powered by a lithium battery pack and has an on-board generator as a backup, ensuring the continuous and stable operation of equipment such as 3D ground penetrating radar, falling weight deflectometer, and synchronous positioning module, and avoiding data loss due to power outages.

[0046] Safety control auxiliary module: Install warning lights on the vehicle platform and set up safety warning signs (closing part of the lanes) before and after the road to be tested to ensure the safety of personnel and equipment during the testing process and avoid traffic interference that may cause interruption of data collection or decrease in accuracy.

[0047] (2) Computational processing subsystem (a) Hardware module: Equipped with an industrial control computer and a graphics processing unit (GPU, such as NVIDIA Tesla V100) parallel computing module to accelerate the template generation of the electromagnetic forward modeling library and improve the efficiency of three-dimensional forward modeling calculation.

[0048] Parallel strategy: Using the CUDA programming model, the three-dimensional temporal finite difference computation domain is spatially divided into threads on the GPU. Each thread is responsible for updating the electromagnetic field components of a Yee cell, and data exchange between adjacent threads is optimized by sharing memory.

[0049] On the NVIDIA Tesla V100 GPU hardware platform, a typical pavement layer disease model containing 100 million grid nodes was tested. The traditional serial three-dimensional finite difference method (such as running on an Intel Xeon Gold 6148 CPU) took 3156.328 seconds to compute, while the parallel three-dimensional finite difference method used in this system only took 90.517 seconds, which greatly shortened the generation time of the forward modeling template.

[0050] (b) Software modules: An electromagnetic forward modeling library based on the finite-difference time-domain method for uniaxial anisotropic fully matched layers: supports three-dimensional electromagnetic forward modeling, proceeding in four steps: "sequentially updating electric field, magnetic field, and auxiliary variables of the uniaxial anisotropic fully matched layer". It achieves stable absorption effect (suppressing spurious boundary reflection energy below -80dB) under the condition that the thickness of the uniaxial anisotropic fully matched layer is ≥10 layers, and improves computational efficiency by ≥90% compared with serial three-dimensional finite-difference time-domain method.

[0051] The dynamic spectrum unit inversion module supports the input of transverse anisotropy parameters (deflection modulus ratio, i.e., the ratio of the equivalent modulus in the horizontal direction to the equivalent modulus in the vertical direction, Eh / Ev). Its prior value range is set between 0.5 and 2.0 according to the soil type and pavement material (such as cement-stabilized crushed stone and graded crushed stone) in the frozen soil area. Based on the layered structure dynamic spectrum unit model, the module inverts the layer thickness and equivalent modulus. The objective function adopts a joint measure of the time domain and frequency domain residuals (the weighting coefficients are dynamically adjusted according to the signal-to-noise ratio), and introduces an L1 regularization term to improve robustness and avoid overfitting of the inversion results.

[0052] Deflection basin index calculation module: Automatically calculates surface layer index Grassroots Index and distal deflection All units are calibrated to micrometers using the displacement sensor of the falling weight deflectometer, supporting comparison of indicators under different load levels or temperature conditions, enhancing the indicative accuracy of strata and solving the problem of ambiguous strata indicative accuracy under a single working condition.

[0053] Composite endpoint grading module: Includes a built-in threshold library and composite grading rules. It can access template libraries and historical engineering case libraries to assist in grading determination, achieving logical fusion of physical defects and structural bearing capacity indicators (i.e., surface layer condition indicators, base layer condition indicators, and subgrade condition indicators). The threshold library supports regression calibration based on historical data by region or road grade, and the template library contains three-dimensional slice templates for typical defects.

[0054] Report generation module: Supports exporting in PDF or HTML format, and can automatically integrate 3D voxel positioning maps, falling weight deflectometer dynamic inversion parameter tables (including layer thickness, equivalent modulus, etc.), and index radar charts (including...). , , Numerical distribution and hierarchical list are used to generate standardized diagnostic reports.

[0055] (3) Data storage subsystem It includes a template library, a threshold library, a historical engineering case library, and a calibration parameter library.

[0056] Template library: Stores three-dimensional slice templates of typical diseases (including spherical cavities, rectangular cavities, loose areas, and water-rich layers) in the longitudinal (X-Z axis plane direction), transverse (Y-Z axis plane direction), and horizontal (X-Y axis plane direction) directions, covering the dielectric constant, conductivity, and geometric scale parameters of different diseases (such as the radius / length, width, and height of cavities, porosity of loose areas, and water content of water-rich layers, etc., discretized sampling data). Threshold library: Stores grading thresholds for deflection basin indicators (e.g., default SCI>130, BDI>30, etc.). >80, based on statistical calibration of 200 km test road section data in North China, decision accuracy is cross-validated >90%); supports importing calibration datasets according to the regional characteristics of permafrost areas for adaptive calibration. Historical Engineering Case Library: Archives past cases of permafrost road diagnosis, including original collected data, processing results, classification conclusions and treatment feedback, providing a reference for the diagnosis of similar diseases; Calibration parameter library: Records equipment calibration data (such as 3D ground penetrating radar antenna parameters, falling weight deflectometer displacement sensor calibration values, and global navigation satellite system / inertial measurement unit positioning accuracy parameters) and key forward / inverse parameters (such as 3D time-domain finite difference grid step size, Courant number, inversion convergence criteria, etc.) to ensure data processing consistency.

[0057] II. Collaborative Data Acquisition (S1) The measurement line is laid out longitudinally along the road section to be measured, and the vehicle platform is controlled to travel at a constant speed (the speed is set according to the data acquisition accuracy requirements to ensure that the data sampling density of the three-dimensional ground penetrating radar and the falling weight deflectometer meets the standards). The three-dimensional ground penetrating radar, the falling weight deflectometer and the synchronous positioning module are started simultaneously.

[0058] Three-dimensional ground-penetrating radar was used to acquire pavement-base-subgrade data of the road under test, and a falling weight deflectometer was used to acquire multi-base distance deflection sequences of the road under test. The above multi-base distance deflection sequences are as follows: ,in Deflection at the point of application of the load. Deflection at remote measuring points; real-time road surface temperature of the road under test; When collecting the above-mentioned pavement-base-subgrade data and the above-mentioned multi-base deflection sequence, the spatial coordinates of the two types of data are aligned in the same location through positioning technology, and the collection time of the two types of data is synchronized through unified time synchronization.

[0059] (1) Ground penetrating radar data acquisition Data on the pavement, base course, and subgrade were acquired, with dimensions of length (X), width (Y), and depth (Z). The spatial coordinates and time scale of the data were recorded. During the data acquisition process, the radar center frequency was adjusted according to the structural characteristics of roads in permafrost regions (such as the thickness of the freeze-thaw layer) to ensure high detection resolution for defects such as loose base course and water-rich subgrade. Parameters such as the distance between the radar transmitting / receiving antennas, sampling step size, and number of stacking operations were also recorded. The radar excitation of the aforementioned 3D ground-penetrating radar used Ricker wavelet or Gaussian-modulated sine waves, and the transmitting and receiving antennas of the aforementioned 3D ground-penetrating radar were equidistantly arranged along the survey line.

[0060] Subsequently, two-dimensional slice images from different directions can be extracted from this three-dimensional data volume for disease interpretation, such as... Figure 6 As shown, Figure 6The extraction method is illustrated schematically, including: longitudinal slice 11, a slice along the vehicle's direction of travel; transverse slice 12, a slice perpendicular to the direction of travel; and horizontal slice 13, a planar view at a certain depth. This multi-angle slicing helps to accurately identify the three-dimensional morphology of the disease.

[0061] (2) Data acquisition of falling weight deflectometer Set a standard impact energy level, apply load according to the preset measurement point spacing (matching the three-dimensional ground penetrating radar survey line), collect instantaneous vertical displacement data of the load application point and each measurement point at the far end to form a deflection basin curve; repeat the acquisition 2-3 times at each measurement point to reduce random errors, record the load time history and displacement response data for each loading to ensure data validity.

[0062] (3) Auxiliary data collection The temperature sensor records the road surface temperature in real time (the sampling interval is synchronized with the data acquisition of the 3D ground penetrating radar and the falling weight deflectometer). The synchronous positioning module obtains the absolute coordinates through the global navigation satellite system and the inertial measurement unit compensates for dynamic errors. It marks each voxel unit of the 3D ground penetrating radar volume data and the measuring point of the falling weight deflectometer with unified spatial coordinates and time scales, ensuring that the two types of data are accurately aligned in the spatiotemporal dimension and avoiding data misalignment problems caused by cross-method fusion.

[0063] III. Data Processing and Indicator Calculation (S2) The electromagnetic forward modeling library is invoked to generate a three-dimensional slice template using candidate defects as simulation objects. This template is then matched with the pavement-base-subgrade data to output the defect category, geometric scale parameters, and similarity score. Based on the multi-base deflection sequence, a falling weight deflectometer dynamic inversion is performed on the pavement-base-subgrade of the road under test, outputting layer thickness and equivalent modulus. Calculations are then performed based on the multi-base deflection sequence to output surface layer condition indicators, base layer condition indicators, and subgrade condition indicators. The aforementioned surface layer condition indicators are... , The above The above-mentioned deflection values ​​are the near-distance measuring points in the multi-base deflection sequence; the above-mentioned base condition indicators are... , The above , The above refers to the deflection value at the mid-distance measuring point in the above multi-base deflection sequence; the above roadbed condition index is... .

[0064] (1) Ground Penetrating Radar Forward Modeling and Three-Dimensional Slice Template Generation and Matching The electromagnetic forward modeling library includes a three-dimensional slice template generation function and a template library. The three-dimensional slice template generation function generates typical disease three-dimensional slice templates through the finite difference time-domain method of a single-axis anisotropic fully matched layer. The generated templates are uniformly stored in the template library.

[0065] (a) Forward modeling parameter settings Grid step size: Follow the "wavelength / 10" principle, that is, the spatial grid step size (Δx, Δy, Δz) is no greater than one-tenth of the wavelength corresponding to the radar center frequency in the material with the smallest dielectric constant (such as the freeze-thaw soil layer) in the road model of the frozen soil area, so as to ensure that the numerical dispersion error is within an acceptable range. At the same time, it is adjusted in combination with the size of the disease (such as the smallest scale of the loose area of ​​the base layer) to ensure spatial accuracy and take into account the computational efficiency. Time step Strictly adhere to the Courant stability condition, setting the Courant number to any value between 0.5 and 0.95, typically set to 0.5, which satisfies the formula. , where c is the propagation speed of electromagnetic waves in the medium, ensuring the numerical stability of the algorithm over a long period of iteration.

[0066] Verification of the absorption effect of uniaxial anisotropic fully matched layer: Numerical experiments were conducted when the boundary thickness of the uniaxial anisotropic fully matched layer was ≥10 layers. By comparing the boundary reflection coefficients at different thicknesses (5 layers, 10 layers, and 15 layers), the results show that when simulating a typical asphalt-gravel-soil subgrade structure, a uniaxial anisotropic fully matched layer of more than 10 layers can suppress the false reflection energy generated at the boundary to below -80dB, meeting the requirements of high-precision simulation.

[0067] (b) Template library construction Candidate defects include cavities (such as spherical cavities and rectangular cavities), loose areas, and water-rich layers; when generating triaxial slice templates, the dielectric constant, conductivity, geometric parameters, porosity, and water content of candidate defects are discretized and sampled to construct a template library.

[0068] An electromagnetic forward modeling library based on the finite-difference time-domain method for uniaxial anisotropic fully matched layers is used to set parameters for candidate defects (such as typical loose base courses, water-rich subgrades, and potential spherical / rectangular cavities in permafrost regions). Medium parameters: Input the empirical values ​​of dielectric constant and conductivity of each structural layer (surface layer, base layer, and roadbed freeze-thaw layer, i.e., pavement, base layer, and roadbed body) in the frozen soil area. Adjust the dielectric constant according to the porosity in the loose area and set the conductivity according to the water content in the water-rich layer. Geometric parameters: Discretized sampling was performed on cavities (spherical radius 5-30cm, rectangular length, width and height 10-50cm×10-50cm×5-20cm), loose areas (range 1-3m×1-3m×0.2-1m), and water-rich layers (thickness 0.3-1.5m); Template generation: Output longitudinal, transverse, and horizontal slice templates for each disease and store them in the template library. The response templates for spherical and rectangular cavities on slices in different directions are shown below. Figures 7-10 As shown, Figures 7-10 It is generated through numerical simulation in step S2 of this invention.

[0069] Figure 7 This is a longitudinal slice diagram of a three-dimensional slice template for a typical pavement defect. The interfaces of the three pavement layers are clearly visible in longitudinal slices at different locations. On different cross-sections of the sphere, the hyperbola is most pronounced and has the largest opening in the straight section directly above. As the radius of the sphere's cross-section decreases in the straight slices, the hyperbola becomes less pronounced and the opening becomes smaller. However, the tangent plane of the rectangular cavity remains the same; when the position of the rectangular cavity cross-section changes, the hyperbola dimensions remain constant.

[0070] Figure 8 This is a schematic diagram of a transverse slice of a three-dimensional slice template for a typical disease. The pattern is the same as that of the longitudinal slice. When the position of the cross section changes, the hyperbola changes at the spherical cavity, while the hyperbola remains unchanged at the rectangular cavity. Figure 7 and Figure 8 The horizontal axis represents spatial distance, that is, the physical position of the 3D ground-penetrating radar antenna along the survey line, in centimeters; the vertical axis represents two-way travel time, in nanoseconds. This axis directly reflects depth. The horizontal stripes in the figure appear at different time points, representing the reflection of different road surface layers, corresponding to different depths. The time point where the vertex of the hyperbola is located corresponds to the depth of the top surface of the underground cavity.

[0071] Figure 9 By taking horizontal slices at different depths at the spherical cavity, it can be seen that as the radius of the sphere's cross-section on the horizontal slice increases, the image becomes more and more obvious, and no image can be found where there is no sphere.

[0072] Figure 10 The image shows horizontal slices at different depths within the rectangular cavity, revealing images at different depths. No images were found where there was no rectangular cavity.

[0073] Figure 9 and Figure 10 The horizontal axis together form a horizontal plane, representing the planar spatial position of the detection area, with the unit being centimeters; the vertical axis represents the depth. Each slice image is displayed on a fixed two-way travel time, showing the amplitude distribution of the reflected signal across the entire plane. By observing the images on different time slices, the planar shape of the disease and its changes with depth can be clearly seen, thus reconstructing the three-dimensional morphology of the disease.

[0074] (c) Data preprocessing: Preprocessing of measured 3D volumetric data from 3D ground-penetrating radar: Spatial alignment: Based on the three dimensions of the template, slices with the same spatial dimension and resolution are cut from the measured data (e.g., if the horizontal slice resolution of the template is 1cm×1cm and the range is 1m×1m, then slices with the same parameters as the measured data are cut accordingly), without the need for additional geometric calibration; Noise suppression: Filtering algorithms (such as background removal and bandpass filtering) are used to eliminate strong reflection noise and electromagnetic interference from road surfaces in frozen soil areas, thereby improving signal purity.

[0075] (d) Template matching and similarity calculation: Matching method: A model incorporating an attention mechanism and a feature extraction module is constructed to perform collaborative analysis on the three-dimensional ground-penetrating radar map and the multi-base deflection data. The feature extraction module adopts a convolutional neural network (CNN), and the attention mechanism adopts a channel attention module to achieve disease template matching in the data.

[0076] Based on the matching results of the above model, a weighted combination of Normalized Cross-Correlation (NCC) and Structural Similarity (SSIM) is used to calculate the similarity for combined verification. The weighting coefficients α (for NCC) and β (for SSIM) in the similarity calculation formula, where α + β = 1, can be optimized by performing Receiver Operating Characteristic (ROC) curve analysis on a dataset containing over 200 known disease samples verified in field excavation, to achieve the best balance between detection accuracy and recall. In a preferred embodiment, α is set to 0.6 and β to 0.4.

[0077] Similarity determination: Normalized cross-correlation reflects the consistency of signal trends (such as the matching degree of reflection peak / trough positions, close to 1 is a match), structural similarity reflects the consistency of structural features (such as the matching degree of disease outline and light-dark contrast, close to 1 is a match), and the final similarity S=α×NCC+β×SSIM.

[0078] The model matching method that incorporates attention mechanism and feature extraction module can be used alone for high-precision matching in complex permafrost regions, while the weighted combination calculation method using normalized cross-correlation and structural similarity can be used alone for fast matching in conventional permafrost regions.

[0079] (e) Disease identification: Threshold standards: Matching thresholds are set for different diseases, such as a threshold of 0.85 for spherical / rectangular cavities (obvious geometric features) and a threshold of 0.75 for loose base course / water-rich subgrade (dispersed features). These thresholds are statistically calibrated through sample datasets to ensure classification accuracy within a 95% confidence interval.

[0080] Output results: When the measured slice similarity is greater than or equal to the corresponding threshold, the presence of this type of disease is determined, and the geometric scale (such as cavity radius, loose area range) and spatial location (depth, planar coordinates) of the disease are estimated by combining the three-dimensional slice matching results.

[0081] (2) Dynamic inversion of falling weight deflectometer Model Establishment: The permafrost road structure layers (surface layer, base layer, and subgrade freeze-thaw layer) are defined as transversely isotropic media. A layered dynamic spectral element model is established. Measured load data from a falling weight deflectometer and the spacing between measuring points are input. A refined integration algorithm is employed, dividing the thick layer into sub-layers and micro-layers and combining Taylor series expansion techniques to solve the transformation domain dynamic equations, avoiding the numerical "exponential overflow" problem under large thickness / high wavenumber conditions. A dynamic algorithm is used instead of a static algorithm for inversion to reduce the risk of modulus bias.

[0082] The aforementioned falling weight deflectometer dynamic inversion considers the transverse anisotropy of permafrost road structures, specifically including: A dynamic spectrum unit model of a layered structure is established. The impact load time history of the falling weight deflectometer and the distance between measuring points are input. The pavement structure layer of the road to be tested is defined as a transversely isotropic medium. The parameters to be determined include the thickness of the pavement-base-subgrade of the road under test, the deflection modulus ratio, the equivalent modulus and Poisson's ratio, and a priori ranges are set for the above parameters. The inversion objective function is the displacement residual in the time domain, the displacement residual in the frequency domain, and... The weighted sum of the regularization terms is solved using a combination of global search and local optimization. Output the thickness and equivalent modulus of the above-mentioned pavement-base-subgrade structure; The road surface temperature in S1 is combined with the average temperature of the road under test over 24 hours, and the temperature at the mid-surface depth of the asphalt layer is predicted using an empirical model. The predicted temperature at the mid-surface depth of the asphalt layer is used as a standard reference temperature, and the equivalent modulus is corrected using a standard temperature correction formula to make the equivalent modulus the modulus value at the standard reference temperature. The average temperature over 24 hours is obtained from meteorological station data near the road section under test or historical data from vehicle temperature recorders. Generate a sensitivity report of the modified equivalent modulus to changes in the deflection modulus ratio.

[0083] (a) Parameter settings: Parameters to be determined: including the thickness of each structural layer, equivalent modulus (E), deflection modulus ratio (Eh / Ev), and Poisson's ratio (set within the range based on empirical values ​​for frozen soil materials). Priorities and boundaries: Set reasonable prior ranges for the deflection (B) / modulus (E) ratio, layer thickness, and Poisson's ratio (e.g., layer thickness ±20% of the design drawings, Poisson's ratio 0.2-0.4) to avoid inversion parameters exceeding the actual range of the project.

[0084] The a priori value range of the transverse anisotropy parameter (deflection modulus ratio Eh / Ev) is set between 0.5 and 2.0 based on the empirical values ​​of regional soil type and pavement structure material (such as cement-stabilized crushed stone and graded crushed stone), and is optimized as a parameter to be determined during the inversion iteration process.

[0085] (b) Inversion solution: Objective function: The displacement residuals in the time domain and frequency domain are used as a joint measure, i.e., residual = ω1 × |measured deflection time history - simulated deflection time history| + ω2 × |measured deflection frequency domain characteristics - simulated deflection frequency domain characteristics|, where ω1 and ω2 are weighting coefficients that are dynamically adjusted according to the signal-to-noise ratio. When the signal-to-noise ratio is low, the weight of ω2 is increased. At the same time, an L1 regularization term is introduced to suppress parameter oscillations. Optimization algorithm: A hybrid strategy combining global search (simulated annealing algorithm) and local optimization (Levenberg-Marquardt algorithm) is adopted. First, the simulated annealing algorithm is used to traverse the global parameter space to find an approximate optimal solution, and then the Levenberg-Marquardt algorithm is used for local fine optimization. Convergence criterion: Stop the inversion when the relative change in the objective function value is less than 1e-6, or when there is no significant improvement in the parameters for three consecutive iterations (which can be set to a change of <0.1%). Temperature correction: By combining the average road surface temperature data collected by temperature sensors over 24 hours, the equivalent modulus of the surface layer obtained by inversion is corrected for temperature, eliminating the influence of temperature on the stiffness of the asphalt surface layer and ensuring the accuracy of the modulus assessment.

[0086] (c) Output of results: Output the thickness, equivalent modulus (E), and deflection modulus ratio (Eh / Ev) of each structural layer, and generate a sensitivity report to analyze the response of each parameter (such as layer thickness and Eh / Ev) to changes in deflection (B) / modulus (E) and identify key influencing parameters.

[0087] (3) Calculation of deflection basin index Data collected by falling weight deflectometer The multi-base deflection sequence is automatically calculated by the deflection basin index calculation module: Surface Index It reflects the integrity of the surface layer structure and is sensitive to damage to the surface layer; Grassroots Index It reflects the carrying capacity of the grassroots level and is sensitive to damage to the grassroots level; Distal deflection It reflects the stiffness of the roadbed and is sensitive to roadbed damage; Comparison of working conditions: If deflection data are collected under different load levels or temperature conditions, calculate the deflection under each working condition. , , The indicators are compared and analyzed to enhance the reliability of stratum fault indication. For example, if the BDI is significantly high under a certain working condition, the focus can be on investigating stratum-level defects.

[0088] IV. Composite endpoint classification (S3) Based on the output of S2 above, a composite endpoint classification is performed, and the disease level and corresponding treatment recommendations are output.

[0089] The specific classification of the above composite endpoints is as follows: In the template matching of S2 above, a similarity score is calculated. A similarity threshold is set based on the data in the historical engineering case library. When the similarity score between the measured data and the three-dimensional slice template of the above candidate disease exceeds the above similarity threshold, the measured data is considered to be the above candidate disease. The above disease categories are divided into moderately loose, severely loose, spherical or rectangular cavities, and severely water-rich.

[0090] Based on the calculation results of the multi-base deflection sequence in S2 above, judgment thresholds for the above surface layer condition index, base layer condition index, and subgrade condition index are set respectively. The judgment thresholds are calibrated by regression using the regional specifications and historical data of the road under test. When the above surface layer condition index is greater than the above judgment threshold, the above pavement is considered to be damaged. When the above base layer condition index is greater than the above judgment threshold, the above base layer is considered to be damaged. When the above subgrade condition index is greater than the above judgment threshold, the above subgrade is considered to be damaged.

[0091] If at least one of the above-mentioned disease categories exists, namely severe loosening, spherical or rectangular cavities, or severe water retention, and at least one of the above-mentioned base course damage or roadbed damage exists, the disease level is determined to be Level III, and the corresponding remediation recommendation is: immediately close the road and prioritize remediation. When the above-mentioned defects are classified as moderately loose and the base layer is damaged, or both the pavement and the base layer are damaged, or the above-mentioned defects are classified as severely loose and the pavement is damaged, the defect level is determined to be Level II, and the corresponding remediation recommendation is: remediation within a time limit to prevent the spread of defects. In other cases, the above-mentioned disease level is determined to be Level I, and the corresponding treatment recommendations are: include it in routine monitoring and conduct regular re-examinations.

[0092] (1) Threshold setting Call the classification threshold after regression calibration based on the specifications and historical data of the road area to be tested (such as...). >130、 >30、 >80 (units are in micrometers as specified by the equipment), and calibrated in conjunction with historical calibration data of the road grade in the permafrost region to ensure that the threshold conforms to the actual engineering situation in the area; if there is previous diagnostic data for the road section, the threshold can be further optimized through regression analysis of historical data to improve the accuracy of the classification.

[0093] The grading threshold here (e.g.) >130、 >30、 The threshold accuracy (>80) is derived from statistical regression analysis of falling weight deflectometer data from over 50 test sections of typical highways in North China, totaling 200 kilometers in length, and subsequent core sampling and excavation verification of actual road damage levels. The decision accuracy has been cross-validated to be greater than 90%. The system supports adaptive calibration of thresholds by importing corresponding calibration datasets based on different regions or road grades.

[0094] (2) Grading determination Based on the damage categories (moderately loose, severely loose, severely water-rich, spherical or rectangular cavities) obtained from 3D ground-penetrating radar template matching and the structural bearing capacity indices (pavement damage, base course damage, subgrade damage, etc.) obtained from falling weight deflectometer dynamic inversion and index calculation, a predefined composite grading rule is applied for determination: Level III: Meets any of the following conditions: the disease category is "severely loose," and the structural bearing capacity index is "base course damaged" (BDI>30) or "subgrade damaged" (…). >80); the disease category is "severe waterlogging", and the structural bearing capacity index is "base course damaged" (BDI>30) or "subgrade damaged" ( >80); the disease category is "spherical void or rectangular void", and the structural bearing capacity index is "base layer damaged" (BDI>30) or "subgrade damaged" ( >80). Recommended action: Immediately close the affected section of road to traffic and organize emergency remediation (such as roadbed drainage and base layer replacement) to prevent road collapse from causing a safety accident.

[0095] Level II: Meets the following conditions: The damage category is "moderately loose," and the structural bearing capacity index is "base course damaged" (BDI>30); the damage category is "moderately loose," and the structural bearing capacity index is both "base course damaged" (BDI>30) and "pavement damaged" (…). >130); the disease category is "severely loose", and the structural bearing capacity index is "pavement damaged" ( >130). Recommendations: Develop a specific treatment plan to reinforce or repair the base layer within a short period (e.g., 1-3 months) to prevent further development of the disease.

[0096] In other cases, it is classified as Level I. Recommended treatment: Record the location and characteristics of the damage, include it in routine inspections, and do not conduct any special treatment at this time.

[0097] V. Diagnostic Report Generation (S4) Based on the output results of S2 and S3 above, a diagnostic report is generated.

[0098] The generation of the above diagnostic report specifically includes: Data Collection: The above data includes the above disease categories, the above geometric scale estimates, the above layer thicknesses, the above equivalent moduli, the above multi-base deflection sequence calculation results, the above disease levels, and corresponding treatment recommendations; Data visualization generation: Fill the numerical data in the above data into the preset parameter table, render the geometric scale estimation data in the above data and the positioning data in S1 into a three-dimensional voxel positioning map, and draw the multi-base deflection sequence calculation results in the above data into a radar map. Data integration: The above parameter tables, 3D voxel localization maps, radar charts, and treatment recommendations are integrated and formatted according to the preset report template to generate a diagnostic report file in PDF or HTML format; Data presentation: The diagnostic report documents are combined with the geographic information system to visualize the spatial distribution and location of the above-mentioned defects in the road, and all data are archived in the historical engineering case library.

[0099] The report generation module automatically integrates all data and results from the entire process to generate a diagnostic report for permafrost road diseases. The core content includes: Project Overview: Location, length, road grade, structural layer design parameters (surface / base layer thickness, material type) of the road section to be tested, and environmental characteristics of the permafrost region (average annual temperature, number of freeze-thaw cycles). Testing equipment and parameters: Models and key parameters of 3D ground penetrating radar, falling weight deflectometer, and synchronous positioning module; load energy level, radar frequency, and sampling step length during data acquisition.

[0100] Data processing results: 3D ground-penetrating radar data results: 3D voxel location map (labeled with plane coordinates, depth, and geometric scale of the disease), disease type (e.g., "severely loose base layer in section K1+200-K1+300") and geometric scale (e.g., "loose area range 2m×1.5m×0.3m"). Insertion is possible. Figure 6 The diagram shown illustrates the extraction of three-dimensional data volume slices. Figures 7-10 The image shows a comparison of three-dimensional slice templates for typical diseases. Falling weight deflectometer data results: Inversion parameter table for each structural layer (layer thickness, equivalent modulus, Eh / Ev), sensitivity analysis chart, deflection basin curve and , , Indicator radar chart; Grading conclusions: Grading results and judgment criteria for each road section with defects; Recommendations: Provide targeted recommendations based on the classification results, including the timing of remediation, technical solutions (such as base grouting reinforcement and subgrade drainage), closure and control requirements, and subsequent monitoring plans; Attachments: Screenshots of original collected data, template matching similarity calculation table, inversion convergence curve, on-site detection photos, and other supporting materials; After the report is generated, it can be exported as a PDF / HTML file. The diagnostic report file is then combined with a geographic information system to visualize the spatial distribution and location of the road defects. The report is archived in a historical engineering case library and simultaneously pushed to the road maintenance department to provide a basis for maintenance decisions.

Claims

1. A rapid diagnostic method integrating three-dimensional GPR-FWD for hidden defects in frozen soil roads, characterized in that, Includes the following steps: S1. Synchronously collect pavement-base-subgrade data and multi-base distance deflection sequence of the road to be tested; the multi-base distance deflection sequence is as follows: ,in Deflection at the point of application of the load. Deflection at the distant measuring point; S2. Call the electromagnetic forward modeling library, generate a three-dimensional slice template with the candidate defects as the simulation object, and then perform template matching between the three-dimensional slice template and the pavement-base-subgrade data, and output the defect category, geometric scale parameters of the defects and similarity score of the pavement-base-subgrade data. Based on the multi-base deflection sequence, a falling weight deflectometer dynamic inversion is performed on the pavement-base-subgrade of the road under test, and the layer thickness and equivalent modulus are output. Based on the multi-base deflection sequence, the surface layer condition index, base layer condition index, and subgrade condition index are calculated and output. The surface condition index is: , The The deflection values ​​of the near-distance measuring points in the multi-base deflection sequence; the base condition index is... , The , The deflection value at the mid-distance measuring point in the multi-base deflection sequence; the subgrade condition index is... ; S3. Based on the disease category, geometric scale parameters and similarity score of the disease; layer thickness and equivalent modulus; surface layer condition index, base layer condition index, and subgrade condition index, perform composite endpoint classification, and output disease level and corresponding treatment suggestions; S4. Based on the output results of S2 and S3, generate a diagnostic report.

2. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in permafrost roads according to claim 1, characterized in that, In S1, the road surface-base-subgrade data is collected by a three-dimensional ground-penetrating radar. The radar excitation of the three-dimensional ground-penetrating radar adopts Ricker wavelet or Gaussian modulated sine wave. The transmitting antenna and receiving antenna of the three-dimensional ground-penetrating radar are equidistantly arranged along the survey line.

3. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in permafrost roads according to claim 1, characterized in that, The electromagnetic forward modeling library called in S2 is constructed based on the finite-difference time-domain method of a single-axis anisotropic fully matched layer, and the following requirements must be met when calling it: The spatial grid step size ranges from 1 / 20 to 1 / 10 of the wavelength; the wavelength is the wavelength of the electromagnetic wave corresponding to the radar center frequency within the material with the lowest dielectric constant in the road structure under test. The time step satisfies the Courant stability condition, wherein the Courant number of the Courant stability condition ranges from 0.5 to 0.95; The boundary thickness of the uniaxial anisotropic fully matched layer is not less than 10 of the spatial grids, so that the false reflection energy at the boundary of the uniaxial anisotropic fully matched layer is suppressed to below -80dB.

4. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in frozen soil roads according to claim 1, characterized in that, In step S2, the candidate diseases include spherical cavities, rectangular cavities, loose areas, and water-rich layers. When generating the three-dimensional slice template, the dielectric constant, conductivity, geometric parameters, porosity, and water content of the candidate diseases are discretized and sampled to construct a template library. The geometric parameters include three-dimensional morphological parameters, the spatial distribution range of the loose areas, and the vertical thickness of the water-rich layers. During the template matching process, a model including an attention mechanism and a feature extraction module is constructed to analyze the pavement-base-subgrade data and the multi-base deflection sequence, and the template matching similarity score is calculated by combining the weighted combination of normalized cross-correlation and structural similarity.

5. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in frozen soil roads according to claim 1, characterized in that, In step S2, the dynamic inversion of the falling weight deflectometer considers the transverse anisotropy of the permafrost road structure, specifically including: A dynamic spectrum unit model of a layered structure is established, and the impact load time history and measuring point spacing of the falling weight deflectometer are input. The pavement structure layer of the road to be tested is defined as a transversely isotropic medium. A priori range is set for the parameters to be determined by the dynamic inversion of the falling weight deflectometer; the parameters to be determined include the layer thickness, equivalent modulus and Poisson's ratio of the pavement-base-subgrade of the road to be tested; The objective function for the dynamic inversion of the falling weight deflectometer is the time-domain displacement residual, the frequency-domain displacement residual, and... The weighted sum of the regularization terms is solved using a combination of global search and local optimization.

6. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in permafrost roads according to claim 1, characterized in that, In S1, while simultaneously collecting pavement-base-subgrade data and multi-base deflection sequences of the road under test, pavement temperature is also collected simultaneously. In step S2, the road surface temperature is combined with the average temperature of the road under test over 24 hours, and the temperature of the asphalt layer mid-surface depth is predicted using an empirical model. The predicted temperature of the asphalt layer mid-surface depth is used as a standard reference temperature, and the equivalent modulus is corrected using a reference temperature correction formula to obtain the corrected equivalent modulus. Generate a sensitivity report of the corrected equivalent modulus to changes in the equivalent modulus.

7. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in permafrost roads according to claim 1, characterized in that, In S3, the specific rules for the composite endpoint classification are as follows: A similarity threshold is set based on historical engineering case database data; when the similarity score exceeds the similarity threshold, it is determined that there are corresponding candidate defects in the pavement-base-subgrade of the road to be tested; the defect categories are divided into moderate looseness, severe looseness, voids, and severe waterlogging; The judgment thresholds for the surface layer condition index, base layer condition index, and subgrade condition index are set respectively, and the judgment thresholds are calibrated based on the regional specifications and historical data of the road under test. The surface layer condition index, base layer condition index, and subgrade condition index are compared with their respective judgment thresholds. When the condition index of any structural layer is greater than the judgment threshold of that structural layer, the structural layer is judged to be damaged. When the disease category is at least one of severe loosening, voiding, or severe water accumulation, and the base course or subgrade is damaged, the disease level is determined to be Level III, and the corresponding treatment recommendation is: immediately close the road and prioritize treatment. When any of the following conditions are met: moderate looseness with damage to the base layer, damage to both the road surface and the base layer, or severe looseness with damage to the road surface, the disease level is determined to be Level II. The corresponding treatment recommendation is: treatment within a specified period to prevent the spread of the disease. In other cases, the disease level is determined to be Level I, and the corresponding treatment recommendation is: include it in routine monitoring and conduct regular re-examinations.

8. The three-dimensional GPR-FWD integrated rapid diagnosis method for hidden defects in permafrost roads according to claim 1, characterized in that, In step S4, the generation of the diagnostic report specifically includes: Data collection: The data includes the disease categories, geometric scale parameters of the diseases, similarity scores, layer thickness, equivalent modulus, and calculation results of the multi-base deflection sequence output by S2, as well as the disease levels and corresponding treatment suggestions output by S3. Data visualization generation: Fill the numerical data in the data into a preset parameter table; associate the geometric scale parameters with the positioning data collected synchronously in S1, render a three-dimensional voxel positioning map, and draw the calculation results of the multi-base deflection sequence as a radar chart; Data integration: The parameter tables, three-dimensional voxel positioning maps, radar charts, disease levels, and treatment recommendations are integrated according to a preset report template to generate a diagnostic report file in PDF or HTML format; Data presentation: The diagnostic report file is combined with a geographic information system to visualize the spatial distribution and location of the defects in the road, and the diagnostic report is archived in the historical engineering case library.

9. A rapid diagnostic system for hidden defects in frozen soil roads using a three-dimensional GPR-FWD integrated approach, characterized in that... include: The data acquisition subsystem is used to perform the following steps: S1. Synchronously collect pavement-base-subgrade data and multi-base distance deflection sequence of the road to be tested; the multi-base distance deflection sequence is as follows: ,in Deflection at the point of application of the load. Deflection at the distant measuring point; The computational processing subsystem is used to perform the following steps: S2. Call the electromagnetic forward modeling library, generate a three-dimensional slice template with the candidate defects as the simulation object, and then perform template matching between the three-dimensional slice template and the pavement-base-subgrade data, and output the defect category, geometric scale parameters of the defects and similarity score of the pavement-base-subgrade data. S3. Based on the multi-base deflection sequence, perform dynamic inversion of the pavement-base-subgrade body of the road under test using a falling weight deflectometer, and output the layer thickness and equivalent modulus. Based on the multi-base deflection sequence, the surface layer condition index, base layer condition index, and subgrade condition index are calculated and output. The surface condition index is: , The The deflection values ​​of the near-distance measuring points in the multi-base deflection sequence; the base condition index is... , The , The deflection value at the mid-distance measuring point in the multi-base deflection sequence; the subgrade condition index is... ; Based on the disease category, geometric scale parameters and similarity score of the disease; layer thickness and equivalent modulus; surface layer condition index, base layer condition index, and subgrade condition index, a composite endpoint classification is performed, and the disease level and corresponding treatment suggestions are output. S4. Based on the output results of S2 and S3, generate a diagnostic report; A data storage subsystem is used to store the data required for the operation of the computing and processing subsystem.

10. The three-dimensional GPR-FWD integrated rapid diagnostic system for hidden defects in frozen soil roads according to claim 9, characterized in that, The data acquisition subsystem includes a three-dimensional ground-penetrating radar array, a falling-weight deflectometer loading device, a synchronous positioning module, and a temperature sensor module; the synchronous positioning module includes a global navigation satellite system and an inertial measurement unit module, supporting second pulse triggering and a unified time scale; The computing subsystem also includes a graphics processor, on which the step of generating the three-dimensional slice template is executed in parallel. The data acquisition subsystem is integrated into a vehicle-mounted or semi-vehicle-mounted platform, and the platform is equipped with an engineering support module, which includes: The power supply module for the work vehicle supplies power to each of the subsystems. A safety control auxiliary module, which deploys traffic safety commands on the road to be tested.

Images