Road repair state monitoring method and system based on edge calculation

Abstract

The invention relates to the technical field of road repair state monitoring, in particular to a road repair state monitoring method and system based on edge calculation. The method comprises the following steps: acquiring planning information of a repaired road, deploying 3D laser scanning vibration meters at a plurality of repairing points, collecting pavement vibration signals in different time periods, and performing disordered distribution time domain feature analysis to obtain a vibration signal disordered distribution time domain graph; then, structure weakening behavior simulation is carried out based on the graph, periodic structure weakening data is quantified, and a pavement water damage cumulative gradient is obtained through water damage cumulative gradient estimation; and finally, in combination with the periodic structure weakening data and the water damage cumulative gradient, carrying out service life evaluation on the restoration state, and sending an evaluation result to the terminal. According to the invention, the road repair state monitoring technology is optimized, so that the road repair state monitoring technology is more perfect.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of road repair state monitoring, and particularly relates to a road repair state monitoring method and system based on edge computing. BACKGROUND

[0002] The road repair state monitoring method based on intelligent sensing has gradually become an important means to improve the efficiency of road management. As a new computing mode, edge computing can process and analyze data at the source of data generation, i.e., the near end, avoiding the delay and bandwidth pressure caused by large-scale data transmission to the remote server, thereby realizing efficient and low-latency real-time monitoring and early warning. By combining edge computing and pavement vibration monitoring technology, through the deployment of various sensors (such as 3D laser scanning vibration meters and accelerometers) in the road repair area, the vibration signals, deformation conditions and environmental changes of the road can be collected in real time. After the data are processed and preliminarily analyzed by the edge computing nodes, they can be quickly fed back to the relevant management personnel, helping to evaluate the effect and quality of road repair in real time, timely discovering potential problems, and effectively intervening and repairing. However, the traditional road repair state monitoring method based on edge computing has the problem of inaccurate analysis of road damage state, thereby causing large errors in the evaluation of road repair state life. SUMMARY

[0003] Therefore, it is necessary to provide a road repair state monitoring method and system based on edge computing to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a road repair state monitoring method based on edge computing comprises the following steps: Step S1: acquiring road repair planning information of repaired roads; extracting a plurality of repaired road defect positions in the planning information; deploying 3D laser scanning vibration meters at the plurality of repaired road defect positions to collect road surface vibration signals of a plurality of traffic periods, and then performing disordered distribution time domain feature analysis to obtain a vibration signal disordered distribution time domain graph; Step S2: performing structure weakening behavior simulation quantification during a use cycle based on the vibration signal disordered distribution time domain graph to obtain cycle structure weakening data; and performing water damage accumulation gradient estimation based on the cycle structure weakening data to obtain a road surface water damage accumulation gradient; Step S3: performing repair state service life evaluation on the cycle structure weakening data and the road surface water damage accumulation gradient based on a reinforcement learning algorithm possessed by an edge computing platform to obtain repair state service life evaluation data; and sending the repair state service life evaluation data to a terminal.

[0005] Preferably, step S1 comprises the following steps: Step S11: Obtain the repaired road repair planning information; extract a plurality of repaired road defect positions in the planning information; Step S12: Deploy 3D laser scanning vibration meters at the plurality of repaired road defect positions to collect road surface vibration signals in a plurality of traffic periods; Step S13: Perform average filtering processing on the road surface vibration signals to obtain road surface vibration filtered signals; Step S14: Perform unordered distribution time domain feature analysis on the road surface vibration filtered signals to obtain an unordered distribution time domain graph of the vibration signals.

[0006] Preferably, step S2 comprises the following steps: Step S21: Extract quality repair design requirements of a plurality of repaired road defect positions according to the repaired road repair planning information, to obtain quality repair design requirements of the plurality of defect positions; Step S22: Perform vertical load intensity conversion on the unordered distribution time domain graph of the vibration signals to obtain vibration vertical load intensity data; Step S23: Simulate and quantify the structural weakening behavior of the quality repair design requirements of the plurality of defect positions based on the vibration vertical load intensity data to obtain periodic structural weakening data; Step S24: Estimate water damage accumulation gradient according to the periodic structural weakening data and the vibration vertical load intensity data to obtain road surface water damage accumulation gradient.

[0007] Preferably, step S22 comprises the following steps: Step S221: Perform amplitude change rate analysis on the unordered distribution time domain graph of the vibration signals to obtain unordered amplitude change rate; Step S222: Perform low-to-high jump slope increment approximate linear correlation analysis on the unordered distribution time domain graph of the vibration signals according to the unordered amplitude change rate to obtain slope increment approximate correlation data; Step S223: Perform multi-scale spectrum entropy decomposition processing on the unordered distribution time domain graph of the vibration signals based on the slope increment approximate correlation data to obtain time series energy spectrum entropy decomposition data; Step S224: Perform equivalent vertical load mechanical state mapping on the time series energy spectrum entropy decomposition data to obtain vertical load mechanical state; Step S225: Perform vertical load intensity conversion according to the vertical load mechanical state to obtain vibration vertical load intensity data.

[0008] Preferably, step S23 comprises the following steps: Step S231: Extract subgrade bearing capacity, compactness and thickness in the quality repair design requirements of the plurality of defect positions; Step S232: coupling non-linear effect to the vibration vertical load intensity data to generate load intensity non-linear coupling data; Step S233: simulating and analyzing the base layer shear deformation mechanical state of the roadbed bearing capacity, density and thickness according to the load intensity non-linear coupling data to obtain the base layer shear deformation mechanical state; Step S234: analyzing the joint adhesion failure degree caused by the base layer shear deformation mechanical state; Step S235: quantifying the periodic structure weakening behavior simulation based on the base layer shear deformation mechanical state and the joint adhesion failure degree to obtain periodic structure weakening data.

[0009] Preferably, step S234 includes the following steps: obtaining the basic characteristics of the repair material; analyzing the material interface adhesion strength based on the basic characteristics of the repair material to obtain the material interface adhesion strength; analyzing the oblique shear stress based on the base layer shear deformation mechanical state to generate the base layer oblique shear stress intensity; simulating the gradient attenuation state along the joint depth based on the base layer oblique shear stress intensity; correlating and identifying the surface interface slip mode of the material joint based on the gradient attenuation state and the material interface adhesion strength to obtain the surface interface slip mode; analyzing the joint adhesion failure based on the surface interface slip mode to obtain the joint adhesion failure degree.

[0010] Preferably, step S24 includes the following steps: Step S241: simulating the base layer crack evolution structure based on the periodic structure weakening data and the multi-scale base layer crack evolution structure to obtain the base layer crack evolution structure; Step S242: mapping the maximum water intrusion value corresponding to the base layer crack evolution structure based on the base layer crack evolution structure to obtain the maximum water intrusion value corresponding to the base layer crack evolution structure; Step S243: analyzing the base layer swelling / softening coupling effect based on the maximum water intrusion value and the base layer crack evolution structure to obtain the base layer swelling / softening coupling effect data; Step S244: estimating the water damage accumulation gradient based on the maximum water intrusion value, the base layer crack evolution structure and the base layer swelling / softening coupling effect data to obtain the road surface water damage accumulation gradient.

[0011] Preferably, step S3 includes the following steps: Step S31: normalizing the periodic structure weakening data and the road surface water damage accumulation gradient respectively to obtain the periodic structure weakening normalized data and the road surface water damage accumulation normalized gradient respectively; Step S32: repair state service life evaluation based on the periodic structure weakening normalized data and the pavement water damage accumulation normalized gradient using the reinforcement learning algorithm possessed by the edge computing platform, to obtain repair state service life evaluation data; Step S33: send the repair state service life evaluation data to the terminal.

[0012] Preferably, the present application also provides an edge computing-based road repair state monitoring system for performing the edge computing-based road repair state monitoring method as described above, which comprises: a vibration signal analysis module for obtaining repaired road repair planning information; extracting a plurality of repaired road defect positions in the planning information; deploying 3D laser scanning vibration meters at the plurality of repaired road defect positions to collect pavement vibration signals of a plurality of traffic periods, and then performing unordered distribution time domain feature analysis to obtain a vibration signal unordered distribution time domain graph; a damage analysis module for simulating and quantifying the structure weakening behavior during the service period based on the vibration signal unordered distribution time domain graph to obtain periodic structure weakening data; and estimating the water damage accumulation gradient based on the periodic structure weakening data to obtain the pavement water damage accumulation gradient; a state service life evaluation module for performing repair state service life evaluation based on the periodic structure weakening data and the pavement water damage accumulation gradient using the reinforcement learning algorithm possessed by the edge computing platform, to obtain repair state service life evaluation data; and sending the repair state service life evaluation data to the terminal.

[0013] The present invention provides the following beneficial effects: by acquiring repair planning information for repaired roads, accurately identifying multiple defect locations within the repaired area, and deploying 3D laser scanning vibrometers to collect pavement vibration signals at different time periods. This approach enables comprehensive monitoring of the dynamic response of repaired roads, particularly the impact of road defects on traffic conditions. By analyzing the collected vibration signals' disordered time-domain characteristics, a disordered time-domain map of the vibration signals is generated. This helps accurately identify the spatiotemporal characteristics of road vibration and the evolution of defects, providing an important basis for subsequent structural weakening and repair status assessment. This process improves the real-time and accuracy of post-repair road monitoring. Based on the disordered time-domain map of the vibration signals, the structural weakening behavior over the service life is further simulated and quantified to obtain periodic structural weakening data. This step quantifies the weakening of roads over long-term use due to traffic loads, climate change, and other environmental factors. Combined with the periodic structural weakening data, the cumulative water damage gradient is estimated to produce the pavement water damage cumulative gradient. This process effectively reflects the changes in road durability under water erosion, further revealing the cumulative effects of water damage on repaired roads, facilitating early detection of potential water damage issues, enabling timely remediation measures and extending road service life. On the edge computing platform, a reinforcement learning algorithm is used to evaluate the repair status and service life of the periodic structural weakening data and the cumulative gradient of water damage. Reinforcement learning can optimize the repair status evaluation model through autonomous learning and accurately predict the long-term performance of roads under different environmental and traffic conditions. This process not only improves the accuracy of the repair status evaluation, but also enables the evaluation process to be gradually optimized as data accumulates, providing dynamic support for road repair decisions. The evaluation results are ultimately fed back to management personnel in real time through terminal equipment, ensuring that road management parties can grasp the repair effect and service life in a timely manner, providing a scientific basis for future road maintenance and resource allocation, reducing maintenance costs, and improving the overall efficiency of road management. Therefore, the present invention is an optimization of a traditional road repair status monitoring method based on edge computing, which solves the problem that a traditional road repair status monitoring method based on edge computing has inaccurate analysis of road loss status, resulting in large errors in road repair status life evaluation, improves the accuracy of road loss status analysis, and reduces the error in road repair status life evaluation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A schematic diagram of the steps of a road repair condition monitoring method based on edge computing; Figure 2 for Figure 1 Detailed implementation steps of step S2 in FIG. Figure 3 for Figure 1 Detailed implementation steps of step S3 in FIG. DETAILED DESCRIPTION

[0015] Referring to Figures 1 to 3 A road repair state monitoring method based on edge computing, the method comprising the following steps: Step S1: Obtain repaired road repair planning information; extract a plurality of repaired road defect positions in the planning information; deploy 3D laser scanning vibration meters at the plurality of repaired road defect positions, thereby collecting road surface vibration signals of a plurality of traffic periods, and then performing disordered distribution time domain feature analysis, thereby obtaining a vibration signal disordered distribution time domain graph; Step S2: Based on the vibration signal disordered distribution time domain graph, simulate and quantify the structure weakening behavior during the use cycle to obtain cycle structure weakening data; estimate the water damage accumulation gradient according to the cycle structure weakening data, thereby obtaining the road surface water damage accumulation gradient; Step S3: Based on the reinforcement learning algorithm possessed in the edge computing platform, evaluate the repair state service life of the cycle structure weakening data and the road surface water damage accumulation gradient, obtain repair state service life evaluation data; send the repair state service life evaluation data to the terminal.

[0016] In the embodiment of the application, reference Figure 1 As shown in the figure, it is a step flowchart of a road repair state monitoring method based on edge computing, in this example, the road repair state monitoring method based on edge computing comprises the following steps: Step S1: Obtain repaired road repair planning information; extract a plurality of repaired road defect positions in the planning information; deploy 3D laser scanning vibration meters at the plurality of repaired road defect positions, thereby collecting road surface vibration signals of a plurality of traffic periods, and then performing disordered distribution time domain feature analysis, thereby obtaining a vibration signal disordered distribution time domain graph; In the embodiment of the application, first, GIS format repair planning information is extracted from the road repair completion archives issued by the municipal road maintenance unit, which should include the start and end latitude and longitude of road repair, construction repair time, defect type, repair technology type, repair thickness and structural layer material type and other structural parameters, after loading the repair rule information layer in ArcGIS 10.7, a plurality of independent defect repair sections are extracted through road chain node numbering, and the geographic location information is exported as CSV format for subsequent device deployment operation, RIEGL VZ-400i type 3D laser scanning vibration meter is deployed in the plurality of defect repair sections, shockproof tripod is set at the deployment point position, and level is adjusted using the level, so that the device remains stable, 3D laser scanning vibration meter and the ground are fixed by high-strength spiral expansion bolts to prevent deviation or vibration interference during measurement, each 3D laser scanning vibration meter needs to be adjusted in angle in a way that is perpendicular to the road surface and the optical axis points to the main direction of vehicle operation. After preliminary alignment using the built-in laser collimator and IMU sensor, the optimal angle is fine-tuned through the feedback of the echo intensity of the real-time collected signal, in order to realize the three-dimensional space unification of multi-point data, GNSS reference points are arranged at both ends of each repair section before deployment, and the coordinates are calibrated using the total station or RTK, and the accurate three-dimensional coordinates of the vibration meter deployment point are recorded. The device has a spatial resolution of 5mm and a sampling frequency of 5000Hz, by continuously collecting road vibration signal data for 30 minutes at 3 traffic periods, i.e. from 7am to 9am, from 12pm to 2pm and from 5pm to 7pm, the original vibration data is stored in the local storage module as a three-dimensional acceleration sequence in XYZ direction, after each collection is completed, it is uploaded to the local edge server through the high-speed LAN interface for data preprocessing, in the preprocessing stage, Butterworth low-pass filter is used to filter the signal with a frequency cutoff of 100Hz to eliminate high-frequency noise, then in the Python environment, a non-uniform sampling time window is constructed using the NumPy library, each window width is 1 second and the step is 0.2 second, after normalizing the signal segment in each window, the local maximum difference sequence is used to extract the disorder degree characteristic value, finally, Matplotlib is used to draw the disorder distribution time domain graph of the vibration signal, which takes time axis as abscissa and disorder degree index as ordinate, and is used to depict the irregular distribution characteristics of the vibration response under the action of traffic load.

[0017] Step S2: Simulate and quantify the structural weakening behavior in the use cycle based on the vibration signal disorder distribution time domain graph to obtain cycle structure weakening data; estimate the water damage accumulation gradient according to the cycle structure weakening data to obtain the road water damage accumulation gradient; In the embodiment of the application, based on the obtained vibration signal disordered distribution time domain graph, the wavelet packet decomposition method is used in the MATLAB R2021a environment to extract the energy distribution of different frequency bands, 5 sub-bands are selected in the range of 5 Hz to 50 Hz, and the energy density change rate of each sub-band is respectively calculated and the periodic fitting residual is calculated. , wherein, represents the energy density of the jth sub-band at time t; represents the number of sampling points, represents the index of the sampling point, ranging from 1 to N, represents the index of the sub-band, ranging from 1 to 5 (5 sub-bands are selected); represents the amplitude of the jth sub-band at the ith sampling point; the variance of the periodic fitting residual is used as the structural periodic weakening index, which is calculated at each defect repair position to form the periodic structure weakening data, including position number, frequency band number, residual variance value, time stamp and other fields. Next, the inverse distance weighted interpolation method is used in the Python environment to perform spatial interpolation processing on the periodic structure weakening data to construct a complete road section structure weakening spatial heat map. On the basis of the map, the water damage cumulative gradient is further estimated. The water damage estimation adopts the soil parameter and structure response joint analysis method. In the section where the structural layer type is cement stabilized gravel layer, the permeability coefficient of the structural layer is set to m / s, and a moisture accumulation function is constructed according to the joint change rate of residual variance and time, and the function expression is: , wherein, represents the cumulative moisture content of position x at time t; represents the initial moisture content, the value range is 3-5%; represents the residual variance-moisture content conversion coefficient, the empirical value is: ; represents the time accumulation coefficient, represents the permeability coefficient correction coefficient, which is a calibration coefficient related to the material permeability coefficient (1.0×10⁻ 5 m / s), represents the differential of the thickness influence index, reflecting the nonlinear influence of thickness on moisture accumulation, α represents the coefficient of moisture migration efficiency, reflecting the response ability of the material to water flow, which is related to the physical properties of soil, environmental conditions (such as rainfall or evaporation) and other factors; the function is used to back calculate the moisture content change trend at different time nodes, and a layer-by-layer moisture accumulation distribution model is constructed combined with the structural layer thickness and drainage slope parameters, and finally the water damage cumulative gradient data of each defect position under different service periods are obtained, including spatial position number, time node, layer moisture content, total moisture gradient and other parameter fields.

[0018] Step S3: Based on the reinforcement learning algorithm possessed in the edge computing platform, the periodic structure weakening data and the pavement water damage cumulative gradient are used to evaluate the service life of the repair state, and repair state service life evaluation data is obtained; the repair state service life evaluation data is sent to the terminal.

[0019] In the embodiment of the application, first, the Pandas library is used to normalize the periodic structure weakening data and the pavement water damage cumulative gradient data, the normalization method adopts the maximum and minimum value standardization method, each group of data is compressed to the interval of 0 to 1, and the normalized data is input to the reinforcement learning analysis module deployed on the edge server in the form of a CSV file. The module is constructed based on TensorFlow, selects the DQN reinforcement learning algorithm to evaluate the state of the road structure service life, defines the state space as the normalized periodic weakening value and the water damage gradient value of each defect position, sets the action space as the state adjustment strategy of the repair level from 1 to 5 levels, designs the reward function as the minimum prediction repair life error and the actual monitoring life difference, the difference data is derived from historical working condition back analysis, the iteration number is set to 1000 in each round of training, the learning rate is set to 0.01, the discount factor is set to 0.9, after the training is completed, the trained strategy network is used to batch infer output for all defect sections, the prediction output is the remaining service period of each section, the unit is day, the output data is the repair state service life evaluation data, finally the evaluation data is packaged into JSON format and uploaded to the municipal road management terminal through the 4G communication module, the terminal visual interface displays the life data of each section in the form of a heat map and a time series chart for subsequent road maintenance decision making.

[0020] Step S1 includes the following steps: Step S11: Obtain the repaired road repair planning information; extract a plurality of repaired road defect positions in the planning information; Step S12: Deploy a 3D laser scanning vibration detector at the plurality of repaired road defect positions, thereby collecting pavement vibration signals in a plurality of traffic periods; Step S13: Perform average filtering processing on the pavement vibration signals to obtain pavement vibration filtered signals; Step S14: Perform unordered distribution time domain feature analysis on the pavement vibration filtered signals, thereby obtaining an unordered distribution time domain graph of the vibration signals.

[0021] In the embodiment of the present application, the road repair planning information is extracted from the "Road Repair Project Completion Information Registration Form" provided by the authorized road maintenance department and the supporting road management information system. This information is exported in XML format and includes road name, starting stake number, ending stake number, defect type number, repair time, structure layer type, repair method code, and structure thickness parameters. The XML file is parsed using the lxml library in Python, and all "structure repair completion" identification fields are identified and located using regular expressions. The corresponding road section start and end stake information is then converted to WGS84 coordinate system latitude and longitude format and exported in CSV format to guide the positioning and deployment of the vibration measurement equipment. Each extracted defect location must have a clear structure type label such as asphalt concrete or cement stabilized gravel, which is used as a precondition for structure response analysis in subsequent steps.

[0022] A three-dimensional laser scanning vibration meter of RIEGL VZ-400i is deployed on the above-mentioned extracted defect location points. This equipment has a spatial sampling frequency of 5000 Hz and a positioning error of less than 5 mm. Before deployment, the point coordinates are calibrated again using an RTK GPS receiver to ensure that the deployment accuracy is not less than 0.2 meters. The collection period is set to 30 minutes of continuous collection during each traffic peak period, with three collection periods per day: 7:00-9:00, 12:00-14:00, and 17:00-19:00. The raw data is stored in the form of three-axis direction acceleration time series, with a TXT format and 5000 groups of acceleration values collected per second. The data is transmitted to the edge server located on site for preprocessing using wired LAN connection. At least 3 days of continuous data collection is required for each defect point to ensure that the data time domain covers the complete short-term traffic disturbance characteristics. After collection is complete, the raw data is named according to the "defect number_date_time period" rule for file naming and archival management.

[0023] The original vibration signal is filtered by using a Python data processing script deployed in the edge server. The filtering operation uses a sliding window average method with a length of 50, that is, an average smoothing processing is performed every 50 consecutive sampling points. The moving step of the sliding window is set to 1. The processed acceleration signal removes the sudden high-frequency noise in the original data while retaining the true response of the vibration main frequency band. The processed data is stored in a JSON format, including fields such as timestamp, three-axis filtered acceleration value, and original file corresponding number. All time period data corresponding to each defect position is saved separately and marked as a filtered version to distinguish the signal source used in subsequent processing steps. In this process, no frequency domain processing such as Fourier transform or wavelet transform is performed to ensure the integrity of the time domain signal. All data is completed on the local edge node without the need for remote cloud transmission operation. First, a time window is constructed in NumPy for the filtered vibration signal, with a window length of 1 second corresponding to 5000 groups of data, and a window sliding step of 0.2 seconds. The standard deviation of the peak-to-peak distance difference of the local maximum value in each window is used as an index to describe the disorder. The local peaks are extracted by the find_peaks function in Scipy, with the minimum peak height limited to the mean value plus 0.5 times the standard deviation to eliminate weak disturbances. One disorder index value is obtained in each time window, and the disorder index sequence is arranged in time order. Then, the disorder distribution time domain graph is drawn using Matplotlib, with the horizontal axis representing time in seconds and the vertical axis representing the disorder index value of the corresponding window.

[0024] Step S2 includes the following steps: Step S21: Extracting quality repair design requirements for a plurality of repaired road defect positions according to the repaired road repair planning information, to obtain quality repair design requirements for the plurality of defect positions; Step S22: Converting the vibration signal disorder distribution time domain graph into vertical load intensity data; Step S23: Simulating and quantifying the structural weakening behavior of the quality repair design requirements for the plurality of defect positions based on the vibration vertical load intensity data, to obtain periodic structural weakening data; Step S24: Estimating the water damage accumulation gradient based on the periodic structural weakening data and the vibration vertical load intensity data, to obtain the road water damage accumulation gradient.

[0025] As an example of the present application, reference is made to Figure 2 In this example, the step S2 includes: Step S21: Extracting quality repair design requirements for a plurality of repaired road defect positions according to the repaired road repair planning information, to obtain quality repair design requirements for the plurality of defect positions; In the embodiment of the application, the road repair records with repair completion date not more than 180 days are retrieved in the road facility archive database, all item records with defect type identification code are extracted in the retrieval results, the construction design data is located according to the repair process field and the number information of the structure design file, the contents in the structure layer design specification table and the construction quality standard template are respectively called, the quality repair design requirements of each defect position are determined, including the total thickness of the structure layer, the thickness of each structure layer, the standard value of the roadbed bearing capacity, the target value of the surface layer and the base layer compaction degree, the standard value of the joint bonding strength, the design axle load passing frequency limit value, etc., in the extraction process, the repair mode code and the structure parameter field are one-to-one corresponding by using the Python dictionary mapping table based on the field mapping rule, taking the cement stabilized macadam base as an example, the design requirements include the base layer thickness ≥180mm, the compaction degree ≥98%, the unconfined compressive strength ≥2MPa, the interlayer bonding strength ≥0.3MPa, etc., such structures are commonly used in urban trunk roads or heavy load logistics channel sections, the quality repair design requirements of each defect section are finally structured into array data format and cached in the edge node for subsequent data comparison and analysis.

[0026] Step S22: vertical load intensity conversion is performed on the vibration signal disordered distribution time domain graph to obtain vibration vertical load intensity data; In the embodiment of the application, the vibration signal disordered distribution time domain graph obtained in the previous stage is deconstructed frame by frame, the timestamp and disordered degree value sequence corresponding to each frame of image are extracted, the pixel points in the curve track in the image are extracted using the OpenCV library in Python, the number of sampling points of each curve is set to be equidistant 100 points, and the sampling frequency is 50 Hz, then the amplitude change rate calculation method is used, the absolute value of the longitudinal coordinate change of each adjacent two points is divided by the horizontal coordinate time interval (0.02 seconds) as the amplitude change rate in unit time, after the change rate sequence of all points is constructed, a smoothing processing is performed using a sliding average window, the window length is 5 points (corresponding to a 0.1 second time window), then a slope increment approximation analysis is performed, a linear trend line is fitted once every 10 consecutive points (0.2 seconds) and the slope change increment is recorded, the slope sequence is normalized in the interval [0, 1] after the slope sequence is obtained, and is used as an input parameter for multi-scale spectrum entropy decomposition processing, the spectrum entropy decomposition processing is performed by EMD algorithm for intrinsic mode decomposition, the screening stop criterion is set to be that the SD value is less than 0.3, the energy spectrum of the first to third order modes is selected and the average spectrum entropy value is calculated, the average spectrum entropy value is used as a vibration energy disturbance level index, and then an equivalent mechanical state mapping operation is performed, the spectrum entropy value interval is divided into 5 levels: 0-0.2 corresponds to load level 1 (10kN), 0.2-0.4 corresponds to load level 2 (25kN), 0.4-0.6 corresponds to load level 3 (40kN), 0.6-0.8 corresponds to load level 4 (55kN), and 0.8-1.0 corresponds to load level 5 (70kN), the mapping reference table extracted from the data set (including 2000 groups of calibration samples) based on the historical traffic test field loading test is used to correspond the average spectrum entropy value and the load level one by one, and finally the vibration vertical load intensity data corresponding to each time point is obtained, the data update frequency is 10Hz, and the label identification is performed on the defect position coordinates.

[0027] Step S23: based on the vibration vertical load intensity data, the quality repair design requirements of the plurality of defect positions are simulated and quantified in the use cycle to obtain cycle structure weakening data; In the embodiment of the present application, in step S23, the vibration vertical load intensity data is organized into a three-dimensional data matrix according to the spatial position and time stamp, and the matrix size is the number of time periods (48 hours periods) x the number of measuring points (15-25 points per section) x the load intensity value (10-70 kN range). The data matrix is subjected to time-frequency analysis by Hilbert-Huang transform, and the empirical mode decomposition technique is used in the transform process (the stop condition is that the residual error is less than 0.05 or the iteration number reaches 12 times), and the instantaneous energy curve of each mode is extracted. The sampling frequency of the instantaneous energy curve is set to 200 Hz, the window width is 5 seconds, and the overlap rate is 75%. The instantaneous energy and the structure parameters are subjected to nonlinear fitting (Levenberg-Marquardt algorithm, convergence threshold 0.001), and the elastic modulus decay rate is calculated, and the decay rate formula parameters are set to the initial elastic modulus 300-500 MPa, the decay coefficient 0.02-0.08, and the power index 0.35-0.65. The fourth-order Runge-Kutta numerical integration method (step length 0.001 seconds) is used to calculate the shear strain accumulation curve, and the part of the cumulative shear strain exceeding the threshold value of 0.6% is defined as the structure weakening index value, and the structure weakening data table is generated; wherein the use cycle refers to a cycle after the road repair is completed and actually put into use, and the acquisition cycle in the embodiment of step S12 is set to continuously acquire 30 minutes in each traffic peak period, 3 time periods per day, 7:00-9:00, 12:00-14:00, and 17:00-19:00, and the data continuously acquired for 3 days is the cycle.

[0028] Step S24: water damage accumulation gradient estimation is performed according to the periodic structure weakening data and the vibration vertical load intensity data, so as to obtain the road surface water damage accumulation gradient.

[0029] In the embodiment of the application, on the basis of the obtained periodic structure weakening data and vibration vertical load intensity data, a two-dimensional finite difference grid is constructed for a typical structure layer profile by using the PDE toolbox in MATLAB, the grid size is set to 5 mm*5 mm, the structure profile is divided into four parts of asphalt surface layer, cement stabilized macadam base layer, sand gravel cushion layer and undisturbed soil layer, the shear modulus attenuation parameter is applied to the middle node of the base layer according to the periodic structure weakening data, the concentrated load simulation is applied at the corresponding time point of the vibration vertical load intensity, the load action range is set to a square area of 0.3 m*0.3 m, the vertical load intensity is taken as the estimated value at the corresponding position, under the time sequence of 72 hours of simulation period, the moisture diffusion analysis is carried out by using Fick's second law, the initial moisture content is set to 8%, the critical moisture content is set to 12%, the permeation boundary condition is applied to the bottom of the structure, the moisture gradient distribution data in the structure layer is output at each time interval, the moisture gradient growth rate at each time interval is integrated to construct a moisture intrusion cumulative layer, the data of each grid point of the layer is the water damage cumulative value per unit area, and then the water damage cumulative gradient value of each defect position is obtained by integrating all section grids, which is recorded as the road surface water damage cumulative gradient data, and used to support the subsequent service life evaluation task.

[0030] Step S22 comprises the following steps: Step S221: amplitude change rate analysis is performed on the vibration signal disordered distribution time domain graph to obtain disordered amplitude change rate; Step S222: low-to-high jump slope increment approximate linear correlation analysis is performed on the vibration signal disordered distribution time domain graph according to the disordered amplitude change rate to obtain slope increment approximate correlation data; Step S223: multi-scale spectrum entropy decomposition processing is performed on the vibration signal disordered distribution time domain graph based on the slope increment approximate correlation data to obtain time sequence energy spectrum entropy decomposition data; Step S224: equivalent vertical load mechanical state mapping is performed on the time sequence energy spectrum entropy decomposition data to obtain the vertical load mechanical state; Step S225: vertical load intensity conversion is performed according to the vertical load mechanical state to obtain vibration vertical load intensity data.

[0031] In the embodiment of the application, firstly, the disordered distribution time domain graph data of the vibration signal corresponding to each defect position is read in the edge computing platform, the graph data is derived from the disordered degree sequence image extracted from the three-dimensional acceleration signal after filtering and window decomposition in the previous stage, each frame of image corresponds to a time window vibration state change curve, the image is converted into two-dimensional coordinate data using the image gray extraction method in OpenCV, the horizontal coordinate is the time index point, and the vertical coordinate is the disordered degree value, then the amplitude change rate in unit time is calculated by dividing the vertical coordinate difference of adjacent points by the horizontal coordinate step, the horizontal coordinate step is 0.2 seconds, the sample interval is fixed, after executing this operation on all points, a complete amplitude change rate sequence is formed, the amplitude change rate is used as a speed index of the disordered degree change to reflect the intensity of short-time vibration energy disturbance under traffic load, in the process, the NumPy library is used for difference operation and unified standardization, the range normalization method is used for standardization to compress all change rate values to 0 to 1, the result data is called disordered amplitude change rate data, the processing flow is independently executed for each defect position and each collection period, and the obtained data is used as the input basis for the next slope increment analysis.

[0032] The processing flow of step S222 is to call the disordered amplitude change rate data obtained in step S221, segment each 100 data points, fit a linear regression curve in each segment by using the least square method, obtain the fitting slope value of each segment of data, sequentially extract the slope of all data segments to form a continuous slope sequence, then perform difference operation between adjacent two segments on the slope sequence to obtain the jump slope increment data, which represents the linear transition change characteristic of the vibration disturbance response, then the trend correlation detection is performed on the jump slope increment data, the Pearson correlation coefficient method is used to perform correlation analysis on the slope increment data and the time period number, the sequence segment with a correlation coefficient greater than 0.6 is selected as a significant jump correlation segment, and the start and end time window numbers are marked, further, local slope trend fitting is performed on these paragraphs to form slope increment approximate correlation data, the data structure includes time period number, start slope, end slope, average increment value, correlation coefficient value and other fields, the result reflects the time sequence relationship between the nonlinear aggravation trend of the vibration signal and the peak value change of the traffic load, and provides a basic variable input for energy disturbance aggregation degree estimation for spectrum entropy decomposition.

[0033] Based on the obtained slope increment approximation correlation data, the multi-scale spectrum entropy decomposition processing is performed on the corresponding vibration signal unordered distribution time domain graph. The processing method uses the empirical mode decomposition (EMD) algorithm. The amplitude change rate sequence of each time period is decomposed by calling the PyEMD module in Python. The decomposition layer is set to 3 layers, that is, the first order to the third order intrinsic mode function (IMF1 to IMF3) is extracted, which respectively represents the high-frequency short-period energy disturbance, the medium-frequency dynamic change and the low-frequency long-term trend. Each order of IMF function is further obtained by short-time Fourier transform to obtain its energy spectrum density function. After the probability normalization processing is performed on the function, the spectrum entropy value is calculated as the spectrum entropy data. The higher the obtained spectrum entropy value represents the more uniform energy distribution and the stronger the irregularity of disturbance, which reflects the discrete degree of traffic disturbance in time. After integrating the spectrum entropy values of all time periods, the complete time series energy spectrum entropy decomposition data is formed. The data structure includes time number, IMF order, energy density, spectrum entropy value, frequency domain main frequency information and other fields. Each defect position repeats this process and transmits the result data to the next stage of mechanical mapping analysis process.

[0034] On the basis of the time series energy spectrum entropy decomposition data, a vertical load mechanical state mapping rule table is established. The rule table establishes a mapping relationship between the typical spectrum entropy response formed by the traffic disturbance loading of the pavement structure and the vertical load mechanical response obtained by the in-situ plate loading test. During the data collection period, a standard plate loading test is conducted in the same defect type area. The test uses a 300 millimeter diameter circular steel plate to apply a concentrated load. The loading levels are 20kN, 40kN, 60kN and 80kN, each level lasts for 10 seconds, and the acceleration response signal is recorded synchronously and the spectrum entropy data is constructed. Finally, the spectrum entropy value distribution interval corresponding to each loading level is statistically calculated. For example, the spectrum entropy value between 0.1 and 0.2 corresponds to the low disturbance low load state, i.e. 20kN level, and the spectrum entropy value between 0.6 and 0.8 corresponds to 60kN level disturbance. The mapping process of spectrum entropy value to vertical load mechanical state is realized by table lookup operation. In actual data processing, each spectrum entropy data is classified and labeled according to the corresponding interval. The labeling result is the vertical load mechanical state value. The types are divided into 5 categories, namely extremely low load, low load, medium load, high load and extremely high load. The corresponding state number sequence is generated for each time period, which is used as input data for vertical load intensity conversion.

[0035] The processing flow of step S225 is to perform vertical load strength conversion operation based on the marked completed vertical load mechanical state, and the conversion process uses a standard load level calibration table as the basis for mapping, and the calibration table is derived from the results of standard load experiments in highway engineering. Each mechanical state number has a one-to-one mapping relationship with the actual equivalent static load strength. The state number 1 corresponds to an equivalent vertical load of 15 kN, the state number 2 corresponds to 30 kN, the state number 3 corresponds to 45 kN, the state number 4 corresponds to 60 kN, and the state number 5 corresponds to 75 kN. In data processing, the mechanical state number of each time period is mapped to its vertical load strength value according to the calibration table, and each load strength value is bound to a specific time period by combining the actual time span of the disordered time domain graph, to realize the precise numerical characterization of the mechanical action strength generated by traffic disturbance in the vibration signal. Finally, a complete vibration vertical load strength data sequence is generated, which includes time number, position number, mechanical state number, equivalent vertical load strength value, etc. fields, which are used as boundary loading inputs in subsequent structure weakening and water damage simulation.

[0036] In another embodiment, in step S221, the vibration signal disordered distribution time domain graph is processed by image processing technology to extract data. First, the Canny edge detection algorithm (low threshold 75, high threshold 150) is used to extract the vibration curve in the time domain graph, and the extracted image resolution is 1920x1080 pixels. After curve extraction, pixel coordinates are converted to physical quantities, the horizontal coordinate is mapped to time (accuracy 0.001 seconds), and the vertical coordinate is mapped to the disordered value (range 0-1). The curve is discretized by using equidistant resampling method, and the sampling interval is set to 0.01 seconds. Each curve is resampled to a fixed 100 points. The amplitude change rate of the resampled point sequence is calculated, and the calculation formula is the absolute value of the difference between the adjacent two points divided by the time interval: amplitude change rate a = |disordered value (i+1)-disordered value (i)| / (time (i+1)-time (i)). The amplitude change rate sequence obtained by calculation is smoothed by Hanning window function (window length 5 points, center weight 1.0, edge weight 0.25). The processed data is saved as an amplitude change rate matrix, and the matrix dimension is [sampling point number, defect position number].

[0037] In step S222, the unordered amplitude change rate data is first sorted by amplitude size, and the sorting uses a quick sorting algorithm with a time complexity of O(n log n). The sorted data is divided into 10 equal intervals, each interval containing an equal number of data points. The least square method is used to linearly fit the data points in each interval, and the fitting formula is amplitude change rate = k x time + b, where k is the slope and b is the intercept. The convergence condition is set to residual sum of squares less than 0.001 or the number of iterations reaching 50 times during linear fitting. The difference value of the slope value of adjacent intervals is calculated to obtain the slope increment data. The slope increment data is correlated with the original time series to form a time-slope increment two-dimensional data table. Each time point in the data table is associated with a slope increment value, and the increment value reflects the change acceleration of the vibration signal at that time. The slope increment data is filled with missing values by cubic spline interpolation method, and the interpolation node interval is 0.05 seconds, and the boundary condition is set to natural boundary condition (second derivative is zero).

[0038] In step S223, the slope increment approximate correlation data is input, and the empirical mode decomposition (EMD) method is used for modal separation, and the EMD process sets the iteration termination condition as the envelope mean value less than 0.05 or the iteration number reaches 12 times. Five intrinsic mode functions (IMF) are obtained by decomposition, and each IMF reflects the signal characteristics of different frequency scales. The instantaneous frequency and instantaneous amplitude are calculated by Hilbert transform for each IMF, and the transform is realized by fast Fourier transform with a sampling frequency of 200 Hz. The instantaneous energy is calculated by the square of the instantaneous amplitude to form a time-energy sequence. The spectral entropy value is calculated for the energy distribution in each time window (window width 5 seconds, overlap rate 75%). The calculation formula is the Shannon entropy formula: wherein Pi is the normalized energy of the ith frequency interval. The multi-scale spectrum entropy is formed by calculating the entropy value at different scales, with the scale factor set to 1 to 5, and finally forming the time sequence energy spectrum entropy decomposition data with a dimension of [number of time windows, number of scales, number of defect positions]. The time sequence energy spectrum entropy decomposition data is first normalized, and the maximum and minimum value normalization method is adopted to make the spectrum entropy value distributed between 0 and 1. The normalized spectrum entropy value and the vertical load mechanical state establish a mapping relationship, and the mapping is based on the pavement engineering test field cumulative load test database (containing 2500 test samples), which records the pavement vibration spectrum entropy characteristics under different loads. The mapping relationship is defined by a piecewise function: the spectrum entropy value 0-0.2 corresponds to the mechanical state 1 (elastic deformation stage), the spectrum entropy value 0.2-0.4 corresponds to the mechanical state 2 (elastic-plastic transition stage), the spectrum entropy value 0.4-0.6 corresponds to the mechanical state 3 (initial stage of plastic deformation), the spectrum entropy value 0.6-0.8 corresponds to the mechanical state 4 (mid-stage of plastic deformation), and the spectrum entropy value 0.8-1.0 corresponds to the mechanical state 5 (late stage of plastic deformation). The mapping adopts a hard classification method, and each spectrum entropy value is strictly mapped to a mechanical state, and there is no state overlap. The mapping result forms a time-position-mechanical state three-dimensional matrix, and the matrix storage adopts a sparse matrix format with a compression rate of 85%.

[0039] In the vertical load intensity conversion process according to the vertical load mechanical state, first, the load mechanical state level is converted into a reference load value, and the conversion relationship is: slight load corresponds to 10 kN, light load corresponds to 25 kN, moderate load corresponds to 40 kN, heavy load corresponds to 55 kN, and extremely heavy load corresponds to 70 kN. The reference load value is corrected, and the correction factors include the pavement type coefficient, the temperature influence coefficient and the humidity influence coefficient. The pavement type coefficient is determined according to the pavement material, and the asphalt concrete pavement is 1.0, the cement concrete pavement is 1.2, and the composite pavement is 1.1. The temperature influence coefficient is calculated by the environmental temperature, and the base temperature is 20 degrees Celsius, the coefficient decreases by 0.05 for every 10 degrees Celsius increase, and the coefficient increases by 0.08 for every 10 degrees Celsius decrease. The humidity influence coefficient is calculated by the relative humidity, and the base humidity is 50%, the coefficient increases by 0.03 for every 10% increase, and the coefficient decreases by 0.02 for every 10% decrease. The reference load value is multiplied by the three correction coefficients to obtain the final vibration vertical load intensity value, and the vibration vertical load intensity data table is arranged according to the time stamp and the position number, which includes four fields of position number, time stamp, load state level and load intensity value.

[0040] Step S23 includes the following steps: Step S231: Extracting the subgrade bearing capacity, density and thickness in the quality repair design requirements of the plurality of defect positions; Step S232: Coupling the vibration vertical load intensity data for nonlinear effects to generate load intensity nonlinear coupling data; Step S233: According to the load intensity nonlinear coupling data, the subgrade bearing capacity, the compactness and the thickness are simulated and analyzed to obtain the base layer shear deformation mechanics state; Step S234: The adhesion failure degree at the joint caused by the base layer shear deformation mechanics state is analyzed; Step S235: Based on the base layer shear deformation mechanics state and the adhesion failure degree at the joint, the periodic structure weakening behavior simulation is quantified to obtain the periodic structure weakening data.

[0041] In the embodiment of the present application, the processing flow of step S231 is as follows: in the quality control stage after the road defect repair is completed, the quality repair design requirement data corresponding to each repair area is extracted through the construction log system and the structural layer construction record table, which includes three key indicators, namely the roadbed bearing capacity design value, the compaction degree design control value and the base layer thickness design value. The roadbed bearing capacity design value is represented by the standard static load test data, with the unit of kPa. The compaction degree design control value is the relative density percentage, with the unit of %. The base layer thickness design value is the vertical structure thickness from the surface layer to the top of the bottom base layer, with the unit of mm. In the data extraction process, a unique number is set for each defect position, and the data is extracted by matching the number with the construction quality control data in the field level. In this embodiment, the data of 52 defect positions is extracted, wherein the roadbed bearing capacity design value is between 140 kPa and 180 kPa, the compaction degree design control value is between 93% and 98%, and the base layer thickness design value ranges from 220 mm to 280 mm. Unified checking logic is performed on the data of all defect positions, a double verification mechanism is adopted to remove and complete the repeated, missing or out-of-limit items in the data source, and all data is structured and stored as a standard three-field data table under the premise of ensuring data consistency and integrity, which is used as the input of subsequent nonlinear coupling processing and mechanical simulation. The processing flow of step S232 is as follows: the vibration vertical load intensity data obtained in step S22 is called, and the data is subjected to nonlinear effect coupling processing with the design control parameters extracted in step S231. The coupling process adopts a nonlinear fitting strategy based on piecewise regression. First, the load intensity sequence of each defect position is segmented by interval, and the division principle is that each 20s is a time window, and a total of 480 window segments are obtained. In each segment, the statistical characteristics such as the mean, peak and range of load intensity are extracted and formed into a feature vector. Then, the design parameters of the same numbered position are standardized, so that the values of bearing capacity, compaction degree and thickness are normalized to the interval [0, 1]. Subsequently, the multivariate piecewise regression method is used to perform nonlinear fitting between the load characteristics and the design parameters of each segment. The stepwise regression method is used to extract the most significant influencing factors, and the F value sorting is used to determine the significance of the variables. For the variable combination with high correlation, principal component analysis is further performed, and the first two principal components are selected as the mapping variables that comprehensively reflect the coupling effect between load disturbance and design structure. Finally, the coupling data set of each time segment is constructed, which includes the window number, the principal component values 1 and 2, the load intensity extreme value and mean value, the fitting residual, the correlation coefficient and other fields. The result is the nonlinear coupling data of load intensity, which is indexed by position number and time window in the data structure for the index access mechanism of the next shear deformation analysis.

[0042] The processing flow of step S233 is to perform a simulation analysis of the base shear deformation mechanical state of each defect position in each time period based on the load intensity nonlinear coupling data generated in step S232. The simulation method uses a two-dimensional shear deformation derivation formula in the finite difference method to calculate the displacement-stress distribution. The material constitutive relationship in the initial state is defined as a linear elastic model. The subgrade bearing capacity design value is used as the starting boundary of the vertical static load. The density is used as the correction factor of the lateral deformation modulus. The thickness is used as the interlayer index coefficient in the stress attenuation function. In the calculation process, the nonlinear coupling principal component variables are applied as disturbance sources in the loading path. The dynamic response of the disturbance deformation variable is controlled by defining the periodicity and peak range of the disturbance load term. The difference step is set to 0.05 m. The calculation area is symmetrically arranged in a 3m x 3m area. The number of units is 60 x 60. The shear strain, shear stress, and principal strain direction change in each unit are recorded. Especially for the node units near the disturbance center area, the strain energy change rate is calculated as the basis for determining whether to enter the shear instability critical state. Finally, the time sequence shear deformation mechanical state data of each defect position is generated. The data structure includes time number, shear strain value, shear stress value, equivalent strain energy density, critical stability coefficient, boundary displacement trend, and other parameter values, which are stored in real time through the edge nodes.

[0043] The processing flow of step S234 is, on the basis of the base layer shear deformation mechanics state obtained in step S233, further analyzing the spatial correspondence between the shear displacement mutation region and the known joint position. The spatial calibration of the joint position is matched by repairing the CAD coordinates in the construction drawing and the LIDAR scanning data recorded in the edge node. The distance threshold control method is used to determine the joint position point set. The point set is centered on each joint and extends 0.3m to both sides to form an analysis region. A total of 68 joint regions are calibrated. Each joint region is divided into 12 equal-width profile units. The shear strain and the principal strain direction change in each profile unit are analyzed by double indicators. The analysis logic is that when the shear strain continuously exceeds 8‰ within 30min and the principal strain direction change accumulates more than 45 degrees, it is determined that the profile region has a potential adhesive slip trend. Further, the strain energy release rate is extracted from the obvious slip trend region. The slip energy discrimination method is used for quantification. The threshold is set to 3.5kJ / m². When the value continuously crosses three time windows and maintains a growth trend, it is defined as moderate adhesive failure. When it exceeds 6.0kJ / m² and a fracture transfer phenomenon occurs, it is defined as severe adhesive failure. All failure types are represented by adhesive failure grade coding. The coding range is 0 to 3 levels. 0 represents no failure, 1 represents light micro-slip, 2 represents moderate interlayer slip, and 3 represents severe adhesive fracture. The failure degree level is mapped to each joint profile unit point by point to construct a two-dimensional spatial matrix, which is used as the initial condition input of the periodic structure weakening simulation. All analysis processes use the strain time sequence and spatial position correspondence method for automatic indexing to avoid human calibration errors.

[0044] The processing flow of step S235 is to establish a weakening behavior quantitative analysis framework of periodic structure response evolution using the shear deformation mechanics state obtained in step S233 and the joint adhesion failure degree obtained in step S234. The framework adopts a time layering promotion strategy and is divided into 12h periodic mechanical response windows according to the period. Key index sequences in each periodic window are extracted as the quantitative basis, including five types of core data such as the maximum shear stress, cumulative shear displacement, maximum value of joint failure level, spatial diffusion rate of joint failure level, and local strain energy density mutation frequency. The evolution trend of the index sequence between periods is extracted using the grey correlation analysis method, and the grey absolute correlation degree of the index change rate between any two periods is calculated. The threshold value 0.75 is selected as the strong correlation judgment standard. When there are more than 3 index change rates between two periods that are greater than the threshold value, it is determined that a significant structural weakening event occurs in the period. A periodic structure weakening index WSI (Weakening Structural Index) is further constructed to quantitatively express the periodic weakening degree. The index is a dimensionless value with a range of 0 to 1, where 0 represents no structural weakening and 1 represents a strong weakening process. In the actual calculation process, the monitoring data of 52 defect areas within 72h is periodically analyzed, resulting in 144 periodic windows. Among them, the number of periodic windows with WSI values greater than 0.8 is 27, accounting for 18.75%. These periodic windows are concentrated in the superposition of shear strain density fluctuation frequent areas and failure level mutation areas. The periodic structure weakening data is finally summarized in the form of a structure body with time index, spatial number, index sequence, and WSI value for subsequent edge rule matching to judge the structure over-limit trend and maintenance intervention scheduling logic input.

[0045] Step S234 includes the following steps: Obtain the basic properties of the repair material; perform interface adhesion strength analysis on the basic properties of the repair material to obtain the material interface adhesion strength; Perform oblique shear stress analysis according to the base shear deformation mechanics state to generate base oblique shear stress strength; Simulate the coupling of base oblique shear stress strength to generate a gradient decay state along the joint depth; Correlate and identify the interface slip mode of the material joint surface according to the gradient decay state of the material interface adhesion strength to obtain the surface layer interface slip mode; Perform joint adhesion failure analysis based on the surface layer interface slip mode to obtain the joint adhesion failure degree.

[0046] In the embodiment of the present application, the basic characteristics of the repair material are obtained, the materials involved include three types of C40 cement concrete repair material, modified asphalt emulsifier, and interfacial bonding resin, and the basic characteristic parameters mainly include five items of material elastic modulus, Poisson's ratio, tensile strength, shear strength and bonding force value. In-situ measurement is carried out on the edge nodes through the embedded multi-channel material characteristic detection unit. The specific operation is to respectively collect the mechanical response curves of the repair material at 6h, 12h, 24h, 48h and 72h after initial setting of mixing. The elastic modulus test is obtained by recording the axial strain and stress ratio through the triaxial pressure sensor. The Poisson's ratio is calculated by the ratio of radial strain to axial strain. The tensile and shear strength test uses the shear loading plate fixation method to gradually load to the limit stress at the time of fracture as the statistical value. The bonding force value is obtained by using the interface peeling unit to carry out the lifting and pulling experiment in the vertical direction of the bonding profile, and the interface fracture point load is converted into the unit bonding strength. The average value is extracted in 5 groups of repeated experiments for subsequent analysis. Then, the interface adhesion strength of the repair material basic characteristics is analyzed. The layered stress analysis method is used to theoretically model the bonding force behavior of the three materials at the interface with the original base material. In the modeling process, the minimum potential energy method is used to construct the interface force distribution, and the average value of the adhesion strength under the unit contact area is extracted. The adhesion force of the asphalt emulsifier is set to 0.85MPa, the adhesion force of the C40 repair material is set to 1.62MPa, and the adhesion force of the bonding resin is set to 2.25MPa, forming the material interface adhesion strength data set.

[0047] Under the premise of obtaining the shear deformation mechanics state of the base layer, the oblique shear stress analysis is carried out. The oblique shear is defined as the shear stress response caused by the inconsistency of the principal strain direction in the base layer and the loading direction in the non-orthogonal direction. The shear stress response is reconstructed by the displacement tensor rotation calculation method. In the operation, the joint along the line is taken as the symmetry axis direction, the shear strain main vector in each 0.2m section is extracted, the shear strain projection component is rotated to the 45° direction along the joint, and the shear stress intensity is obtained by multiplying the Young's modulus of the material. This operation is carried out within 0.6m on both sides of the joint along the line, a total of 220 stress paths are extracted, the maximum value of the oblique shear stress intensity is 3.42MPa, and the minimum value is 0.96MPa. Further, the attenuation state of the oblique shear stress intensity along the depth direction of the joint is simulated. The exponential order fitting method is used to construct the stress depth attenuation function, the initial value of the attenuation coefficient is set to 0.015m⁻¹, and the average error is adjusted to be less than 0.07MPa through the fitting residual convergence method. The attenuation trend of the oblique shear stress in the depth direction is determined to form a gradient distribution field. The distribution field is mapped to the longitudinal profile of the joint to form a three-dimensional tensor matrix, which is used for interface slip mode analysis.

[0048] Subsequently, the interface slip mode of the joint surface is correlated and identified, the oblique shear stress intensity of each depth point is operated by the slip threshold ratio method, and the interface slip area is determined when the ratio is greater than 1.0, the transition zone is 0.8 to 1.0, and the stable adhesion area is less than 0.8. Each joint is divided into a depth unit every 10 mm, a total of 6800 points in the profile area are analyzed, the slip type is marked, and the slip mode is divided according to the slip area form. The specific mode includes three types of through type, layered type and point type. The through type is that the slip area is continuous along the depth and exceeds 80%, the layered type is that there is an intermediate slip zone in the local area, and the point type is a scattered and discontinuous slip point. Each slip mode forms a one-to-one data record under each joint number.

[0049] Finally, the adhesive failure analysis of the joint is carried out based on the above surface interface slip mode. The slip mode is coupled and analyzed with the actual shear stress fluctuation data by using the segmented failure accumulation method. The slip state is divided according to the profile layer in each joint area, the occurrence frequency and area change rate of the slip point in each shear stress peak value period are judged, which is attributed to the energy release type, interface deterioration type and stress accumulation type in the adhesive failure mechanism, and the failure level is quantified according to the slip area evolution trend. The level is divided into 0 to 3 levels. The slip area of the through type slip mode accounts for more than 90% during the peak stress period, which is defined as severe failure. The point type area is less than 10%, which is defined as mild failure. Finally, a complete spatial distribution data of the adhesive failure degree of the joint is constructed to serve as the input basic data structure for subsequent periodic structure weakening analysis.

[0050] Step S24 includes the following steps: Step S241: Multi-scale base crack evolution structure simulation is performed on the periodic structure weakening data according to the vibration vertical load intensity data, and a base crack evolution structure is obtained; Step S242: The base crack evolution structure is subjected to water intrusion maximum value equivalent mapping, and the water intrusion maximum value corresponding to the base crack evolution structure is obtained; Step S243: Base expansion / softening coupling effect analysis is performed according to the water intrusion maximum value and the base crack evolution structure, and base expansion / softening coupling effect data is obtained; Step S244: Water damage cumulative gradient estimation is performed according to the water intrusion maximum value, the base crack evolution structure and the base expansion / softening coupling effect data, and a road surface water damage cumulative gradient is obtained.

[0051] In the embodiment of the application, the multi-scale base layer crack evolution structure simulation is performed on the periodic structure weakening data according to the vibration vertical load intensity data. In the simulation process, the hierarchical crack propagation analysis method is used. Firstly, a 1m*1m*0.5m three-dimensional profile unit block is constructed with the identified region in the periodic structure weakening data as the center. The vibration vertical load intensity data is embedded in the element grid unit divided in the unit block with a step of 20mm. The data is derived from the vertical dynamic load obtained by converting the pavement surface vibration acceleration. The amplitude range is 1.1kN to 5.8kN, and the frequency range is 6Hz to 23Hz. The fatigue strength threshold is set to 75% of the tensile strength of the repair material in each grid element, and the vibration energy accumulation rate is used as the time advancing function to calculate the crack initiation time of each grid. Then, according to the strain concentration zone between the crack initiation points, the crack evolution paths are formed by connecting them according to the direction consistency. All paths extend to the adjacent unit and merge to form the crack network topology structure. In the microscale, the crack tip area is refined with a 5mm grid. The number of microcrack branches, branch length, expansion angle and intersection frequency are recorded. The crack resistance coefficient and crack reconstruction geometry are adjusted in combination with the micro-porosity distribution in the material. The simulation and construction of the base layer crack evolution structure are completed.

[0052] The maximum water intrusion equivalent mapping operation is performed on the constructed base layer crack evolution structure. The crack channel length, depth and opening are extracted in the crack profile with a unit of 10mm. The crack opening is calculated from the displacement difference sensing data, and the range is 0.2mm to 3.4mm. Based on the crack channel geometric parameters, the capillary guided water migration method is used to analyze the water permeation path. The initial soil humidity is set to 0.21m³ / m³, and the crack bottom capillary suction boundary is 38kPa. The 10h permeation test simulation is performed under the condition of standard rainfall intensity 35mm / h. The maximum water intrusion depth and water volume concentration peak in each crack channel are recorded. The crack node water intrusion equivalent mapping field is generated. The water superposition effect of crack intersection points and intersection sections is compared and analyzed in combination with the original crack topology data. The water intrusion maximum value spatial distribution result in the crack structure is extracted.

[0053] According to the obtained maximum water intrusion and base layer crack evolution structure, the base layer swelling and softening coupling effect is analyzed, 30mm*30mm coupling response unit cells are extracted in the internal distribution area of the crack path, the crack number, water intrusion value and material moisture absorption expansion coefficient in the area are read, the expansion coefficient of the asphalt stabilized macadam base is set as 0.014mm / mm·% moisture content, the volume strain is calculated by using the expansion response conversion formula and is mapped to the grid area, on this basis, the dynamic elastic modulus attenuation coefficient 0.28 is introduced to describe the softening behavior, and then the regional level swelling-softening coupling tensor field is drawn based on each coupling response unit cell, the volume expansion ratio and elastic modulus attenuation percentage in the unit cell are recorded, the tensor field is used to reflect the composite deformation behavior caused by the combined action of water and cracks, and the coupling data is accumulated step by step in the edge processing unit as input for the subsequent gradient analysis module. Finally, the water damage cumulative gradient is estimated based on the maximum water intrusion, the base layer crack evolution structure and the base layer swelling and softening coupling effect data. The operation adopts a three-dimensional layered gradient weight superposition method, divides the base layer thickness direction into 20 layers, each layer is 10mm thick, and three types of parameters of water intrusion concentration, expansion strain ratio and elastic modulus attenuation rate are extracted in each layer. The weight factors are set as 0.45, 0.35 and 0.20 respectively. The basis for setting the weight factors as 0.45 (water intrusion concentration), 0.35 (expansion strain ratio) and 0.20 (elastic modulus attenuation rate) is to comprehensively balance the relative influence degree of each parameter on different physical effects in the base layer water damage evolution mechanism. Specifically, the water intrusion concentration directly determines the effective penetration depth and range of water, and is the dominant factor causing the expansion and softening effect, so it is given the highest weight (0.45); the expansion strain ratio reflects the deformation degree of the material due to moisture absorption and expansion, which has a significant influence on the structural integrity, so it is given a medium weight (0.35); and the elastic modulus attenuation rate is an indirect reflection of the material performance degradation, which has a lagging influence and cumulative characteristics, so the weight is relatively low (0.20). The cumulative influence value is calculated by assigning weights to each layer, the cumulative value is superimposed on the equal value surface in the structural weakening path area, and the gradient direction change rate is respectively calculated according to the coordinate axis direction. Combined with the continuity of the crack space path and the water retention area, the water damage gradient is calculated layer by layer downward every 20mm step, the water damage cumulative gradient distribution data is derived, and is stored in the monitoring task node structure for subsequent evaluation and alarm logic module analysis.

[0054] Step S3 includes the following steps: Step S31: The periodic structure weakening data and the pavement water damage cumulative gradient are normalized respectively to obtain periodic structure weakening normalized data and pavement water damage cumulative normalized gradient respectively. Step S32: Perform repair state service life assessment on the periodic structure weakening normalized data and the pavement water damage accumulation normalized gradient based on the reinforcement learning algorithm possessed in the edge computing platform, to obtain repair state service life assessment data; Step S33: Send the repair state service life assessment data to the terminal.

[0055] As an example of the present application, reference is made to Figure 3 As shown in the example, the step S3 includes: Step S31: Perform normalization processing on the periodic structure weakening data and the pavement water damage accumulation gradient respectively, to obtain periodic structure weakening normalized data and pavement water damage accumulation normalized gradient respectively; In the embodiment of the present application, the periodic structure weakening data and the pavement water damage accumulation gradient are normalized respectively. The periodic structure weakening data is derived from four indexes in the base crack evolution structure, including crack distribution density, crack length distribution, crack expansion direction consistency factor and crack network connectivity. The crack distribution density is in units of strips / m², the length distribution ranges from 20mm to 410mm, the direction consistency factor is calculated as the ratio of the standard deviation of crack angle deviation to the maximum angle deviation, and the crack network connectivity is normalized by the number of maximum crack paths. In the normalization process, the original data is mapped to the interval [0, 1] using the maximum-minimum normalization method. In the operation, the maximum and minimum values of the observation area samples are selected for standard linear normalization transformation. All parameters are fused with the same weight as the periodic structure weakening normalized data. The pavement water damage accumulation gradient is derived from the maximum water intrusion, water retention layer depth, modulus attenuation degree caused by water and expansion strain gradient distribution. The water retention layer depth is in units of mm, the modulus attenuation degree is represented by the ratio of the initial value to the current value of the material elastic modulus, and the expansion strain is in units of %. Each parameter is linearly normalized using the maximum-minimum normalization method. After normalization, the first principal component is extracted as the pavement water damage accumulation normalized gradient using the principal component analysis method.

[0056] Step S32: Perform repair state service life assessment on the periodic structure weakening normalized data and the pavement water damage accumulation normalized gradient based on the reinforcement learning algorithm possessed in the edge computing platform, to obtain repair state service life assessment data; In the embodiment of the present application, the periodic structure weakening normalized data and the pavement water damage cumulative normalized gradient are used for repair state service life evaluation based on the reinforcement learning algorithm possessed by the edge computing platform. The deterioration trend of the road repair state is optimally estimated by the discrete state space reinforcement training method in the Q learning algorithm. Specifically, the periodic structure weakening normalized data and the pavement water damage cumulative normalized gradient are spliced as a state vector input. The state vector is divided into 20x20, i.e. 400 state nodes at an interval of 0.05. Each state node corresponds to a repair state grade. The behavior selection includes continuing service and arranging secondary repair. The reward value is calculated as the inverse of the state deterioration speed per unit time. The learning rate is set to 0.2, the discount factor is set to 0.9, the initial Q table is all zero, the exploration is performed by the epsilon-greedy strategy, the epsilon value is set to 0.1, the iteration round is set to 10000 times, the service time of the simulated road is set to 0 to 180 days in each round, and the historical sequence of the periodic structure weakening and water damage gradient deterioration actually collected in the observed road is introduced as a training sample during the period. In each iteration, the maximum Q value action is selected from the Q table according to the current state and behavior, and the strategy table is updated. Finally, the expected residual life corresponding to each state node in the Q table is formed. The life expectation is extracted as the repair state service life evaluation data according to the corresponding position of the state vector in all training samples.

[0057] Step S33: Send the repair state service life evaluation data to the terminal.

[0058] In the embodiment of the present application, the obtained repair state service life evaluation data is sent to the terminal. The sending operation packs the evaluation data in the form of a binary array through the CAN bus module of the edge computing node. The structure includes four parts: data identifier, sampling timestamp, corresponding state vector index position and life evaluation value. The data identifier is 8 bits, the timestamp is 64-bit UNIX time format, the index position is encoded as two 8-bit unsigned integers representing the two-dimensional coordinates of the state vector, and the life evaluation value is represented in the form of a 32-bit floating point number. The overall length of the data packet is 20 bytes, the sending period is 60s, the sending path is forwarded to the roadside information interaction terminal through the local link relay station and cached in the terminal database, the terminal database stores and backtracks the life data in a structured index mode based on time series, each data is bound to the corresponding geographical positioning identifier and the unique identification code of the processing node, and the integrity and reliability of the data link are ensured.

[0059] The foregoing is considered as illustrative only of the principles of the application. Numerous modifications and changes will readily occur to those skilled in the art, and it is intended to embrace all such modifications and changes that fall within the scope of the application. Accordingly, the application is not to be restricted in scope to the specific embodiments disclosed herein but is to be accorded the full scope that the principles and novel features request appropriately granted.

Claims

1. A road repair status monitoring method based on edge computing, characterized in that: The following steps are involved: Step S1: Obtaining repair planning information of repaired roads; Extract multiple repaired road defect locations from planning information; 3D laser scanning vibrometers were deployed at multiple repaired road defect locations to collect road vibration signals during multiple traffic periods. The disordered distribution time domain characteristics were then analyzed to obtain a disordered distribution time domain map of the vibration signals. Step S2: simulating and quantifying the structural weakening behavior during the service cycle based on the disordered distribution time domain diagram of the vibration signal to obtain periodic structural weakening data; estimating the water damage cumulative gradient based on the periodic structural weakening data to obtain the pavement water damage cumulative gradient; Step S3: Based on the reinforcement learning algorithm in the edge computing platform, the periodic structural weakening data and the cumulative gradient of pavement water damage are evaluated for the repair status service life to obtain the repair status service life evaluation data; and the repair status service life evaluation data is sent to the terminal.

2. The road repair status monitoring method based on edge computing according to claim 1 is characterized in that: Step S1 includes the following steps: Step S11: Acquire repair planning information of repaired roads; extract multiple locations of repaired road defects in the planning information; Step S12: deploying 3D laser scanning vibrometers at multiple repaired road defect locations to collect road vibration signals during multiple traffic periods; Step S13: performing average filtering on the road vibration signal to obtain a road vibration filtered signal; Step S14: performing disordered distribution time domain feature analysis on the road surface vibration filter signal, thereby obtaining a disordered distribution time domain graph of the vibration signal.

3. The road repair status monitoring method based on edge computing according to claim 1 is characterized in that: Step S2 includes the following steps: Step S21: extracting quality design requirements for multiple repaired road defect locations based on the repair planning information of the repaired roads to obtain quality repair design requirements for the multiple defect locations; Step S22: performing vertical load intensity conversion on the disordered distribution time domain diagram of the vibration signal to obtain vibration vertical load intensity data; Step S23: simulating and quantifying the structural weakening behavior during the service cycle of the quality repair design requirements of multiple defect locations based on the vibration vertical load intensity data to obtain periodic structural weakening data; Step S24: Estimating the water damage cumulative gradient based on the periodic structural weakening data and the vibration vertical load intensity data, thereby obtaining the pavement water damage cumulative gradient.

4. The road repair status monitoring method based on edge computing according to claim 3 is characterized in that: Step S22 includes the following steps: Step S221: performing amplitude change rate analysis on the disordered distribution time domain graph of the vibration signal to obtain the disordered amplitude change rate; Step S222: performing an approximate linear correlation analysis of the rising slope increment from low to high on the disordered distribution time domain diagram of the vibration signal according to the disordered amplitude change rate, and obtaining slope increment approximate correlation data; Step S223: performing multi-scale spectral entropy decomposition processing on the disordered distribution time domain graph of the vibration signal based on the slope increment approximate correlation data to obtain time series energy spectral entropy decomposition data; Step S224: performing equal vertical load mechanical state mapping on the time series energy spectrum entropy decomposition data to obtain the vertical load mechanical state; Step S225: Perform vertical load intensity conversion according to the vertical load mechanical state to obtain vibration vertical load intensity data.

5. The road repair status monitoring method based on edge computing according to claim 3 is characterized in that: Step S23 includes the following steps: Step S231: extracting the roadbed bearing capacity, density and thickness in the quality repair design requirements of multiple defect locations; Step S232: performing nonlinear effect coupling on the vibration vertical load intensity data to generate load intensity nonlinear coupling data; Step S233: performing a simulation analysis of the shear deformation mechanical state of the base layer on the roadbed bearing capacity, density and thickness according to the load intensity nonlinear coupling data to obtain the shear deformation mechanical state of the base layer; Step S234: analyzing the degree of adhesive failure at the joint caused by the shear deformation mechanical state of the base layer; Step S235: Based on the shear deformation mechanical state of the base layer and the degree of adhesive failure at the joint, the periodic structural weakening behavior is simulated and quantified to obtain periodic structural weakening data.

6. The road repair status monitoring method based on edge computing according to claim 5 is characterized in that: Step S234 includes the following steps: Obtaining basic properties of the repair material; performing interface adhesion strength analysis on the basic properties of the repair material to obtain the material interface adhesion strength; Perform oblique shear stress analysis based on the shear deformation mechanical state of the base layer to generate the oblique shear stress intensity of the base layer; The intensity of the oblique shear stress of the coupled base layer is simulated to show a gradient attenuation along the depth of the joint; Correlation identification of the interface slip mode on the surface of the material joint is performed on the material interface adhesion strength according to the gradient attenuation state to obtain the surface interface slip mode; The adhesion failure analysis at the joint is performed based on the surface interface slip mode to obtain the degree of adhesion failure at the joint.

7. The road repair status monitoring method based on edge computing according to claim 6 is characterized in that: Step S24 includes the following steps: Step S241: performing a multi-scale base crack evolution structure simulation on the periodic structural weakening data according to the vibration vertical load intensity data to obtain the base crack evolution structure; Step S242: performing equal mapping of the maximum water intrusion value on the base crack evolution structure to obtain the maximum water intrusion value corresponding to the base crack evolution structure; Step S243: performing a base expansion / softening coupling effect analysis based on the maximum water intrusion value and the base crack evolution structure to obtain base expansion / softening coupling effect data; Step S244: Estimating the water damage cumulative gradient based on the maximum water intrusion value, the base crack evolution structure, and the base expansion / softening coupling effect data, thereby obtaining the pavement water damage cumulative gradient.

8. The road repair status monitoring method based on edge computing according to claim 1 is characterized in that: Step S3 includes the following steps: Step S31: normalizing the periodic structure weakening data and the pavement water damage cumulative gradient respectively to obtain the periodic structure weakening normalized data and the pavement water damage cumulative normalized gradient respectively; Step S32: Based on the reinforcement learning algorithm in the edge computing platform, the normalized data of periodic structural weakening and the normalized gradient of pavement water damage accumulation are evaluated for the service life of the repair state to obtain the evaluation data of the service life of the repair state; Step S33: Send the repair status service life assessment data to the terminal.

9. A road repair status monitoring system based on edge computing, characterized in that: The method for monitoring a road repair state based on edge computing according to any one of claims 1 to 8 is configured to be executed. The system for monitoring a road repair state based on edge computing comprises: The vibration signal analysis module is used to obtain repair planning information for repaired roads; extract multiple locations of repaired road defects from the planning information; deploy 3D laser scanning vibrometers at multiple repaired road defect locations to collect road vibration signals during multiple traffic periods; and then perform disordered distribution time domain feature analysis to obtain a disordered distribution time domain map of the vibration signal; The damage analysis module is used to simulate and quantify the structural weakening behavior during the service cycle based on the disordered distribution time domain diagram of the vibration signal to obtain periodic structural weakening data; the water damage cumulative gradient is estimated based on the periodic structural weakening data to obtain the pavement water damage cumulative gradient; The condition service life assessment module is used to perform repair condition service life assessment on periodic structural weakening data and pavement water damage cumulative gradient based on the reinforcement learning algorithm in the edge computing platform to obtain repair condition service life assessment data; and send the repair condition service life assessment data to the terminal.

Images