A pavement quality detection method and system for engineering supervision

Abstract

This invention relates to the field of engineering management technology, specifically to a method and system for road surface quality inspection in engineering supervision. The method includes the following steps: acquiring the load-displacement response sequence of road surface measuring points; extracting items with displacement differences greater than a threshold to construct abrupt change features; integrating the deviation between the actual rebound and the theoretical recovery displacement to obtain attenuation features; weighting the attenuation features to obtain a joint attenuation value; generating hazard identifiers based on three-dimensional coordinates; and classifying defects based on variance clustering and binding them to supervision station numbers to generate control strategies. In this invention, by extracting abrupt change features to quantify deformation gradients, combining displacement deviation integrals to construct attenuation features and mapping spatial identifiers, a degradation model is driven to predict evolution trends. This breaks the static sampling mode to achieve continuous, non-destructive, dynamic tracking of structural mechanical attenuation, accurately locating the spatial distribution of internal void defects, providing early warning of evolution trends, and generating targeted handling strategies, significantly improving the timeliness of supervision quality control.

Description

Technical Field

[0001] This invention relates to the field of engineering management technology, and in particular to a method and system for road surface quality testing in engineering supervision. Background Technology

[0002] The field of engineering management technology covers core aspects of engineering projects, including organization and coordination, quality control, schedule management, and cost control, from project initiation, design, construction to acceptance. It involves the formulation of construction process specifications, construction process monitoring, material performance testing, structural safety assessment, and on-site data collection and analysis. Especially in road engineering, pavement quality testing is an important part of engineering quality control. It systematically tests and evaluates indicators such as pavement smoothness, compaction, thickness, and strength through a combination of on-site testing equipment and manual labor to ensure that the project meets the requirements of design standards and construction specifications.

[0003] One traditional method for road surface quality inspection in engineering supervision involves using manual inspection combined with specialized testing equipment to assess road construction quality during the engineering supervision process. This method involves placing a three-meter straightedge at intervals on the road surface and measuring the gaps to determine the smoothness; using a core drill to extract core samples at designated locations on the road surface and measuring their thickness with calipers; using the sand filling method to fill pre-excavated pits with standard sand and weighing the sand to calculate the compaction degree; and using a falling weight deflectometer to apply impact loads to the road surface and record the deflection values ​​to evaluate the structural bearing capacity. All of the above tests are carried out point by point according to the established test point layout rules, and the test records are manually compiled and compared for analysis.

[0004] Current technologies for road quality inspection mainly rely on manual inspections combined with traditional equipment for discrete point measurements. They use rulers to measure gaps and core drilling to obtain parameters. Local sampling damages structural integrity, and the sparse measurement points make it difficult to reflect hidden defects in the entire road section. The static evaluation model that relies on manual data reading and comparison with specifications is greatly affected by subjective experience. It also lacks dynamic monitoring of the deep mechanical response characteristics of road materials, making it easy to miss internal voids during the engineering supervision process. This results in a lack of prediction of the evolution trend of hidden dangers and a lag in quality control, leading to increased maintenance costs in the later stages. Summary of the Invention

[0005] To address the shortcomings of existing technologies in road quality inspection, which primarily rely on manual inspections combined with traditional equipment for discrete point measurements, using rulers to measure gaps and core drilling to obtain parameters, resulting in localized sampling that damages structural integrity and sparse measurement points that fail to reflect hidden defects across the entire road section, and the static evaluation model based on manual data reading and comparison with specifications that is highly susceptible to subjective experience interference, and insufficient dynamic monitoring of the deep mechanical response characteristics of road materials, this invention provides a road quality inspection method for engineering supervision.

[0006] To achieve the above objectives, the present invention employs a road surface quality inspection method for engineering supervision, comprising the following steps:

[0007] S1: Obtain the impact load sequence and pavement deformation sequence for the road surface measuring points of the engineering supervision, and perform differentiation to extract the load peak position and maximum displacement position, and construct the load displacement response sequence;

[0008] S2: Calculate the displacement change difference between adjacent test cycles in the load displacement response sequence, extract the terms whose displacement change difference is greater than the preset deformation gradient threshold, and construct local deformation abrupt change features;

[0009] S3: Obtain the local deformation abrupt change characteristics, collect the actual rebound displacement and rebound recovery time of the corresponding period, calculate the displacement deviation between the actual rebound displacement and the preset theoretical recovery displacement, and perform integral calculation on the displacement deviation along the rebound recovery time to construct the rebound attenuation deviation characteristics.

[0010] S4: Based on the local deformation abruptness features, perform weighted coupling on the rebound attenuation deviation features to obtain the joint attenuation value, filter the local maximum value items to extract the corresponding three-dimensional coordinate coefficients and associate them with the joint attenuation value to generate a pavement structure hidden danger identifier;

[0011] S5: Based on the discrete variance numerical clustering of the pavement structure hazard identification, the hazards are divided into interlayer loose category and structural void category. After binding the on-site supervision station number, the evolution trend of the hazards is predicted based on the preset degradation mechanism model, and an engineering monitoring and management strategy is generated.

[0012] As a further aspect of the present invention, the load displacement response sequence includes the load peak time, maximum displacement amplitude, and phase difference; the local deformation abrupt change characteristics include the abrupt change amplitude, abrupt change location, and abrupt change frequency; the rebound attenuation deviation characteristics include the displacement deviation integral area, attenuation rate, and recovery hysteresis, where the recovery hysteresis specifically refers to the time difference between the actual rebound recovery time and the theoretical recovery time, reflecting the degree of hysteresis in the elastic recovery capability of the pavement material; the pavement structure hazard identification includes extreme three-dimensional coordinates, joint attenuation value, and spatial distribution density, where the joint attenuation value specifically refers to the comprehensive attenuation quantification value obtained by weighted coupling of the rebound attenuation deviation characteristics with the local deformation abrupt change characteristics; and the engineering monitoring and management strategy includes hazard category attribution, evolution trend prediction results, and corresponding disposal measures recommendations.

[0013] As a further aspect of the present invention, the specific steps of S1 are as follows:

[0014] S101: Collect the impact load sequence and road deformation sequence of the road measurement points in the sample dataset, compare the timestamps of the two sequences frame by frame, remove outlier measurement point samples whose time difference is greater than the deviation benchmark value, map the measurement points with overlapping timestamps to key-value pairs according to the measurement point number and the corresponding timestamp, and generate a time-aligned state array.

[0015] S102: Call the time-aligned state array, extract the node derivative values ​​of the impact load sequence by taking the derivative and compare them with the zero reference value. Extract the nodes whose derivatives cross zero and whose second derivatives are negative as candidate load extreme points. Extract the zero reference crossing points of the road deformation sequence by differential extraction as candidate displacement extreme points and generate an extreme value location index set.

[0016] S103: Based on the extreme value location index set, call the candidate load extreme value points and candidate displacement extreme value points, retrieve the node values ​​of the impact load sequence and the road deformation sequence at the corresponding extreme value point positions, perform key-value pair mapping and pairing of the load node values ​​and displacement node values ​​in time order, and generate the load displacement response sequence.

[0017] As a further aspect of the present invention, the specific steps of S2 are as follows:

[0018] S201: Call the load displacement response sequence to extract the timestamp and displacement node values. Based on the timestamp, divide the displacement node values ​​into multiple test periods along the time axis. Perform subtraction on the corresponding displacement node values ​​in the data of two adjacent test periods to calculate the difference in change. Arrange the displacement difference set in chronological order.

[0019] S202: Extract the change difference within the displacement difference set item by item and compare it with the preset deformation gradient threshold. Filter the extraction items whose difference is greater than the preset deformation gradient threshold. Extract the associated timestamp and displacement value. Perform aggregation and encapsulation on the timestamp, change difference, and displacement value to generate a mutation data vector.

[0020] S203: Based on the timestamps in the mutation data vector, match and locate them in the load displacement response sequence, extract the corresponding load values, establish a ternary key value mapping of the load values, change differences and displacement values ​​according to the timestamp index, and perform feature splicing and one-dimensional sequence flattening transformation to generate local deformation mutation features.

[0021] As a further aspect of the present invention, the preset deformation gradient threshold is determined based on the statistical distribution of all variation differences within the displacement difference set. The statistical distribution includes the maximum value, minimum value, and average value of the variation differences. The variation differences located in the preset quantile interval are selected as the preset deformation gradient threshold after sorting the variation differences according to their numerical size.

[0022] As a further aspect of the present invention, the specific steps of S3 are as follows:

[0023] S301: Obtain the local deformation abruptness characteristics, extract the periodic index to read historical test data, retrieve the actual rebound displacement value and rebound recovery time sequence, reduce the noise of the actual rebound displacement value and align it along the rebound recovery time sequence, and establish rebound time response parameters.

[0024] S302: Call the rebound time response parameters, extract the actual rebound displacement value and time node, input the time node into the preset attenuation benchmark model to calculate the preset theoretical recovery displacement, subtract the preset theoretical recovery displacement from the actual rebound displacement value to obtain discrete displacement deviation and rearrange it to generate a dynamic displacement deviation sequence.

[0025] S303: Extract discrete displacement deviation from the dynamic displacement deviation sequence, obtain the node spacing as the integration step size, multiply the discrete displacement deviation by the integration step size to obtain the micro-element area, accumulate the micro-element surface along the rebound recovery time sequence to obtain the cumulative deviation amount, and reshape the cumulative deviation amount to obtain the rebound attenuation deviation characteristics.

[0026] As a further aspect of the present invention, the preset attenuation benchmark model obtains the preset theoretical recovery displacement by multiplying the initial displacement by a natural exponential attenuation term constructed based on time nodes and attenuation coefficients.

[0027] As a further aspect of the present invention, the specific steps of S4 are as follows:

[0028] S401: Obtain the rebound attenuation deviation feature, extract the amplitude component of the local deformation abrupt feature, map the amplitude component to a preset interval through normalization to obtain a weight matrix, multiply the rebound attenuation deviation feature with the weight matrix to obtain a weighted tensor, perform mean filtering smoothing on the weighted tensor, and generate a joint attenuation value set.

[0029] S402: Call the joint attenuation value set configuration sliding window and move to read the internal values, compare the center value with the neighboring values, mark the node where the center value is greater than the neighboring value as a local maximum value, parse the grid index of the local maximum value to map the longitude, latitude and elevation coordinates, and establish a three-dimensional coordinate coefficient set.

[0030] S403: Call the three-dimensional coordinate coefficient set, retrieve the joint attenuation value corresponding to the local maximum value, pair it with the coordinate data to establish a spatial mapping key-value pair, construct feature patches based on the adjacency relationship of the spatial mapping key-value pair, compress the dimension of the feature patches and perform vector encapsulation to obtain the road structure hazard identification.

[0031] As a further aspect of the present invention, the specific steps of S5 are as follows:

[0032] S501: Obtain the attenuation distribution vector of the pavement structure hazard identification, evaluate the discrete variance value based on the degree of deviation of the vector value, compare it with the preset classification benchmark threshold, and if it is less than the preset classification benchmark threshold, it is classified into the interlayer loose category; otherwise, it is classified into the structural void category, and a hazard attribute division set is generated.

[0033] S502: Call the aforementioned hazard attribute division set, parse the spatial geographic coordinates of the elements, read the site supervision station number table to extract the coordinates of the foundation calibration point, evaluate the distance parameter based on the positional difference between the spatial geographic coordinates and the coordinates of the foundation calibration point, bind the hazard information with the calibration point corresponding to the minimum value in the distance parameter, and establish a station number association file.

[0034] S503: Extract the initial state parameters of the category from the chainage-related files, obtain the decay coefficient matrix through a preset degradation mechanism model, modulate the initial state parameters of the category based on the decay coefficient matrix to obtain the evolution rate vector, define the intervention time nodes according to the numerical level of the evolution rate vector, and logically integrate with the chainage-related files to generate an engineering monitoring and management strategy.

[0035] A road surface quality inspection system for engineering supervision, comprising:

[0036] The load deformation monitoring module acquires the impact load sequence and pavement deformation sequence for the road surface measuring points of the engineering supervision, and performs differentiation operation to extract the load peak position and maximum displacement position, and constructs the load displacement response sequence.

[0037] The deformation mutation extraction module calculates the displacement change difference between adjacent test cycles in the load displacement response sequence, extracts the terms whose displacement change difference is greater than a preset deformation gradient threshold, and constructs local deformation mutation features.

[0038] The attenuation deviation calculation module obtains the local deformation abrupt change characteristics, collects the actual rebound displacement and rebound recovery time of the corresponding period, calculates the displacement deviation between the actual rebound displacement and the preset theoretical recovery displacement, and performs integral calculation of the displacement deviation along the rebound recovery time to construct the rebound attenuation deviation characteristics.

[0039] The hazard identification generation module performs weighted coupling on the rebound attenuation deviation feature based on the local deformation abruptness feature to obtain the joint attenuation value, filters the local maximum value item to extract the corresponding three-dimensional coordinate coefficient and associates it with the joint attenuation value to generate a pavement structure hazard identification;

[0040] The control strategy generation module divides the pavement structure hazard identification into interlayer loose category and structural void category through discrete variance numerical clustering. After binding the on-site supervision station number, it predicts the evolution trend of hazard based on the preset degradation mechanism model and generates the engineering monitoring and management strategy.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0042] In this invention, impact load and displacement deformation response values ​​of road surface measuring points are obtained to construct time-series response features. Local deformation gradients are quantified by extracting characteristic abrupt changes within adjacent periods. Attenuation deviation features are constructed by combining theoretically recovered displacement and actual rebound parameter integrals. Attenuation deviations are weighted based on deformation abrupt changes and mapped to three-dimensional space to generate structural hazard identifiers. Evolution trends are predicted by variance clustering and binding to the supervision station number to drive the degradation model. This breaks the static discrete sampling mode to achieve non-destructive dynamic tracking of deep mechanical attenuation of the road surface, accurately locates the distribution area of ​​internal interlayer loosening and void defects, provides early warning of the evolution of structural hazards, and generates targeted control and disposal strategies, significantly improving the timeliness of quality control in the engineering supervision process. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of the steps of the present invention;

[0045] Figure 2 This is a detailed schematic diagram of S1 of the present invention;

[0046] Figure 3 This is a detailed schematic diagram of S2 of the present invention;

[0047] Figure 4 This is a detailed schematic diagram of S3 of the present invention;

[0048] Figure 5 This is a detailed schematic diagram of S4 of the present invention;

[0049] Figure 6 This is a detailed schematic diagram of S5 of the present invention;

[0050] Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0051] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0052] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0053] Please see Figure 1 This invention provides a method for road surface quality testing for engineering supervision, comprising the following steps:

[0054] S1: Obtain the impact load sequence and pavement deformation sequence for the road surface measuring points of the engineering supervision, and perform differentiation to extract the load peak position and maximum displacement position, and construct the load displacement response sequence;

[0055] S2: Calculate the displacement change difference between adjacent test periods in the load displacement response sequence, extract the terms whose displacement change difference is greater than the preset deformation gradient threshold, and construct local deformation abrupt change features.

[0056] S3: Obtain local deformation abrupt change characteristics, collect the actual rebound displacement and rebound recovery time of the corresponding period, calculate the displacement deviation between the actual rebound displacement and the preset theoretical recovery displacement, and perform integral calculation of the displacement deviation along the rebound recovery time to construct the rebound attenuation deviation characteristics;

[0057] S4: Based on the local deformation abruptness characteristics, the rebound attenuation deviation characteristics are weighted and coupled to obtain the joint attenuation value. Local maximum values ​​are selected to extract the corresponding three-dimensional coordinate coefficients and associate them with the joint attenuation value to generate pavement structure hidden danger indicators.

[0058] S5: Based on the discrete variance numerical clustering of pavement structure hazard identification, the hazards are divided into interlayer loose category and structural void category. After binding the on-site supervision station number, the evolution trend of hazards is predicted based on the preset degradation mechanism model, and an engineering monitoring and management strategy is generated.

[0059] The characteristics of local deformation mutation include mutation amplitude, mutation location, and mutation frequency. The characteristics of rebound attenuation deviation include displacement deviation integral area, attenuation rate, and recovery hysteresis. The recovery hysteresis specifically refers to the time difference between the actual rebound recovery time and the theoretical recovery time, reflecting the degree of lag in the elastic recovery ability of the pavement material. The pavement structure hazard identification includes extreme value three-dimensional coordinates, joint attenuation value, and spatial distribution density. The joint attenuation value specifically refers to the comprehensive attenuation quantification value obtained by weighted coupling of the local deformation mutation characteristics and the rebound attenuation deviation characteristics. The engineering monitoring and management strategy includes hazard category classification, evolution trend prediction results, and corresponding disposal measures recommendations.

[0060] Please see Figure 2 The specific steps of S1 are as follows:

[0061] S101: Collect the impact load sequence and road deformation sequence of the road measurement points in the sample dataset, compare the timestamps of the two sequences frame by frame, remove outlier measurement point samples whose time difference is greater than the deviation benchmark value, map the measurement points with overlapping timestamps to key-value pairs according to the measurement point number and the corresponding timestamp, and generate a time-aligned state array.

[0062] Dynamic response data of the pavement structure are continuously collected at a fixed sampling frequency of 2000 Hz using the dynamic mechanical sensor of a falling weight deflectometer and a high-frequency displacement gauge, directly obtaining the impact load sequence and pavement deformation sequence of the pavement measuring points. The first timestamp of each data frame is read from the impact load sequence, and the second timestamp of the same frame number is simultaneously read from the pavement deformation sequence. The difference between the first and second timestamps is calculated to the microsecond level to obtain the time difference parameter. Multiple field tests have confirmed that when the time deviation caused by data transmission delay or sensor jitter exceeds a certain limit, the physical causal relationship between load and deformation will be misaligned. Statistical regression was performed on a large amount of time-series data under normal and abnormal operating conditions, and the deviation benchmark value was set at 1.5 milliseconds. This value covers 99.5% of the normal signal delay range. The time difference parameter is compared item by item with the deviation benchmark value. If the time difference parameter value is greater than 1.5 milliseconds, it is determined that the data frame of that measuring point has an irreparable synchronization fault, and the outlier measuring point sample is completely removed from the original sample dataset. For example, for road surface measuring point number 15, its impact load sequence first timestamp is 45000.5 milliseconds, and its road surface deformation sequence second timestamp is 45002.8 milliseconds. The difference between these two timestamps yields a time difference parameter of 2.3 milliseconds, which is greater than the set deviation benchmark of 1.5 milliseconds. Therefore, a rejection operation is performed. For road surface measuring point number 16, the first timestamp is 45010.2 milliseconds, and the second timestamp is 45011.0 milliseconds, with a time difference parameter of 0.8 milliseconds, satisfying the comparison condition. A hash key-value pair mapping relationship is established between the measuring point numbers that meet the condition and their corresponding benchmark timestamp data. All verified key-value pair data are then multidimensionally reorganized and arranged according to the spatially deployed measuring point number order and the time increment order, ultimately generating a standardized time-aligned state array in memory.

[0063] S102: Call the time-aligned state array, extract the node derivative values ​​from the impact load sequence, compare them with the zero reference value, extract the nodes whose derivatives cross zero and whose second derivatives are negative as candidate load extreme points, extract the zero reference crossing points from the pavement deformation sequence as candidate displacement extreme points, and generate an extreme point location index set.

[0064] The aligned impact load sequence data is extracted from the aforementioned time-aligned state array, and continuous load values ​​are read at a fixed sampling interval of 0.01 seconds. For two adjacent time nodes in the sequence, the load value of the current time node is extracted and subtracted from the load value of the previous time node. The resulting load difference is then divided by the 0.01-second time step parameter between the two time nodes to calculate the first-order nodal derivative value. This first-order nodal derivative value is compared with a preset zero reference value. Considering the slight fluctuations in actual sensing data, when the first-order nodal derivative value is within a small range of -0.05 to +0.05, it is determined to be a zero-crossing state. Next, the first-order nodal derivative values ​​immediately before and after this zero-crossing point are extracted, and the difference operation is performed again and divided by the time step parameter to derive and calculate the second-order nodal derivative value of the corresponding node. Nodes with second-order nodal derivative values ​​that are exactly less than 0 are extracted and directly marked as candidate load extreme points. For example, the load value at 0.05 seconds is 50,000 Newtons, and at 0.06 seconds it is 50,100 Newtons. The load difference of 100 Newtons divided by the time step of 0.01 seconds yields a first-order nodal derivative value of 10,000. As the sequence moves forward, the first-order nodal derivative value at 0.12 seconds decreases to 0.02, satisfying the zero-crossing condition. The corresponding second-order nodal derivative value is calculated to be -25,000, which is less than 0, thus confirming it as an extreme point. For the road deformation sequence, continuous displacement node measurements are extracted, and the difference between the previous and subsequent displacement measurements is calculated to obtain the displacement difference result. The sign change of two adjacent displacement difference results is monitored. When a situation occurs where the previous value is greater than the zero reference value and the immediately following value is less than the zero reference value, a zero reference crossing phenomenon has occurred. This crossing time node is extracted as a candidate displacement extreme point. By integrating the spatial coordinates and temporal location labels of all the candidate load extreme points and candidate displacement extreme points identified above, and performing ordered stacking and duplicate item cleanup according to the time stamp order, a unified extreme location index set is generated.

[0065] S103: Based on the extreme value location index set, call the candidate load extreme value points and candidate displacement extreme value points, retrieve the node values ​​of the impact load sequence and the road deformation sequence at the corresponding extreme value point positions, perform key-value pair mapping and pairing of the load node values ​​and displacement node values ​​in time order, and generate the load displacement response sequence.

[0066] The previously constructed extreme value location index set is read, and the location labels corresponding to the calibrated candidate load extreme value points and candidate displacement extreme value points are extracted item by item. Using the location labels with timestamp information as address pointers, the actual measured values ​​of the load nodes for the impact load sequence at the corresponding time are retrieved from the underlying database, along with the actual measured values ​​of the displacement nodes for the road deformation sequence at the corresponding time. Due to the microsecond-level difference in sensor response time, the extracted load extreme value occurrence time and displacement extreme value occurrence time are often not exactly the same. By setting a maximum tolerance search window parameter, based on a stress wave propagation rate measurement value of 5 milliseconds for the road material, the closest load extreme value and displacement extreme value are found and associated within a 5-millisecond window. The retrieved load node values ​​are extracted and set as primary key elements, and the corresponding displacement node values ​​are extracted and set as secondary key elements. For example, the index set location tags reveal a load extremum at 45000 milliseconds with a load node value of 65000 Newtons, and a displacement extremum at 45003 milliseconds with a displacement node value of 1.2 millimeters. Since the time difference of 3 milliseconds is less than the 5 millisecond tolerance window, 65000 Newtons and 1.2 millimeters are paired. Following the monotonically increasing order of the underlying timestamps, the paired primary and secondary key elements are mapped and assembled using key-value pairs to form a combined data element containing both the applied mechanical force and the deformation response. This process is repeated, iterating through all tags in the entire extremum location index set to complete the data retrieval and mapping assembly of all extremum locations. All generated combined data elements are then arranged in chronological order to construct a continuous load-displacement response sequence.

[0067] Please see Figure 3 The specific steps of S2 are as follows:

[0068] S201: Call the load displacement response sequence to extract the timestamp and displacement node values. Based on the timestamp, divide the displacement node values ​​into multiple test periods along the time axis. Perform subtraction on the corresponding displacement node values ​​in the data of two adjacent test periods to calculate the difference in change. Arrange the displacement difference set in chronological order.

[0069] From the constructed load-displacement response sequence, the timestamp parameter and displacement node values ​​at the key positions are extracted for each combined data element. Based on the duration reflected by the extracted timestamp parameter, the continuous displacement node values ​​are segmented along the main axis of the time series. The benchmark time parameter for dividing the test cycle is set to 10 seconds, and the value of this benchmark time parameter is obtained from the actual measured statistics of a single loading and reset complete cycle of the drop hammer device. Using this 10-second step length, the long sequence is divided into multiple independent test cycle data blocks. Within the adjacent first and second test cycle data blocks, the displacement node values ​​at the same phase time point are extracted. The displacement node values ​​of the corresponding phase in the second test cycle data block are used as the minuend, and the displacement node values ​​of the same phase in the first test cycle data block are used as the subtrahend. Basic subtraction is performed to calculate the difference in displacement change of the nodes of the same phase during these two loading processes. For example, the displacement node value with a relative time point of 2.5 seconds in the first test cycle is extracted as 1.3 mm, and the displacement node value with the same relative time point of 2.5 seconds in the immediately following second test cycle is extracted as 1.5 mm. By performing a subtraction operation, subtracting 1.3 mm from 1.5 mm, the change difference corresponding to this phase position is calculated as 0.2 mm. This process is repeated for all phase-corresponding nodes in these two test cycles to calculate the change difference for all nodes. The resulting phase change differences are then linearly arranged according to their relative time order within the test cycle, forming a difference data column reflecting the evolution characteristics of a single cycle. The above cross-cycle phase alignment and subtraction operation is repeated for the displacement node values ​​of all adjacent cycles to generate a complete displacement difference set.

[0070] S202: Extract the change difference within the displacement difference set item by item and compare it with the preset deformation gradient threshold. Filter the extraction items whose difference is greater than the preset deformation gradient threshold. Extract the associated timestamp and displacement value. Perform aggregation and encapsulation on the timestamp, change difference and displacement value to generate a mutation data vector.

[0071] All variation difference data records are read item by item from the aforementioned displacement difference set, and a statistical feature scan of the global data is initiated. The maximum boundary value, minimum boundary value, and arithmetic mean of all variation difference data within the displacement difference set are calculated. A sorting algorithm is used to monotonically sort all variation differences in the set from low to high based on their absolute numerical magnitude, constructing an ordered difference distribution array. A preset quantile interval parameter is defined within the ordered difference distribution array. By selecting historical damaged road section displacement difference samples for probability density fitting analysis, the preset quantile interval parameter is strictly set to the 85% to 90% range. The arithmetic mean of each variation difference data within this quantile interval is extracted, and this mean is assigned as the preset deformation gradient threshold. For example, if the set contains 1000 variation differences, after sorting by size, the 851st to 900th difference data are precisely extracted. If the arithmetic mean of these 50 difference data is 0.8 mm, then 0.8 mm is directly set as the preset deformation gradient threshold. Subsequently, the entire displacement difference set is re-traversed, and each independent change difference is rigorously compared with a preset deformation gradient threshold of 0.8 mm. Data extraction items corresponding to all change differences with values ​​exactly greater than 0.8 mm are precisely selected. The original timestamp parameters associated with the selected data extraction items, as well as the basic displacement values ​​within the corresponding test period, are extracted. For discrete parameters, the timestamp parameters are used as retrieval identifiers, and the change difference and displacement values ​​are used together as feature loads. Multi-field combination, arrangement, and structured encapsulation operations are performed in contiguous memory addresses to ultimately generate a high-dimensional mutation data vector.

[0072] S203: Match and locate the corresponding load values ​​in the load displacement response sequence based on the timestamp in the mutation data vector, establish a ternary key value mapping of the load values, change difference and displacement values ​​according to the timestamp index, and perform feature splicing and one-dimensional sequence flattening transformation to generate local deformation mutation features.

[0073] The generated mutation data vector is obtained, and the timestamp parameter used as a retrieval identifier is extracted. Using this timestamp parameter as the positioning anchor, a full sequence traversal search is performed in the earliest generated load-displacement response sequence stored at the underlying level. Through a timestamp-complete consistency matching rule, the response data node corresponding to the same time point is located, and the original load numerical parameter at that node is extracted. After extracting the load numerical parameters, the retrieved load numerical parameters are structurally integrated with the existing change difference parameters and displacement numerical parameters in the mutation data vector. Using the positioning anchor timestamp as the unique index key, the load value, change difference, and displacement value are set as mapping attribute values, establishing a ternary key-value mapping table containing one primary key and three attribute fields. For example, using the timestamp of 45000 milliseconds as the index key for matching and positioning, the applied load value at that time is extracted to be 65000 Newtons, the original change difference in the mutation data vector is 0.9 mm, and the basic displacement value is 1.4 mm. These three physical parameters are strictly bound to the 45000 millisecond timestamp to form a complete ternary mapping data set. For each set of established ternary mapping data, the load values, variation differences, and displacement values ​​are concatenated using a memory-level vector feature concatenation operation in a fixed order to generate an initial feature matrix with three dimensions. Then, a one-dimensional sequence flattening transformation is performed on this initial feature matrix, which forcibly stretches and rearranges the originally multi-column feature data according to row and column order into a single-row continuous one-dimensional floating-point array. Finally, standardized local deformation mutation features are output to the system.

[0074] Please see Figure 4 The specific steps of S3 are as follows:

[0075] S301: Obtain local deformation abruptness characteristics, extract periodic index to read historical test data, retrieve actual rebound displacement value and rebound recovery time series, denoise the actual rebound displacement value and align it along the rebound recovery time series, and establish rebound time response parameters.

[0076] The previously flattened local deformation abrupt change feature data stream is read, and a periodic index parameter containing time stamp information is extracted. This periodic index parameter is used as a retrieval command to connect to the road surface detection historical database server to read the corresponding historical test data archive. In the historical test data archive, records of each loading and unloading process that match the spatial station number of the current test section are accurately retrieved, and the actual rebound displacement value sequence and the corresponding rebound recovery time sequence of the unloading stage are extracted. Since the actual rebound displacement values ​​are mixed with high-frequency environmental interference noise caused by the vibration of the test vehicle engine, a median filtering algorithm is used to smooth and reduce noise in the extracted actual rebound displacement values. Specifically, a data sliding window with a length of 5 sampling points is set, and the center value of the 5 actual rebound displacement values ​​within the window is taken to replace the original value of the center node, thereby eliminating pulse spike noise. For example, the rebound displacement values ​​extracted for 5 consecutive time nodes are 0.6 mm, 0.7 mm, 1.5 mm, 0.65 mm, and 0.68 mm, where 1.5 mm is a significant noise abrupt change spike. After sorting by size, the median value is selected as 0.68 mm. Therefore, 1.5 mm is replaced with 0.68 mm to complete local noise reduction. All actual rebound displacement values ​​after filtering and noise reduction are strictly aligned frame-by-frame along the timestamp axis of the corresponding rebound recovery time series. This ensures that each displacement data has a precise and unique recovery time label, thereby establishing a complete set of comprehensive rebound time response parameters in the storage module. The advantage of this operational logic is that, through targeted retrieval from the historical database combined with the powerful noise reduction method of median sliding window, electromechanical coupling clutter interference attached to weak rebound signals is eliminated, significantly restoring the true elastic recovery trajectory of the road surface structure.

[0077] S302: Call the rebound time response parameters, extract the actual rebound displacement value and time node, input the time node into the preset attenuation benchmark model to calculate the preset theoretical recovery displacement, subtract the preset theoretical recovery displacement from the actual rebound displacement value to obtain the discrete displacement deviation and rearrange it to generate a dynamic displacement deviation sequence.

[0078] The previously constructed rebound time response parameter set is invoked, and the actual rebound displacement value after noise reduction and the strictly paired time node parameters are extracted from it. The extracted time node parameters are directly input into a pre-calibrated preset attenuation benchmark model to calculate the preset theoretical recovery displacement parameter that the road surface should reach at that moment. The operation mechanism of the preset attenuation benchmark model is as follows: the initial maximum displacement parameter at the instant the road surface begins to unload is obtained, and combined with the currently input time node parameter and the fixed attenuation coefficient parameter determined by the mechanical properties of the road material, the result of multiplying the fixed attenuation coefficient by the time node parameter is negative, and the negative value of the natural constant is calculated to construct the natural exponential attenuation term. Then, the initial maximum displacement parameter is multiplied by the natural exponential attenuation term to calculate and output the preset theoretical recovery displacement parameter. For example, if the initial maximum displacement parameter is read as 2.0 mm and the fixed attenuation coefficient parameter is set to 0.5, when the input time node parameter is 1.2 seconds, the attenuation coefficient 0.5 is multiplied by the time 1.2 seconds to obtain 0.6. The natural constant raised to the power of -0.6 is approximately equal to 0.548. Multiplying the initial displacement of 2.0 mm by this natural exponential decay term of 0.548 yields a pre-defined theoretical recovery displacement of 1.096 mm. The actual rebound displacement measured at 1.2 seconds, at the same time point, is extracted and assumed to be 0.8 mm. Subtracting the theoretical value of 1.096 mm from 0.8 mm yields a discrete displacement deviation of -0.296 mm. This theoretical value calculation and subtraction operation is repeated for each time point to obtain a large number of discrete displacement deviation results. All discrete displacement deviation results are then reordered and reassembled in memory according to the ascending order of the time points, ultimately generating a continuous dynamic displacement deviation sequence.

[0079] S303: Extract discrete displacement deviation from dynamic displacement deviation sequence, obtain node spacing as integration step size, multiply discrete displacement deviation by integration step size to obtain micro-element area, accumulate micro-element surface along rebound recovery time sequence to obtain cumulative deviation amount and reshape cumulative deviation amount to obtain rebound attenuation deviation characteristics.

[0080] From the aforementioned constructed and arranged dynamic displacement deviation sequence, independent discrete displacement deviation parameter data are extracted sequentially. Based on the sampling frequency setting parameters of the underlying high-frequency measurement and control system, a strictly fixed time interval parameter between two adjacent consecutive acquisition time nodes is read and assigned as the basic integration step size parameter for numerical integration. The discrete displacement deviation parameter value at the current time node is extracted and multiplied by the preset basic integration step size parameter to calculate the rectangular micro-element area parameter enclosed by the displacement deviation curve within this extremely short time segment on the coordinate plane. For example, if the actual value of the discrete displacement deviation parameter at a certain time node is precisely extracted to be 0.3 mm, and the node spacing set by the sensor hardware clock is read to be 0.02 seconds, this is set as the integration step size parameter. By directly multiplying the deviation value of 0.3 mm by the integration step size of 0.02 seconds, the micro-element area parameter corresponding to this tiny instant is derived and calculated to be 0.006 mm / s. For hundreds or thousands of valid data nodes within the dynamic displacement deviation sequence, the precise multiplication and differential element operation described above is repeatedly performed cyclically. Subsequently, strictly following the positive direction of the monotonic extension of the rebound recovery time sequence, all calculated differential element area parameters are continuously summed to obtain a cumulative deviation parameter that reflects the overall kinetic energy loss deviation throughout the entire unloading elastic recovery cycle. Finally, this scalar cumulative deviation parameter is expanded using data structure tensor quantization, and by introducing blank padding vectors, it is reshaped and expanded into an input tensor with multi-dimensional matrix dimensions, ultimately generating a rebound attenuation deviation feature with spatiotemporal representation capabilities.

[0081] Please see Figure 5 The specific steps of S4 are as follows:

[0082] S401: Obtain the rebound attenuation deviation feature, extract the amplitude component of the local deformation abrupt feature, map the amplitude component to the preset interval through normalization to obtain the weight matrix, multiply the rebound attenuation deviation feature with the weight matrix to obtain the weighted tensor, perform mean filtering smoothing on the weighted tensor, and generate a joint attenuation numerical set.

[0083] Read the aforementioned multidimensional rebound attenuation deviation features and simultaneously extract the amplitude component parameters contained in the local deformation mutation features from the previous steps. Start the extreme value normalization mapping algorithm program to search for the absolute maximum and absolute minimum values ​​of all extracted amplitude component parameter sets. Subtract the minimum value from the current amplitude component value and divide the difference by the range between the maximum and minimum values ​​to strictly suppress and map the original amplitude component data to a preset dimensionless interval of 0 to 1. Construct a structured weight matrix parameter from the numerical set generated after the above normalization process according to the original spatial node coordinates. To demonstrate the mapping distribution of the amplitude components, mapping interval comparison data is introduced as an empirical supplement; the specific parameter correspondence is shown in Table 1.

[0084] Table 1: Normalized Mapping Interval Parameter Table

[0085] Test point number Original amplitude components Minimum value within a set Maximum value in set Normalization operation logic Weight matrix numerical values Measurement point 1 2.5 mm 0.5 mm 4.5 mm (2.5-0.5) / (4.5-0.5) 0.50 Measurement point 2 1.5 mm 0.5 mm 4.5 mm (1.5-0.5) / (4.5-0.5) 0.25 Measurement point 3 4.1 mm 0.5 mm 4.5 mm (4.1-0.5) / (4.5-0.5) 0.90

[0086] Referring to the specific calculation mapping logic listed in Table 1, when the original amplitude component of measurement point number 1 is 2.5 mm, combined with the minimum value of 0.5 mm and the maximum value of 4.5 mm, a dimensionless weight value of 0.50 can be generated after mapping operation and filled into the corresponding matrix node. Subsequently, the read rebound attenuation deviation feature tensor matrix and the weight matrix parameter just calculated are subjected to element-wise multiplication intersection operation. By multiplying each deviation feature with the sensitivity weight of the corresponding spatial location, a weighted tensor parameter that enhances the features of the sudden anomaly area is calculated. Mean filtering smoothing is performed on the generated weighted tensor parameter. Specifically, a 3×3 sliding window with a fixed receptive field is configured and slides across the weighted tensor parameter matrix with a step size of 1. Nine floating-point values ​​within the window coverage area are extracted, and the arithmetic mean of these nine values ​​is calculated. This arithmetic mean is used to replace the original numerical feature at the center of the window. After completing the mean replacement operation across the entire tensor plane, a joint attenuation value set that condenses the local smooth attenuation attributes is output and generated.

[0087] S402: Call the joint attenuation numerical set configuration sliding window and move to read the internal values, compare the central value with the neighboring values, mark the node where the central value is greater than the neighboring value as a local maximum value, parse the grid index of the local maximum value to map the longitude, latitude and elevation coordinates, and establish a three-dimensional coordinate coefficient set.

[0088] The previously processed joint decay value set is invoked, and a dynamic search window with a size of 5×5 nodes is configured in memory for the two-dimensional data plane of this value set. This search window is controlled to move alternately to the right and downwards according to a set unit step, starting from the initial origin coordinates of the data plane, to read the 25 feature values ​​covered within the window. When the window stops, the central value parameter located at the exact center of the window is extracted, along with the 24 neighboring value parameters distributed around this central node. A traversal comparison is performed, comparing the central value parameter with each of the 24 neighboring value parameters. Only when the central value parameter is absolutely greater than all 24 neighboring value parameters is the grid node containing the central value parameter marked as a local maximum. For example, when the sliding window moves to a specific monitoring area, the central numerical parameter reading is 8.6, while the largest among the 24 neighboring numerical parameters is only 7.9. Since 8.6 is definitely greater than 7.9, this node is calibrated as satisfying the local maximum condition, and its two-dimensional grid index parameter at row 120 and column 35 is recorded. All grid index parameters bound to the confirmed calibrated local maximum conditions are extracted. The underlying mapping spatial database conversion interface is connected, and the row and column grid index parameters are substituted into the spatial inverse calculation process. Combining the pre-stored road segment reference point data and grid physical resolution parameters, the corresponding real-world longitude, latitude, and elevation coordinates of this node are directly derived and converted analytically. The spatial three-dimensional positioning coordinates after the analysis of all local maximum conditions are summarized and arranged according to their distance from the origin, establishing a three-dimensional coordinate coefficient set with spatial pointing capability in the underlying control system.

[0089] S403: Call the three-dimensional coordinate coefficient set, retrieve the joint attenuation value corresponding to the local maximum value, pair it with the coordinate data to establish a spatial mapping key-value pair, construct feature patches based on the adjacency relationship of the spatial mapping key-value pair, compress the dimension of the feature patches and perform vector encapsulation to obtain the road structure hidden danger identification.

[0090] The process involves invoking the previously constructed set of 3D coordinate coefficients through reverse computation, while simultaneously tracing back and retrieving the joint attenuation numerical parameters corresponding to each local maximum value in the previous processing flow. For each independent spatial hazard hotspot, the extracted 3D latitude, longitude, and elevation coordinate data are used as the primary identifier key for geospatial data, and the corresponding joint attenuation numerical parameter is used as the subordinate numerical key for physical attributes. A data binding operation based on the same spatiotemporal dimension is performed, precisely and forcibly pairing the primary identifier key and subordinate numerical key to establish a spatial mapping key-value pair data structure containing location and intensity information. For example, if the extracted 3D coordinates are found to be 116.3 degrees East longitude, 39.9 degrees North latitude, and 55 meters elevation, and the associated joint attenuation numerical parameter subordinate numerical key is level 8.6, the attenuation intensity of level 8.6 is strictly attached to this latitude, longitude, and elevation point through program control, completing the assembly of a single paired data. All generated spatial mapping key-value pair information is then collected. Using a spatial connectivity criterion, the physical straight-line distance between the three-dimensional coordinates of each key-value pair is calculated. When the Euclidean distance between two points is less than a set proximity threshold of 0.5 meters, the two nodes are considered to have a physical adjacency relationship. Based on this spatial adjacency relationship, discrete maxima points that are close in location and suffer severe attenuation are delineated by polygonal edge envelopes and merged with regional connectivity to construct closed feature patch parameters that can intuitively reflect the scope of the damage within the virtual geographic space plane. Subsequently, the geometric feature extraction algorithm interface is activated to extract three core parameters of the feature patch parameters: edge perimeter, encapsulation area, and centroid of the attenuation gradient. This forcibly compresses and reorganizes the originally complex high-dimensional graphic structure. The compressed and extracted core parameters are then sequentially loaded with linear sequences and encapsulated into one-dimensional continuous vectors, ultimately outputting a simplified and high-quality pavement structure hazard indicator.

[0091] Please see Figure 6 The specific steps of S5 are as follows:

[0092] S501: Obtain the attenuation distribution vector of the pavement structure hazard identification, evaluate the discrete variance value based on the degree of deviation of the vector value, compare it with the preset classification benchmark threshold, and if it is less than the preset classification benchmark threshold, it is classified into the interlayer loose category; otherwise, it is classified into the structural void category, and a hazard attribute division set is generated.

[0093] The pavement structure hazard identifiers output by the aforementioned dimensionality reduction encapsulation are obtained, and the attenuation distribution vector data containing information on the degree of structural damage is extracted from them. For this multidimensional continuous vector, the values ​​of all distribution elements are extracted, and the arithmetic mean of these values ​​is calculated. Based on this mean, the value of each independent element in the attenuation distribution vector is subtracted from the mean to obtain the deviation data. Each deviation data point is squared, and the results of all squares are summed. Finally, the sum is divided by the total number of elements in the vector. Based on this deviation assessment, the discrete variance value, representing the severity of numerical fluctuations, is calculated. To visually illustrate the calculation process of the discrete variance value, specific test evaluation experimental data is listed in Table 2.

[0094] Table 2: Evaluation Test Table for Discrete Variance of Hidden Danger and Disease Parameters

[0095] Test station number Extracting the elements of the attenuation distribution vector Arithmetic mean Sum of squared deviations Numerical Discrete Variance Section 1 1.2、1.5、1.8、1.4 1.475 0.1975 0.0494 Section 2 2.5、6.8、3.1、8.2 5.150 25.1300 6.2825 Section 3 1.1、1.3、1.2、1.4 1.250 0.0500 0.0125

[0096] Referring to the calculation results listed in Table 2, the discrete variance of section 2 is as high as 6.2825, indicating that its internal attenuation distribution is extremely uneven. A pre-calibrated classification benchmark threshold was extracted. This benchmark threshold was obtained by statistically analyzing the variance of core samples from 50 known defects and was explicitly set to 1.5. The calculated discrete variance values ​​of each road section were logically compared with the preset classification benchmark threshold of 1.5. When the calculated discrete variance value of the target is clearly less than the 1.5 threshold (e.g., 0.0494 for section 1), it is determined that the internal medium of the pavement exhibits relatively uniform overall degradation, and it is classified into the loose interlayer category. Conversely, when the discrete variance value is greater than or equal to the 1.5 threshold (e.g., 6.2825 for section 2), it indicates that extreme subsidence and voids have occurred in local areas, and it is decisively classified into the structural void category. The hazard attribute labels of the classified road sections were summarized and stored to generate a complete hazard attribute classification set.

[0097] S502: Call the hazard attribute division set, parse the spatial geographic coordinates of the elements, read the site supervision station number table to extract the coordinates of the foundation calibration point, evaluate the distance parameter based on the positional difference between the spatial geographic coordinates and the coordinates of the foundation calibration point, bind the hazard information with the calibration point corresponding to the minimum value in the distance parameter, and establish a station number association file.

[0098] The system retrieves the attribute classification record file generated by attribute identification, extracts the attribute tags for each specific defect, and analyzes the spatial geographic absolute coordinate parameters of the elements bound to them. It then connects to the on-site deployment and control information network of the engineering supervision bureau and reads the on-site supervision stationing table file stored in the database. From this stationing table file, it extracts the coordinate parameters of each foundation calibration point fixed at 50-meter intervals along the highway centerline. Using the spherical distance great circle calculation logic, it calculates the spherical straight-line distance between the obtained spatial geographic absolute coordinates of the defect element and the coordinates of all foundation calibration points in the stationing table. It compares all the calculated straight-line distance data and obtains the spacing parameter with the smallest value. The specific foundation calibration point corresponding to this smallest spacing parameter is identified as the physical jurisdiction reference closest to the defect location. For example, the spatial geographic absolute coordinates of a structural void hazard are extracted. By sequentially calculating the spherical distances to the K100+000, K100+050, and K100+100 calibration points along the line, the corresponding distances are found to be 45 meters, 12 meters, and 38 meters, respectively. Through numerical comparison, 12 meters is extracted as the minimum spacing parameter. Based on this, the nearest basic calibration point for this hazard is determined to be at K100+050. An information binding operation is performed, forcibly anchoring and associating the hazard information data segment containing core parameters such as hazard type and void severity, along with the 12-meter minimum spacing parameter, with the identification code of the K100+050 basic calibration point. All hazard alarm data items and station number identifiers corresponding to the completed point bindings are integrated and permanently entered into the backend server in a unified archive data package format, formally establishing a station number association archive usable on-site for engineering supervision. The advantage of this operational logic is that, through the inverse solution of spatial spherical distance and the minimum distance search algorithm, it successfully converts the latitude and longitude coordinates of potential hazards that are difficult to locate directly into the linear highway mileage markers that construction supervisors are accustomed to using.

[0099] S503: Extract the initial state parameters of the category from the chainage-related files, obtain the decay coefficient matrix through the preset degradation mechanism model, modulate the initial state parameters of the category based on the decay coefficient matrix to obtain the evolution rate vector, define the intervention time nodes according to the numerical level of the evolution rate vector, and logically integrate with the chainage-related files to generate an engineering monitoring and management strategy.

[0100] A step-by-step search was conducted on the aforementioned station-related files to extract the initial state parameters representing the current deterioration level of potential hazards. A pre-built degradation mechanism model, fitted using 100,000 fatigue loading experiments, was activated within the system. The environmental temperature variables, average daily traffic load pressure, and material gradation parameters of the target area's pavement were input into the model. The model then used the Weibull distribution fatigue equation to calculate and output a decay coefficient matrix reflecting the fatigue damage resistance characteristics of different pavement layers. This multidimensional decay coefficient matrix was extracted and subjected to a matrix multiplication modulation operation with the previously obtained initial state parameters. By amplifying or reducing the characteristic amplitude of the original state parameters through coefficient weighting, an evolution rate vector accurately representing the future expansion speed of pavement defects was obtained. To clearly demonstrate the guiding role of the evolution rate in subsequent intervention actions, relevant intervention grading verification data were introduced for interpretation. The specific decision grading criteria are shown in Table 3.

[0101] Table 3: Evolution Rate Hierarchical Control Intervention Decision Table

[0102] Extract the mean of the evolution rate vector Rate level determination criteria Disease spread assessment level Define intervention time points Greater than 0.8 mm per month Rapid deterioration level Dangerous critical state Forced milling within 3 days 0.3 to 0.8 mm per month Medium-speed expansion tier Moderate alert status Grouting reinforcement will be arranged within 15 days. Less than 0.3 mm per month Slow evolutionary hierarchy Observe follow-up status Surface sealing to be performed within 60 days

[0103] Referring to the logical hierarchical specifications listed in Table 3, the intervals are divided based on the mean value of the elements within the evolution rate vector obtained through matrix calculation, compared with the set numerical levels. If the calculated mean rate is 0.6 mm per month, falling into the medium-speed expansion level, then grouting reinforcement is directly designated as a mandatory intervention time node parameter within 15 days from the strategy library. The generated specific intervention time nodes, supporting treatment plans, and corresponding spatial stationing associated archive data are comprehensively integrated and packaged along the execution logic dimension to uniformly generate an output engineering monitoring and management strategy document for direct dispatch and scheduling by the engineering quality supervision department.

[0104] Please see Figure 7 A road surface quality inspection system for engineering supervision, comprising:

[0105] The load deformation monitoring module acquires the impact load sequence and pavement deformation sequence for the road surface measuring points of the engineering supervision, and performs differentiation operation to extract the load peak position and maximum displacement position, and constructs the load displacement response sequence.

[0106] The deformation mutation extraction module calculates the displacement change difference between adjacent test cycles in the load displacement response sequence, extracts the terms whose displacement change difference is greater than the preset deformation gradient threshold, and constructs local deformation mutation features.

[0107] The attenuation deviation calculation module obtains the characteristics of sudden changes in local deformation, collects the actual rebound displacement and rebound recovery time of the corresponding period, calculates the displacement deviation between the actual rebound displacement and the preset theoretical recovery displacement, and performs integral calculation of the displacement deviation along the rebound recovery time to construct the rebound attenuation deviation characteristics.

[0108] The hazard identification generation module performs weighted coupling on the rebound attenuation deviation features based on the local deformation abruptness characteristics to obtain the joint attenuation value, filters the local maximum value items to extract the corresponding three-dimensional coordinate coefficients and associates them with the joint attenuation value to generate pavement structure hazard identification;

[0109] The control strategy generation module categorizes road structure hazards into loose interlayer and void structural categories using discrete variance numerical clustering based on their identification. After binding the site supervisor's station number, it predicts the evolution trend of hazards based on a preset degradation mechanism model and generates an engineering monitoring and management strategy.

[0110] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of protection of the described technical solutions.

Claims

1. A method for road surface quality inspection in engineering supervision, characterized in that, Includes the following steps: S1: Obtain the impact load sequence and pavement deformation sequence for the road surface measuring points of the engineering supervision, and perform differentiation to extract the load peak position and maximum displacement position, and construct the load displacement response sequence; S2: Calculate the displacement change difference between adjacent test cycles in the load displacement response sequence, extract the terms whose displacement change difference is greater than the preset deformation gradient threshold, and construct local deformation abrupt change features; S3: Obtain the local deformation abrupt change characteristics, collect the actual rebound displacement and rebound recovery time of the corresponding period, calculate the displacement deviation between the actual rebound displacement and the preset theoretical recovery displacement, and perform integral calculation on the displacement deviation along the rebound recovery time to construct the rebound attenuation deviation characteristics. S4: Based on the local deformation abruptness features, perform weighted coupling on the rebound attenuation deviation features to obtain the joint attenuation value, filter the local maximum value items to extract the corresponding three-dimensional coordinate coefficients and associate them with the joint attenuation value to generate a pavement structure hidden danger identifier; S5: Based on the discrete variance numerical clustering of the pavement structure hazard identification, the hazards are divided into interlayer loose category and structural void category. After binding the on-site supervision station number, the evolution trend of the hazards is predicted based on the preset degradation mechanism model, and an engineering monitoring and management strategy is generated.

2. The method for road surface quality testing for engineering supervision according to claim 1, characterized in that, The load-displacement response sequence includes the load peak time, maximum displacement amplitude, and phase difference. The local deformation abrupt change characteristics include the abrupt change amplitude, abrupt change location, and abrupt change frequency. The rebound attenuation deviation characteristics include the displacement deviation integral area, attenuation rate, and recovery hysteresis. The recovery hysteresis specifically refers to the time difference between the actual rebound recovery time and the theoretical recovery time, reflecting the degree of hysteresis in the elastic recovery ability of the pavement material. The pavement structure hazard identification includes extreme three-dimensional coordinates, joint attenuation value, and spatial distribution density. The joint attenuation value specifically refers to the comprehensive attenuation quantification value obtained by weighted coupling of the local deformation abrupt change characteristics and the rebound attenuation deviation characteristics. The engineering monitoring and management strategy includes hazard category attribution, evolution trend prediction results, and corresponding disposal measures suggestions.

3. The method for road surface quality testing for engineering supervision according to claim 1, characterized in that, The specific steps of S1 are as follows: S101: Collect the impact load sequence and road deformation sequence of the road measurement points in the sample dataset, compare the timestamps of the two sequences frame by frame, remove outlier measurement point samples whose time difference is greater than the deviation benchmark value, map the measurement points with overlapping timestamps to key-value pairs according to the measurement point number and the corresponding timestamp, and generate a time-aligned state array. S102: Call the time-aligned state array, extract the node derivative values ​​of the impact load sequence by taking the derivative and compare them with the zero reference value. Extract the nodes whose derivatives cross zero and whose second derivatives are negative as candidate load extreme points. Extract the zero reference crossing points of the road deformation sequence by differential extraction as candidate displacement extreme points and generate an extreme value location index set. S103: Based on the extreme value location index set, call the candidate load extreme value points and candidate displacement extreme value points, retrieve the node values ​​of the impact load sequence and the road deformation sequence at the corresponding extreme value point positions, perform key-value pair mapping and pairing of the load node values ​​and displacement node values ​​in time order, and generate the load displacement response sequence.

4. The road surface quality testing method for engineering supervision according to claim 3, characterized in that, The specific steps of S2 are as follows: S201: Call the load displacement response sequence to extract the timestamp and displacement node values, divide the displacement node values ​​into multiple test cycles along the time axis according to the timestamp, perform subtraction operation on the corresponding displacement node values ​​in the data of two adjacent test cycles to calculate the difference in change, and arrange them in time order to generate a displacement difference set. S202: Extract the change difference within the displacement difference set item by item and compare it with the preset deformation gradient threshold. Filter the extraction items whose difference is greater than the preset deformation gradient threshold. Extract the associated timestamp and displacement value. Perform aggregation and encapsulation on the timestamp, change difference, and displacement value to generate a mutation data vector. S203: Based on the timestamps in the mutation data vector, match and locate them in the load displacement response sequence, extract the corresponding load values, establish a ternary key value mapping of the load values, change differences and displacement values ​​according to the timestamp index, and perform feature splicing and one-dimensional sequence flattening transformation to generate local deformation mutation features.

5. The method for road surface quality testing for engineering supervision according to claim 4, characterized in that, The preset deformation gradient threshold is determined based on the statistical distribution of all variation differences within the displacement difference set. The statistical distribution includes the maximum, minimum, and average values ​​of the variation differences. The variation differences located in the preset quantile interval are selected as the preset deformation gradient threshold after sorting the variation differences according to their numerical size.

6. The method for road surface quality testing for engineering supervision according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Obtain the local deformation abruptness characteristics, extract the periodic index to read historical test data, retrieve the actual rebound displacement value and rebound recovery time sequence, reduce the noise of the actual rebound displacement value and align it along the rebound recovery time sequence, and establish rebound time response parameters. S302: Call the rebound time response parameters, extract the actual rebound displacement value and time node, input the time node into the preset attenuation benchmark model to calculate the preset theoretical recovery displacement, subtract the preset theoretical recovery displacement from the actual rebound displacement value to obtain discrete displacement deviation and rearrange it to generate a dynamic displacement deviation sequence. S303: Extract discrete displacement deviation from the dynamic displacement deviation sequence, obtain the node spacing as the integration step size, multiply the discrete displacement deviation by the integration step size to obtain the micro-element area, accumulate the micro-element surface along the rebound recovery time sequence to obtain the cumulative deviation amount, and reshape the cumulative deviation amount to obtain the rebound attenuation deviation characteristics.

7. The method for road surface quality testing for engineering supervision according to claim 6, characterized in that, The preset attenuation benchmark model obtains the preset theoretical recovery displacement by multiplying the initial displacement by a natural exponential attenuation term constructed based on time nodes and attenuation coefficients.

8. The method for road surface quality testing for engineering supervision according to claim 6, characterized in that, The specific steps of S4 are as follows: S401: Obtain the rebound attenuation deviation feature, extract the amplitude component of the local deformation abrupt feature, map the amplitude component to a preset interval through normalization to obtain a weight matrix, multiply the rebound attenuation deviation feature with the weight matrix to obtain a weighted tensor, perform mean filtering smoothing on the weighted tensor, and generate a joint attenuation value set. S402: Call the joint attenuation value set configuration sliding window and move to read the internal values, compare the center value with the neighboring values, mark the node where the center value is greater than the neighboring value as a local maximum value, parse the grid index of the local maximum value to map the longitude, latitude and elevation coordinates, and establish a three-dimensional coordinate coefficient set. S403: Call the three-dimensional coordinate coefficient set, retrieve the joint attenuation value corresponding to the local maximum value, pair it with the coordinate data to establish a spatial mapping key-value pair, construct feature patches based on the adjacency relationship of the spatial mapping key-value pair, compress the dimension of the feature patches and perform vector encapsulation to obtain the road structure hazard identification.

9. The method for road surface quality testing for engineering supervision according to claim 8, characterized in that, The specific steps of S5 are as follows: S501: Obtain the attenuation distribution vector of the pavement structure hazard identification, evaluate the discrete variance value based on the degree of deviation of the vector value, compare it with the preset classification benchmark threshold, and if it is less than the preset classification benchmark threshold, it is classified into the interlayer loose category; otherwise, it is classified into the structural void category, and a hazard attribute division set is generated. S502: Call the aforementioned hazard attribute division set, parse the spatial geographic coordinates of the elements, read the site supervision station number table to extract the coordinates of the foundation calibration point, evaluate the distance parameter based on the positional difference between the spatial geographic coordinates and the coordinates of the foundation calibration point, bind the hazard information with the calibration point corresponding to the minimum value in the distance parameter, and establish a station number association file. S503: Extract the initial state parameters of the category from the chainage-related files, obtain the decay coefficient matrix through a preset degradation mechanism model, modulate the initial state parameters of the category based on the decay coefficient matrix to obtain the evolution rate vector, define the intervention time nodes according to the numerical level of the evolution rate vector, and logically integrate with the chainage-related files to generate an engineering monitoring and management strategy.

10. A road surface quality inspection system for engineering supervision, characterized in that, The system is used to implement the road surface quality testing method for engineering supervision as described in any one of claims 1-9, the system comprising: The load deformation monitoring module acquires the impact load sequence and pavement deformation sequence for the road surface measuring points of the engineering supervision, and performs differentiation operation to extract the load peak position and maximum displacement position, and constructs the load displacement response sequence. The deformation mutation extraction module calculates the displacement change difference between adjacent test cycles in the load displacement response sequence, extracts the terms whose displacement change difference is greater than a preset deformation gradient threshold, and constructs local deformation mutation features. The attenuation deviation calculation module obtains the local deformation abrupt change characteristics, collects the actual rebound displacement and rebound recovery time of the corresponding period, calculates the displacement deviation between the actual rebound displacement and the preset theoretical recovery displacement, and performs integral calculation of the displacement deviation along the rebound recovery time to construct the rebound attenuation deviation characteristics. The hazard identification generation module performs weighted coupling on the rebound attenuation deviation feature based on the local deformation abruptness feature to obtain the joint attenuation value, filters the local maximum value item to extract the corresponding three-dimensional coordinate coefficient and associates it with the joint attenuation value to generate a pavement structure hazard identification; The control strategy generation module divides the pavement structure hazard identification into interlayer loose category and structural void category through discrete variance numerical clustering. After binding the on-site supervision station number, it predicts the evolution trend of hazard based on the preset degradation mechanism model and generates the engineering monitoring and management strategy.

Images