Nondestructive testing method and system for deep compaction degree of large-thickness water-stable base

Abstract

The present application relates to the field of engineering detection technology, in particular to a deep compaction degree nondestructive testing method and system for large-thickness water-stable base course. A multi-source heterogeneous dataset of a to-be-detected area is obtained using an integrated detection device; a construction process compliance index and a process uniformity index of the to-be-detected area are obtained; each data source in the multi-modal nondestructive testing data is deeply corrected to obtain a deep correction compaction degree estimation set of each data source; the deep correction compaction degree estimation set of each data source is input into an adaptive weight fusion model to obtain a preliminary compaction degree distribution map; a fusion compaction degree distribution map of the to-be-detected area and a dynamic confidence score of each spatial point are calculated; the compaction quality of the to-be-detected area is determined, and a detection strategy optimization instruction is generated for the area with a confidence score lower than a confidence threshold. The present application can realize high-precision, high-confidence and engineering-actual nondestructive testing of deep compaction degree of large-thickness water-stable base course.

Description

Technical Field

[0001] This invention relates to the field of engineering testing technology, and in particular to a method and system for non-destructive testing of deep compaction degree for thick water-stabilized base courses. Background Technology

[0002] Currently, non-destructive testing technology for deep compaction of thick (≥30cm) water-stabilized base courses is urgently needed in engineering construction. Existing methods, such as the general multi-source fusion scheme represented by patent publication number CN121142017A, improve data comprehensiveness, but still have shortcomings, including: designing for shallow asphalt pavement testing, failing to address core physical challenges such as drastic signal attenuation with depth and strong interference from water content in water-stabilized materials; the data-driven fusion model it relies on lacks physical constraints for deep testing scenarios, resulting in insufficient confidence and poor interpretability of the results in engineering applications; and such schemes completely disregard the decisive influence of construction process parameters on compaction state, thus deviating from engineering practice.

[0003] Therefore, developing a specialized method that deeply integrates physical mechanisms and process knowledge, and can quantify the confidence level of test results and provide decision guidance, has become an urgent need to break through the bottleneck of quality control of thick water-stabilized base courses. Summary of the Invention

[0004] This invention addresses the bottlenecks in deep testing, result interpretation, and engineering adaptability in existing technologies by providing a non-destructive testing method and system for deep compaction of thick water-stabilized base courses.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0006] In a first aspect, the present invention provides a method and system for non-destructive testing of deep compaction degree for thick water-stabilized base courses, comprising: using integrated testing equipment to acquire a multi-source heterogeneous dataset of the area to be tested, wherein the multi-source heterogeneous dataset includes at least a construction process parameter set, a material property parameter set, a multimodal non-destructive testing dataset, and a process image dataset;

[0007] Based on the construction process parameter set and the process image dataset, the construction process compliance index and process uniformity index of the area to be detected are obtained through analysis.

[0008] The multimodal nondestructive testing dataset is preprocessed, and the depth of each data source in the multimodal nondestructive testing dataset is corrected based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source.

[0009] The depth-corrected compaction degree estimation sets from each data source are input into the adaptive weight fusion model to obtain a preliminary compaction degree distribution map. The adaptive weight fusion model dynamically adjusts the fusion weights of different data sources based on the depth of the detection points and material property parameters.

[0010] Based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point are calculated.

[0011] Based on preset compaction thresholds and confidence thresholds, and combining the fused compaction distribution map with the dynamic confidence score, the compaction quality of the area to be detected is determined, and detection strategy optimization instructions are generated for areas with confidence scores lower than the confidence threshold.

[0012] Optionally, an integrated detection device can be used to acquire a multi-source heterogeneous dataset of the region to be detected, including:

[0013] An integrated detection equipment network consisting of construction parameter acquisition terminals, multi-sensor detection arrays, data communication interfaces, and high-definition camera equipment is configured at the construction site.

[0014] The construction parameter acquisition terminal collects the vibration frequency, amplitude, travel speed, number of compaction passes, and compaction trajectory coordinates obtained by the satellite positioning system in real time during the operation of the road roller, forming the construction process parameter set.

[0015] Through the data communication interface, the cement dosage, key parameters of the gradation curve, and material transportation time of each batch of mixture are obtained in real time from the production control system of the mixing plant, forming the set of material characteristic parameters;

[0016] The multi-sensor detection array synchronously collects Rayleigh wave data generated by the acceleration sensor array arranged on the surface of the water-stabilized base course, compaction index inverted by the vibration wheel response generated by the continuous compaction control system, and dielectric constant profile data obtained by multi-frequency ground-penetrating radar scanning, forming the multi-modal non-destructive testing dataset.

[0017] The high-definition camera equipment is used to collect high-definition texture images of the surface of the paved mixture and continuous video images of the roller compaction process, forming the process image dataset.

[0018] Optionally, based on the construction process parameter set and the process image dataset, the construction process compliance index and process uniformity index of the area to be inspected are analyzed and obtained, including:

[0019] The vibration frequency, amplitude, and travel speed collected in real time are compared with the pre-set process design specifications for the construction of thick water-stabilized base courses, and the vibration frequency compliance score, amplitude compliance score, and travel speed compliance score are calculated respectively.

[0020] Based on the rolling trajectory coordinates, the coverage integrity of the actual rolling trajectory over the designed rolling area is calculated, and the overlap width between adjacent rolling zones is analyzed to meet the theoretical requirements, thus obtaining the trajectory overlap rate.

[0021] Image analysis is performed on the high-definition texture image of the surface of the paved mixture to identify the uneven texture areas in the image caused by aggregate segregation, and the percentage of the area of ​​the uneven texture area to the total area of ​​the image is calculated as the segregation area percentage.

[0022] The construction process compliance index is calculated by weighting the vibration frequency compliance score, the amplitude compliance score, and the travel speed compliance score together with the trajectory overlap rate.

[0023] The value obtained by subtracting the percentage of the segregated area from 1 is defined as the process uniformity index.

[0024] Optionally, based on a preset effective detection depth attenuation coefficient, depth correction is performed on each data source in the multimodal nondestructive testing data to obtain a depth-corrected compaction estimate set for each data source, including:

[0025] The method for constructing the effective detection depth attenuation coefficient includes: during construction of a representative test section, density sensing elements are buried at different preset depths, and the original signals of various non-destructive testing devices on the surface are collected simultaneously. A correlation model between the characteristic values ​​of various signals and the actual density of the buried points is established. The variation of the parameters of the correlation model with detection depth, material moisture content and gradation is analyzed, and the effective detection depth attenuation coefficient is obtained through function fitting.

[0026] Optionally, the depth correction of each data source in the multimodal nondestructive testing data is performed based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source, further comprising:

[0027] For any detection data in the multimodal nondestructive testing dataset, based on the corresponding signal type and frequency, the corresponding effective detection depth attenuation coefficient function is called, and combined with the target detection depth of the current detection point, material moisture content and gradation parameters, the correction coefficient for converting the multimodal nondestructive testing data from the surface response value to the compaction estimate at the target depth is calculated. The depth of each data source in the multimodal nondestructive testing data is then corrected, and finally the depth-corrected compaction estimate set is obtained.

[0028] Optionally, the depth-corrected compaction degree estimates from each data source are input into an adaptive weighted fusion model to obtain a preliminary compaction degree distribution map, which includes:

[0029] Each nondestructive testing data source participating in the fusion is assigned a basic weight based on historical verification data, which is determined according to the average reliability level of the corresponding data source under ideal conditions;

[0030] Based on the current depth of the detection point, the basic weight of each data source is dynamically adjusted: for data sources whose physical principle sensitivity to the surface meets a preset standard, their weight is reduced as the depth increases according to a preset decay function; for data sources whose physical principle penetration capability to deep information meets a preset standard, their weight is adjusted within a specific depth range, and the adjusted weight is greater than or equal to the basic weight corresponding to the data source, wherein the specific depth range is a high-precision depth range set based on the physical principle characteristics of each data source detection method.

[0031] The weights are adjusted a second time based on the set of material property parameters: if, according to a preset standard, a certain data source shows that it is sensitive to the current material property parameter in historical data, then the fusion weight of the corresponding data source for that material property parameter is reduced accordingly.

[0032] The adjusted weights of each data source are normalized, and the depth-corrected compaction estimate of the corresponding data source is weighted and averaged to obtain the compaction estimate of each spatial point in the preliminary compaction distribution map.

[0033] Based on the estimated compaction degree, the preliminary compaction degree distribution map is generated.

[0034] Optionally, based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, a fused compaction distribution map of the area to be tested and a dynamic confidence score for each spatial point are calculated, including:

[0035] A Bayesian data fusion framework was adopted, and the spatial compaction estimate and its uncertainty provided by the preliminary compaction distribution map were used as the observation likelihood function.

[0036] The construction process compliance index and the process uniformity index are spatially interpolated and mapped to generate a probability map reflecting the prior quality distribution of construction, which serves as the prior distribution.

[0037] The prior distribution and the observed likelihood function are fused using Bayes' theorem to calculate the posterior probability distribution. The expected value map of this posterior probability distribution is then used as the fused compaction distribution map.

[0038] Optionally, based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, calculating the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point further includes:

[0039] The first confidence level is obtained based on the dispersion of the estimated values ​​after depth correction from various data sources at the target spatial points.

[0040] The construction process compliance index of the area where the target spatial point is located is used as the second confidence level;

[0041] The third confidence level is calculated based on the spatiotemporal rate of change of material moisture content in the region where the target spatial point is located.

[0042] The first confidence level, the second confidence level, and the third confidence level are fused based on preset weighting coefficients to calculate the dynamic confidence score for each spatial point.

[0043] Optionally, based on preset compaction thresholds and confidence thresholds, and combining the fused compaction distribution map with the dynamic confidence score, the compaction quality of the area to be tested is determined, and a detection strategy optimization instruction is generated for areas with confidence scores lower than the confidence threshold, including:

[0044] Pre-set the lower and upper thresholds for the acceptable compaction range, as well as the minimum confidence threshold required for the judgment result to be reliable;

[0045] Traverse the fused compaction distribution map, and for each spatial point in the map, determine whether its compaction estimate falls within the preset qualified range, and at the same time determine whether its dynamic confidence score reaches or exceeds the minimum confidence threshold.

[0046] If the estimated compaction degree of a spatial point is within the acceptable range and the dynamic confidence score meets the standard, then the compaction quality of the corresponding spatial point is deemed acceptable.

[0047] If the compaction estimate of a spatial point is within the acceptable range, but its dynamic confidence score does not meet the standard, the quality of the corresponding spatial point is determined to be unverified, and one or more detection strategy optimization instructions are automatically generated. The detection strategy optimization instructions include at least: suggesting the use of a destructive method of core drilling at the corresponding spatial point for result calibration and verification; suggesting the adjustment of the parameter settings of the non-destructive testing equipment and rescanning the corresponding area to be tested; and prompting the use of spatial statistical analysis in conjunction with the data of the surrounding qualified spatial points to assist in inference.

[0048] If the estimated compaction value of a spatial point is lower than the lower limit of the acceptable range or higher than the upper limit, the quality of the corresponding spatial point is directly determined to be unacceptable, and the corresponding spatial point is prominently marked in the visualization map output by the system. At the same time, a construction remedial instruction containing specific location information is generated.

[0049] Secondly, the present invention provides a non-destructive testing system for deep compaction of thick water-stabilized base courses, comprising:

[0050] The multi-source data acquisition module is used to acquire a multi-source heterogeneous dataset of the area to be inspected using integrated testing equipment. The multi-source heterogeneous dataset includes at least a construction process parameter set, a material property parameter set, a multimodal nondestructive testing dataset, and a process image dataset.

[0051] The construction process evaluation module is used to analyze and obtain the construction process compliance index and process uniformity index of the area to be inspected based on the construction process parameter set and the process image dataset.

[0052] The original data correction module is used to preprocess the multimodal nondestructive testing dataset and perform depth correction on each data source in the multimodal nondestructive testing data based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source.

[0053] The multi-source data fusion module is used to input the depth-corrected compaction degree estimation set of each data source into the adaptive weight fusion model to obtain a preliminary compaction degree distribution map. The adaptive weight fusion model dynamically adjusts the fusion weights of different data sources according to the depth of the detection point and the material property parameters.

[0054] The test result analysis module is used to calculate the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point based on the preliminary compaction distribution map, the construction process compliance index and the process uniformity index.

[0055] The compaction quality determination module is used to determine the compaction quality of the area to be tested based on a preset compaction degree threshold and confidence degree threshold, combined with the fused compaction degree distribution map and the dynamic confidence degree score, and to generate detection strategy optimization instructions for areas with confidence degree scores lower than the confidence degree threshold.

[0056] By implementing this invention, it is possible to acquire a multi-source heterogeneous dataset of the area to be inspected using integrated testing equipment. The multi-source heterogeneous dataset includes at least a set of construction process parameters, a set of material property parameters, a multimodal nondestructive testing dataset, and a process image dataset. This breaks the limitations of single data sets, provides comprehensive and realistic engineering scenario data support for subsequent analysis, and avoids detection bias caused by missing data.

[0057] By implementing this invention, it is possible to analyze and obtain the construction process compliance index and process uniformity index of the area to be inspected based on the construction process parameter set and the process image dataset; to associate the construction process with the compaction quality, so that the test results are consistent with the actual project, and to provide a priori basis for the construction level for subsequent compaction analysis.

[0058] By implementing this invention, it is possible to preprocess the multimodal nondestructive testing dataset and perform depth correction on each data source in the multimodal nondestructive testing data based on a preset effective detection depth attenuation coefficient, thereby obtaining a depth-corrected compaction degree estimation set for each data source; this solves the core physical problem of deep detection of thick water-stabilized base courses and improves the accuracy and reliability of compaction degree estimation for each data source.

[0059] By implementing this invention, it is possible to input the depth-corrected compaction degree estimation sets of various data sources into an adaptive weighted fusion model to obtain a preliminary compaction degree distribution map. The adaptive weighted fusion model dynamically adjusts the fusion weights of different data sources according to the depth of the detection points and material property parameters, thereby avoiding the limitations of fixed weight fusion and allowing data sources with different depths and material properties to play their optimal role, generating a more accurate preliminary compaction degree distribution.

[0060] By implementing this invention, it is possible to calculate the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index; to achieve deep integration of physical mechanisms, process knowledge, and test data, while clarifying the credibility of the test results and improving the interpretability of the results.

[0061] By implementing this invention, it is possible to determine the compaction quality of the area to be tested based on preset compaction thresholds and confidence thresholds, combined with the fused compaction distribution map and the dynamic confidence score, and generate detection strategy optimization instructions for areas with confidence scores lower than the confidence threshold; accurately distinguish between qualified, unverified, and unqualified areas, providing a clear decision basis for quality control, and improving subsequent detection efficiency through optimization instructions.

[0062] In summary, by implementing this invention, high-precision, high-confidence, and engineering-practice-compliant non-destructive testing of the deep compaction degree of thick water-stabilized base courses can be achieved. It can accurately quantify the compaction quality distribution, clarify the reliability of the results, and guide subsequent testing optimization, thus breaking through the bottlenecks of traditional technologies in deep testing, result interpretability, and engineering adaptability. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating the non-destructive testing method for deep compaction degree of thick water-stabilized base courses provided by the present invention.

[0064] Figure 2 This is a schematic diagram of the structure of the non-destructive testing system for deep compaction of thick water-stabilized base courses provided by the present invention.

[0065] In the attached diagram, the components represented by each number are as follows:

[0066] Multi-source data acquisition module 11, construction process evaluation module 12, raw data correction module 13, multi-source data fusion module 14, test result analysis module 15, and compaction quality judgment module 16. Detailed Implementation

[0067] Example 1, as Figure 1 As shown, this embodiment of the invention provides a non-destructive testing method for deep compaction of thick water-stabilized base courses, including:

[0068] S100: Use integrated testing equipment to acquire a multi-source heterogeneous dataset of the area to be tested. The multi-source heterogeneous dataset includes at least a construction process parameter set, a material property parameter set, a multimodal nondestructive testing dataset, and a process image dataset.

[0069] S200: Based on the construction process parameter set and the process image dataset, analyze and obtain the construction process compliance index and process uniformity index of the area to be detected;

[0070] S300: Preprocess the multimodal nondestructive testing dataset, and perform depth correction on each data source in the multimodal nondestructive testing data based on the preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimation set for each data source;

[0071] S400: Input the depth-corrected compaction degree estimation set of each data source into the adaptive weight fusion model to obtain a preliminary compaction degree distribution map, wherein the adaptive weight fusion model dynamically adjusts the fusion weight of different data sources according to the depth of the detection point and the material property parameters.

[0072] S500: Based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, calculate the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point;

[0073] S600: Based on the preset compaction threshold and confidence threshold, and combining the fused compaction distribution map with the dynamic confidence score, the compaction quality of the area to be detected is determined, and a detection strategy optimization instruction is generated for areas with confidence scores lower than the confidence threshold.

[0074] In step S100 of this embodiment, the multi-source heterogeneous dataset of the region to be detected is obtained using an integrated detection device, including:

[0075] An integrated detection equipment network consisting of construction parameter acquisition terminals, multi-sensor detection arrays, data communication interfaces, and high-definition camera equipment is configured at the construction site.

[0076] The construction parameter acquisition terminal collects the vibration frequency, amplitude, travel speed, number of compaction passes, and compaction trajectory coordinates obtained by the satellite positioning system in real time during the operation of the road roller, forming the construction process parameter set.

[0077] Through the data communication interface, the cement dosage, key parameters of the gradation curve, and material transportation time of each batch of mixture are obtained in real time from the production control system of the mixing plant, forming the set of material characteristic parameters;

[0078] The multi-sensor detection array synchronously collects Rayleigh wave data generated by the acceleration sensor array arranged on the surface of the water-stabilized base course, compaction index inverted by the vibration wheel response generated by the continuous compaction control system, and dielectric constant profile data obtained by multi-frequency ground-penetrating radar scanning, forming the multi-modal non-destructive testing dataset.

[0079] The high-definition camera equipment is used to collect high-definition texture images of the surface of the paved mixture and continuous video images of the roller compaction process, forming the process image dataset.

[0080] In this embodiment, the purpose of step S100 is to comprehensively collect all kinds of key data required for the testing of thick water-stabilized base courses, providing complete and accurate basic data support for subsequent construction process evaluation, original data correction, multi-source data fusion, and compaction quality judgment. By integrating four types of data—construction process parameters, material characteristic parameters, non-destructive testing data, and process images—the problem of traditional testing methods having single data and being detached from engineering reality is solved. This ensures that subsequent compaction degree calculation and quality judgment can combine physical mechanisms and process knowledge, improving the confidence and interpretability of the test results.

[0081] To achieve the above objectives, it is first necessary to configure an integrated detection equipment network at the construction site, consisting of construction parameter acquisition terminals, multi-sensor detection arrays, data communication interfaces, and high-definition camera equipment.

[0082] An integrated testing equipment network was set up at the construction site according to a pre-designed layout. This network consists of four core types of equipment: construction parameter acquisition terminals, multi-sensor arrays, data communication interfaces, and high-definition cameras. All devices work together through a unified data transmission protocol to ensure synchronous acquisition, real-time transmission, and centralized storage of different types of data, avoiding analysis errors caused by data acquisition delays or asynchrony. Specifically, the construction parameter acquisition terminals are installed on the road roller; the multi-sensor arrays are arranged in a grid pattern on the surface of the water-stabilized base course; the data communication interfaces are connected to the mixing plant's production control system via a dedicated line; and the high-definition cameras are mounted at high points around the testing area to achieve comprehensive, blind-spot-free imaging.

[0083] Next, the vibration frequency, amplitude, travel speed, number of compaction passes, and compaction trajectory coordinates obtained by the satellite positioning system are collected in real time through the construction parameter acquisition terminal to form the construction process parameter set.

[0084] This means that by installing a construction parameter acquisition terminal on the road roller, key operating parameters and position trajectory parameters of the road roller during operation can be captured in real time, so as to comprehensively reflect the execution of the construction process.

[0085] The specific parameters collected include: vibration frequency, amplitude, travel speed, number of compaction passes, and the real-time coordinates of the compaction trajectory obtained through a satellite positioning system.

[0086] The data acquisition principle is as follows: The construction parameter acquisition terminal has built-in vibration sensors, speed sensors and positioning modules. The vibration sensors sense the vibration frequency and amplitude of the road roller in real time, the speed sensors record the travel speed, and the positioning module updates the rolling trajectory coordinates once per second. The terminal packages and uploads the data to the data storage center at a preset frequency, and accumulates to form a set of construction process parameters.

[0087] For example, when the road roller is operating, the vibration sensor collects a vibration frequency of 38Hz and an amplitude of 1.8mm, and the speed sensor records that the travel speed is stable at 2.5km / h. After 1 hour of operation, the terminal counts 5 compaction passes, and the satellite positioning system records the compaction trajectory coordinate sequence as (X118.7632, Y32.0547), (X118.7635, Y32.0549)...(X118.7689, Y32.0592). This continuous data fully presents the operating status and coverage area of ​​the road roller.

[0088] Then, through the data communication interface, the cement dosage, key parameters of the gradation curve, and material transportation time of each batch of mixture are obtained in real time from the production control system of the mixing plant, forming the set of material characteristic parameters;

[0089] This involves establishing a real-time data interaction channel with the mixing plant's production control system via a data communication interface. This allows for the direct acquisition of core characteristic parameters of each batch of mixture from production to transportation, ensuring the authenticity and timeliness of the data and providing a basis for subsequent analysis of the material's impact on compaction.

[0090] Specific parameters collected include: cement dosage, key parameters of the gradation curve, and material transportation time.

[0091] The data collection principle is as follows: After each batch of mixed material is produced, the production control system of the mixing plant automatically records the cement addition ratio of that batch, i.e., the cement dosage, and the passing rate of each level of sieve, i.e., the key parameters of the gradation curve. After the material is loaded onto the truck, the transport vehicle sends the departure time to the detection system through the on-board terminal and the arrival time at the construction site. The system automatically calculates the material transportation time, and all data is synchronized in real time to the material property database of the detection system through the data communication interface.

[0092] For example, during the production of a certain batch of mixture, the production control system of the mixing plant recorded a cement dosage of 4.5%, and the passing rates of the 2.36mm sieve and 0.075mm sieve in the key parameters of the gradation curve were 35% and 7%, respectively. The material was dispatched from the mixing plant at 9:00 AM and arrived at the construction site at 9:35 AM. The system automatically calculated the material transportation time as 35 minutes. These data together constitute the set of material characteristic parameters for this batch of mixture.

[0093] Furthermore, through the multi-sensor detection array, Rayleigh wave data generated by the acceleration sensor array arranged on the surface of the water-stabilized base course, compaction index inverted by the vibration wheel response generated by the continuous compaction control system, and dielectric constant profile data obtained by multi-frequency ground-penetrating radar scanning are synchronously collected to form the multi-modal non-destructive testing dataset;

[0094] This involves using a multi-sensor array deployed on the surface of the water-stabilized base layer to simultaneously collect three types of non-destructive testing data based on different principles. This data reflects the compaction status of the water-stabilized base layer from different dimensions, providing multi-source data support for subsequent in-depth correction and fusion analysis.

[0095] The specific data collected include: Rayleigh wave data generated by the accelerometer array, compaction index inverted from the vibration wheel response generated by the continuous compaction control system, and dielectric constant profile data obtained from multi-frequency ground-penetrating radar scanning.

[0096] The acquisition principle is as follows: an accelerometer array is arranged in a 2m×2m grid, and Rayleigh waves are generated by a vibration device. The sensors record the propagation time and amplitude of the Rayleigh waves in the water-stabilized base course, forming Rayleigh wave data; the continuous compaction control system is linked with the vibrating wheel of the road roller, and the interaction response between the vibrating wheel and the base course is sensed by the sensor, and the compaction index is obtained by inversion; the multi-frequency ground-penetrating radar emits electromagnetic waves of different frequencies, receives the electromagnetic wave signals reflected by the base course, analyzes and obtains the dielectric constant at different depths, forming dielectric constant profile data. The three types of data are collected synchronously and correlated to the same spatial coordinate.

[0097] For example, an accelerometer array collected Rayleigh wave propagation speed of 280 m / s at a certain detection point, the continuous compaction control system inverted the compaction index of the point to be 93, and the multi-frequency ground-penetrating radar detected a dielectric constant of 5.8 at a depth of 5 cm, 6.2 at a depth of 15 cm, and 6.5 at a depth of 30 cm. These data reflect the compaction state of the detection point from three dimensions: wave propagation characteristics, mechanical response, and electromagnetic characteristics.

[0098] Finally, the high-definition camera equipment is used to collect high-definition texture images of the surface of the paved mixture and continuous video images of the roller compaction process to form the process image dataset.

[0099] This involves using high-definition cameras installed around the testing area to capture real-time images of the surface condition of the mixture after paving and the entire compaction process by the road roller, providing visual evidence for subsequent analysis of construction uniformity and identification of segregation phenomena.

[0100] The specific data collected includes: high-resolution texture images of the surface of the paved mixture and continuous video footage of the roller compaction process.

[0101] The data acquisition method is as follows: The high-definition camera equipment uses an industrial camera with a resolution of no less than 4K, and the shooting frequency is set to 30 frames per second. After the paving is completed, the camera takes full-coverage pictures of the detection area to obtain surface texture images. When the road roller is compacting, the camera tracks and captures the compaction trajectory and changes in the base surface to form continuous video images. The image data is stored according to timestamps and spatial regions and is associated with other datasets.

[0102] For example, after the mixture is laid, the high-definition camera equipment captures a high-definition texture image of the surface of a certain area, in which the distribution of aggregates can be clearly observed; during the compaction process, continuous video footage records the complete process of the roller compacting from the edge of the detection area to the center, as well as the change in the smoothness of the base surface after compaction. This visual data provides a direct basis for subsequent calculation of the segregation area ratio and verification of the compaction trajectory coverage.

[0103] In step S200 of this embodiment, based on the construction process parameter set and the process image dataset, the construction process compliance index and process uniformity index of the area to be detected are analyzed and obtained, including:

[0104] The vibration frequency, amplitude, and travel speed collected in real time are compared with the pre-set process design specifications for the construction of thick water-stabilized base courses, and the vibration frequency compliance score, amplitude compliance score, and travel speed compliance score are calculated respectively.

[0105] Based on the rolling trajectory coordinates, the coverage integrity of the actual rolling trajectory over the designed rolling area is calculated, and the overlap width between adjacent rolling zones is analyzed to meet the theoretical requirements, thus obtaining the trajectory overlap rate.

[0106] Image analysis is performed on the high-definition texture image of the surface of the paved mixture to identify the uneven texture areas in the image caused by aggregate segregation, and the percentage of the area of ​​the uneven texture area to the total area of ​​the image is calculated as the segregation area percentage.

[0107] The construction process compliance index is calculated by weighting the vibration frequency compliance score, the amplitude compliance score, and the travel speed compliance score together with the trajectory overlap rate.

[0108] The value obtained by subtracting the percentage of the segregated area from 1 is defined as the process uniformity index.

[0109] In this embodiment, the purpose of step S200 is to quantitatively assess the basic construction quality of the thick water-stabilized base course from two key dimensions: construction execution standardization and process uniformity. The construction process compliance index is used to determine whether the roller operating parameters and compaction trajectory meet the preset process requirements, and the process uniformity index is used to identify the uniformity of aggregate distribution after the mixture is laid. Together, they constitute the quality assessment indicators for the construction process. These indices not only directly reflect the compliance and rationality of the construction process but also provide important prior information for subsequent compaction degree calculations. This ensures that the final compaction quality judgment is based on both test data and the actual construction process, improving the engineering interpretability and confidence of the test results and avoiding misjudgments caused by relying solely on test data without considering the construction context.

[0110] To achieve the above objectives, firstly, the real-time collected vibration frequency, amplitude, and travel speed need to be compared with the pre-set process design specifications for the construction of thick water-stabilized base courses, and the vibration frequency compliance score, amplitude compliance score, and travel speed compliance score should be calculated respectively.

[0111] That is, based on the pre-set construction process design specifications for thick water-stabilized base courses, the key operational parameters collected in real time during the construction process are compared with these specifications, and the degree of compliance of each parameter is obtained through quantitative calculation.

[0112] It is necessary to formulate process design specifications in advance for the construction of thick water-stabilized base courses. For example, the vibration frequency specification is 35Hz to 45Hz, the amplitude specification is 1.5mm to 2.5mm, and the travel speed specification is 2km / h to 4km / h.

[0113] For example, the calculation method can adopt a deviation rate quantitative scoring method. If the actual collected value is within the standard value range, the deviation rate is 0 and the score is 100 points; if it exceeds the standard value, points are deducted linearly according to the deviation rate, the larger the deviation rate, the lower the score, and the lowest score is 0 points. For example, the deviation rate calculation formula is: when the actual value exceeds the upper limit: deviation rate = (actual value - upper limit of standard value) / upper limit of standard value; when the actual value is lower than the lower limit: deviation rate = (lower limit of standard value - actual value) / lower limit of standard value.

[0114] For example, if the actual vibration frequency is 38Hz, which is within the standard range of 35Hz to 45Hz, the vibration frequency compliance score is 100 points; if the actual amplitude is 1.4mm, which is lower than the standard lower limit of 1.5mm, the deviation rate is (1.5-1.4) / 1.5≈6.67%, and the amplitude compliance score is 100-6.67≈93.33 points; if the actual travel speed is 4.5km / h, which exceeds the standard upper limit of 4km / h, the deviation rate is (4.5-4) / 4=12.5%, and the travel speed compliance score is 100-12.5=87.5 points.

[0115] Next, based on the rolling trajectory coordinates, it is necessary to calculate the coverage integrity of the actual rolling trajectory over the designed rolling area, and analyze the degree of conformity between the overlap width between adjacent rolling zones and the theoretical requirements to obtain the trajectory overlap rate.

[0116] Based on the compaction trajectory coordinates of the construction process parameter set, the coverage integrity and overlap rationality of the actual compaction operation are analyzed, and the compliance of the compaction trajectory is quantified.

[0117] For example, the calculation method can be as follows: First, determine the total area to be covered based on the boundary coordinates of the designed compaction area; then, draw the actual compaction trajectory map through the compaction trajectory coordinate sequence, calculate the area covered by the actual trajectory, and the ratio of the two is the coverage integrity; at the same time, extract the edge coordinates of adjacent compaction zones, calculate the actual overlap width between adjacent zones, compare it with the theoretical overlap width, and obtain the overlap width conformity; finally, merge the coverage integrity and overlap width conformity with a weight of 5:5 to obtain the trajectory overlap rate.

[0118] For example, if the designed compaction area is rectangular with a total area of ​​1000 square meters, and the actual compaction area is 980 square meters based on the compaction trajectory coordinate analysis, the coverage is 980 / 1000 = 98%; the theoretical overlap width is 20cm, and the actual average overlap width of adjacent compaction zones is 18cm, resulting in an overlap width compliance of 18 / 20 = 90%; therefore, the trajectory overlap rate is 98% × 0.5 + 90% × 0.5 = 94%.

[0119] Then, image analysis is required on the high-definition texture image of the surface of the paved mixture to identify the uneven texture areas in the image caused by aggregate segregation, and to calculate the percentage of the area of ​​the uneven texture area to the total area of ​​the image as the segregation area percentage.

[0120] This involves performing image analysis on high-resolution texture images of the surface of the mixture after paving in a centralized process image dataset to identify areas of uneven texture caused by aggregate segregation and quantify the uniformity defects of the mixture paving.

[0121] The analysis logic is as follows: image segmentation and texture feature recognition algorithms are used to preprocess high-definition texture images to enhance the contrast between aggregates and mortar; by extracting features such as grayscale values, texture entropy, and local variance of the image, a texture uniformity judgment model is established to identify texture non-uniform areas caused by aggregate aggregation or absence, i.e., segregation areas; the pixel area of ​​the segregation area is calculated, and the ratio of the pixel area of ​​the segregation area to the total pixel area of ​​the image is the segregation area ratio.

[0122] For example, if the total pixel area of ​​a high-definition texture image of a certain paved area is 1 million pixels, and the pixel area of ​​the dissimilar area with uneven texture identified by image analysis is 30,000 pixels, then the proportion of the dissimilar area is 30,000 / 1 million = 3%.

[0123] Furthermore, the weighted average of the vibration frequency compliance score, the amplitude compliance score, and the travel speed compliance score needs to be combined with the trajectory overlap rate in a second weighted fusion to calculate the construction process compliance index.

[0124] The aforementioned scores and trajectory overlap rates are weighted and integrated to comprehensively quantify the overall compliance level of the construction process.

[0125] For example, the calculation logic is as follows: first, the vibration frequency compliance score, amplitude compliance score, and travel speed compliance score are calculated as a weighted average with a weight of 3:3:4. Then, this average value is combined with the trajectory overlap rate with a weight of 6:4 for a second weighted fusion. The final result is the construction process compliance index, with a value range of 0 to 100.

[0126] For example, given that the vibration frequency compliance score is 100, the amplitude compliance score is 93.33, and the travel speed compliance score is 87.5, the weighted average of the three is 100×0.3+93.33×0.3+87.5×0.4=30+27.999+35=92.999; the trajectory overlap rate is 94% (i.e., 94 points); then the construction process compliance index is 92.999×0.6+94×0.4≈55.799+37.6=93.4 points.

[0127] Finally, the value obtained by subtracting the segregation area percentage from 1 is defined as the process uniformity index.

[0128] This means directly quantifying the uniformity level of the mixture paving process based on the proportion of segregated area.

[0129] The calculation logic is that the process uniformity index is negatively correlated with the segregation area ratio, defined as 1 minus the segregation area ratio, with a value range of 0 to 1. The closer the value is to 1, the better the paving uniformity.

[0130] For example, if the segregation area accounts for 3%, the process uniformity index is 1-0.03=0.97; if the segregation area accounts for 8%, the process uniformity index is 1-0.08=0.92.

[0131] In step S300 of this embodiment, the depth of each data source in the multimodal nondestructive testing data is corrected based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source, including:

[0132] The method for constructing the effective detection depth attenuation coefficient includes: during construction of a representative test section, density sensing elements are buried at different preset depths, and the original signals of various non-destructive testing devices on the surface are collected simultaneously. A correlation model between the characteristic values ​​of various signals and the actual density of the buried points is established. The variation of the parameters of the correlation model with detection depth, material moisture content and gradation is analyzed, and the effective detection depth attenuation coefficient is obtained through function fitting.

[0133] For any detection data in the multimodal nondestructive testing dataset, based on the corresponding signal type and frequency, the corresponding effective detection depth attenuation coefficient function is called, and combined with the target detection depth of the current detection point, material moisture content and gradation parameters, the correction coefficient for converting the multimodal nondestructive testing data from the surface response value to the compaction estimate at the target depth is calculated. The depth of each data source in the multimodal nondestructive testing data is then corrected, and finally the depth-corrected compaction estimate set is obtained.

[0134] In this embodiment, the purpose of step S300 is to solve the core physical problems of severe signal attenuation with depth and interference from moisture content and gradation parameters in nondestructive testing of thick water-stabilized base courses. By constructing an effective detection depth attenuation coefficient, targeted depth correction is performed on multimodal nondestructive testing data, eliminating the deviation between the surface detection signal and the actual deep compaction state. This allows the detection results from each data source to accurately reflect the compaction level at the target depth, providing accurate and reliable basic data for subsequent adaptive weight fusion, avoiding distortion in deep compaction estimation caused by signal attenuation, and improving the accuracy and reliability of the final detection results.

[0135] To achieve the above objectives, it is first necessary to construct an effective detection depth attenuation coefficient.

[0136] The first step is to embed density sensing elements at predetermined depths during the construction of representative test sections. These predetermined depths must cover the typical testing range of a thick, water-stabilized base course, such as 5cm, 15cm, 30cm, and 40cm. Multiple density sensing elements should be evenly embedded in different areas at each depth to ensure data representativeness. Simultaneously, various non-destructive testing devices should be deployed on the surface of the test sections, working synchronously with the density sensing elements to collect raw surface signals.

[0137] For example, in three different areas of the test road section, piezoelectric density sensing elements were buried at depths of 5cm, 15cm, 30cm, and 40cm, respectively, for a total of 24 elements; accelerometers, continuous compaction control equipment, and multi-frequency ground-penetrating radar were arranged on the surface to simultaneously collect Rayleigh wave raw signals, vibration wheel response signals, and electromagnetic wave reflection signals.

[0138] The second step is to extract the feature values ​​of the original signals from various surface non-destructive testing equipment, such as the propagation speed of Rayleigh waves, the amplitude of the vibration wheel response signal, and the dielectric constant of electromagnetic wave signals. These feature values ​​are then matched with the actual density measured by the density sensing element at the corresponding depth to establish a correlation model between the signal feature values ​​and the actual density.

[0139] For example, for Rayleigh wave data, the Rayleigh wave propagation velocity corresponding to different depths is extracted and fitted with the actual density measured by the density sensing element at that depth to establish a linear correlation model between propagation velocity and actual density. The model expression is ρ=0.02v+1.6, where ρ is the actual density in grams per cubic centimeter and v is the Rayleigh wave propagation velocity in meters per second.

[0140] The third step involves repeating the data acquisition and modeling process by changing the material moisture content and gradation parameters of the test road section, and analyzing the variation of the correlation model parameters with detection depth, material moisture content, and gradation. For example, it was observed that as the detection depth increases, the slope of the correlation model between Rayleigh wave propagation velocity and true density gradually decreases; and as the moisture content increases, the model intercept slightly increases. Based on these patterns, a multivariate function fitting method was used to construct an effective detection depth attenuation coefficient function, which includes detection depth, material moisture content, and gradation parameters as independent variables, and the correlation model correction coefficient as the dependent variable.

[0141] For example, by fitting multiple sets of experimental data, the effective detection depth attenuation coefficient function of Rayleigh wave data is obtained as k = 0.98 - 0.002h + 0.001w - 0.0005g, where k is the attenuation coefficient, h is the detection depth in cm, w is the material moisture content in %; and g is the gradation parameter, taken as the 0.075mm sieve aperture pass rate in %.

[0142] Next, it is necessary to perform depth correction on each data source in the multimodal nondestructive testing data to obtain a depth-corrected compaction estimate set for each data source.

[0143] That is, by calling the existing effective detection depth attenuation coefficient function and combining it with the parameters of the actual detection scenario, the multimodal nondestructive testing data is corrected for each data source to obtain a set of estimated values ​​that accurately reflect the compaction degree of the target depth.

[0144] The first step is to identify the corresponding signal type and frequency for any test data in the multimodal nondestructive testing dataset, such as Rayleigh wave signal or 100MHz ground-penetrating radar signal. Then, extract the material moisture content and gradation parameters of the current test point from the material property parameter set, such as a moisture content of 6% and a 0.075mm sieve pass rate of 7%. At the same time, determine the target test depth of the current test point, such as 25cm.

[0145] The second step involves calling the corresponding effective detection depth attenuation coefficient function based on the signal type and frequency. The target detection depth, material moisture content, and gradation parameters are substituted into the function to calculate the correction coefficient for the detection data.

[0146] For example, if the target detection depth at a certain detection point is 25cm, the material moisture content is 6%, and the 0.075mm sieve aperture passing rate in the gradation parameters is 7%, then for the Rayleigh wave detection data at this point, the corresponding attenuation coefficient function k=0.98-0.002h+0.001w-0.0005g is called, and the calculated k=0.98-0.002×25+0.001×6-0.0005×7=0.98-0.05+0.006-0.0035=0.9325.

[0147] The third step involves converting the surface response values ​​of the multimodal nondestructive testing data into initial compaction estimates, which are then multiplied by a calculated correction coefficient to obtain the corrected compaction estimate at the target depth. This correction process is performed on all data sources within the multimodal nondestructive testing dataset, ultimately forming a depth-corrected compaction estimate set for each data source.

[0148] For example, the initial compaction estimate after converting the surface response value of a Rayleigh wave detection data is 92%. Substituting this into the correction coefficient of 0.9325, the calculated corrected compaction estimate at a target depth of 25cm is 92% × 0.9325 ≈ 85.8%. The initial compaction estimate after converting the dielectric constant data of the ground-penetrating radar at the same detection point is 94%. Using the attenuation coefficient function corresponding to the ground-penetrating radar, a correction coefficient of 0.91 is calculated, and the corrected compaction estimate is 94% × 0.91 ≈ 85.5%. The corrected estimates from each data source are summarized to form the depth-corrected compaction estimate set for that detection point. The whole set constitutes the depth-corrected compaction estimate set for each data source in the area to be detected.

[0149] In step S400 of this embodiment, the depth-corrected compaction degree estimation sets from each data source are input into the adaptive weighted fusion model to obtain a preliminary compaction degree distribution map, which includes:

[0150] Each nondestructive testing data source participating in the fusion is assigned a basic weight based on historical verification data, which is determined according to the average reliability level of the corresponding data source under ideal conditions;

[0151] Based on the current depth of the detection point, the basic weight of each data source is dynamically adjusted: for data sources whose physical principle sensitivity to the surface meets a preset standard, their weight is reduced as the depth increases according to a preset decay function; for data sources whose physical principle penetration capability to deep information meets a preset standard, their weight is adjusted within a specific depth range, and the adjusted weight is greater than or equal to the basic weight corresponding to the data source, wherein the specific depth range is a high-precision depth range set based on the physical principle characteristics of each data source detection method.

[0152] The weights are adjusted a second time based on the set of material property parameters: if, according to a preset standard, a certain data source shows that it is sensitive to the current material property parameter in historical data, then the fusion weight of the corresponding data source for that material property parameter is reduced accordingly.

[0153] The adjusted weights of each data source are normalized, and the depth-corrected compaction estimate of the corresponding data source is weighted and averaged to obtain the compaction estimate of each spatial point in the preliminary compaction distribution map.

[0154] Based on the estimated compaction degree, the preliminary compaction degree distribution map is generated.

[0155] In step S400 of this embodiment, the purpose of the above step is to scientifically integrate the depth-corrected compaction estimates from various data sources through an adaptive weighted fusion model, eliminating the limitations of a single data source and obtaining a more accurate and comprehensive preliminary compaction distribution. By dynamically adjusting the weights, different data sources can play their optimal role under suitable depth ranges and material characteristics, solving the problem that traditional fixed-weighted fusion cannot adapt to complex testing scenarios of thick water-stabilized base courses. This provides more reliable intermediate data support for subsequent fusion compaction calculations and quality judgments, improving the stability and accuracy of the overall testing results.

[0156] To achieve the above objectives, it is first necessary to assign a basic weight based on historical verification data to each non-destructive testing data source participating in the fusion. This basic weight is determined according to the average reliability level of the corresponding data source under ideal conditions.

[0157] This involves assigning a basic weight to each nondestructive testing data source participating in the fusion process. The magnitude of this basic weight reflects the average reliability level of the data source under ideal conditions, which include scenarios with uniform material, stable moisture content, and appropriate testing depth.

[0158] The basic weight setting logic is to calculate the degree of deviation between the estimated value of each data source and the actual value of the borehole core in the historical detection data. The smaller the deviation, the higher the basic weight. The sum of the basic weights of all data sources is 1.

[0159] For example, the data sources involved in the fusion include Rayleigh wave data, vibration wheel response inversion compaction index, and dielectric constant profile data. Based on historical data verification, the average deviation of Rayleigh wave data is 2.1%, the average deviation of vibration wheel response inversion compaction index is 1.8%, and the average deviation of dielectric constant profile data is 2.5%. Based on this, the basic weights are allocated as follows: Rayleigh wave data is 0.35, vibration wheel response inversion compaction index is 0.4, and dielectric constant profile data is 0.25.

[0160] Next, based on the depth of the current detection point, the basic weight of each data source is dynamically adjusted: for data sources whose physical principle sensitivity to the surface meets a preset standard, their weight is reduced as the depth increases according to a preset decay function; for data sources whose physical principle penetration capability to deep information meets a preset standard, their weight is adjusted within a specific depth range, and the adjusted weight is greater than or equal to the basic weight corresponding to the data source, wherein the specific depth range is a high-precision depth range set based on the physical principle characteristics of each data source detection method;

[0161] That is, based on the current depth of the detection point and combined with the physical characteristics of each data source, the basic weights are dynamically adjusted for the first time, so that the weights can adapt to the detection needs at different depths.

[0162] The adjustment logic is as follows: First, the physical characteristics of each data source are clearly categorized. For example, the physical principle of the vibration wheel response inversion compaction index shows that its sensitivity to the surface meets a preset standard, but the detection accuracy decreases with increasing depth. The physical principle of Rayleigh wave data shows that its penetration ability to deep information meets a preset standard, and its performance is stable within a depth range of 10cm to 35cm. Therefore, the corresponding specific depth range is 10cm to 35cm. Dielectric constant profile data has a strong penetration ability to deep layers, and its performance is better within a depth range of 25cm to 40cm. Therefore, the corresponding specific depth range is 25cm to 40cm. The preset attenuation function is linear attenuation, for example, the weighted attenuation coefficient = 1 - 0.02 × (depth - surface critical depth) (when the depth exceeds the surface critical depth).

[0163] For example, the target depth of a certain detection point is 30cm, and the critical depth of the surface layer is set to 10cm. The basic weight of the compaction index in the vibration wheel response inversion is 0.4. Since the depth exceeds the critical depth of the surface layer by 20cm, the attenuation coefficient is calculated according to the attenuation function as 1 - 0.02 × 20 = 0.6. After adjustment, the weight is 0.4 × 0.6 = 0.24. The basic weight of Rayleigh wave data is 0.35, which is within the advantageous depth range of 10cm to 35cm. The basic weight remains unchanged at 0.35. The basic weight of dielectric constant profile data is 0.25, which is within the advantageous depth range of 25cm to 40cm. The weight is increased according to the preset rules, and after adjustment, it is 0.3.

[0164] Then, the weights are adjusted a second time according to the set of material property parameters: if, according to the preset standard, a certain data source shows that it is sensitive to the current material property parameter in the historical data, then the fusion weight of the corresponding data source for that material property parameter is reduced accordingly.

[0165] This involves combining parameters from the material property parameter set to analyze the sensitivity of each data source to the current material properties, and then performing a secondary correction on the weights after deep adjustment to avoid introducing additional errors from sensitive data sources.

[0166] The adjustment logic is as follows: a preset sensitivity judgment standard is set. If, in historical data, the estimated value deviation of a material characteristic parameter increases by more than 30% when the change range of a certain data source exceeds 5%, then the data source is determined to be sensitive to that material characteristic parameter. For sensitive data sources, the weight is reduced by a preset proportion, ranging from 10% to 30%, depending on the strength of the sensitivity.

[0167] For example, in the current material property parameters at the testing point, the moisture content is 7.2%, which is approximately 24.1% different from the historical average moisture content of 5.8%. Historical data shows that the dielectric constant profile data is sensitive to moisture content; when the moisture content changes by more than 5%, the estimated value deviation increases by 35%. Therefore, the dielectric constant profile data is adjusted a second time, reducing the percentage by 25%, and the adjusted weight is 0.3 × (1 - 25%) = 0.225. The Rayleigh wave data and the vibration wheel response inversion compaction index are not sensitive to the current moisture content change, and their weights remain unchanged at 0.35 and 0.24, respectively.

[0168] Furthermore, the adjusted weights of each data source are normalized, and the depth-corrected compaction estimate of the corresponding data source is weighted and averaged using these weights to calculate the compaction estimate of each spatial point in the preliminary compaction distribution map.

[0169] That is, the weights of each data source after two adjustments are normalized to ensure that the sum of the weights is 1, and then the depth-corrected compaction estimate of each data source is weighted and averaged using the normalized weights to obtain the compaction estimate of each spatial point.

[0170] The calculation logic is as follows: Normalized weight = Adjusted weight of a certain data source ÷ Sum of adjusted weights of all data sources; Spatial point compaction estimate = Σ (Normalized weight of a certain data source × Depth-corrected compaction estimate of that data source).

[0171] For example, after two adjustments, the weights are: Rayleigh wave data 0.35, vibratory wheel response inversion compaction index 0.24, and dielectric constant profile data 0.225, with a total weight of 0.35 + 0.24 + 0.225 = 0.815. After normalization, the weights are: Rayleigh wave data ≈ 0.35 ÷ 0.815 ≈ 0.429, vibratory wheel response inversion compaction index ≈ 0.24 ÷ 0.815 ≈ 0.294, and dielectric constant profile data ≈ 0.225 ÷ 0.815 ≈ 0.276. The depth-corrected compaction estimates for each data source are 86.2% for Rayleigh wave data, 85.7% for vibratory wheel response inversion compaction index, and 87.1% for dielectric constant profile data. The estimated compaction degree of this spatial point is approximately 36.98% + 25.20% + 24.04% ≈ 86.22%.

[0172] Finally, based on the estimated compaction degree, the preliminary compaction degree distribution map is generated.

[0173] That is, traverse all spatial points in the area to be tested, repeat the above process of weight adjustment, normalization and compaction degree estimation to obtain the compaction degree estimate of each spatial point, and then use a spatial interpolation algorithm to convert the discrete spatial point estimates into a continuous distribution map, i.e., a preliminary compaction degree distribution map.

[0174] For example, the area to be tested is divided into a spatial grid of 2m×2m, containing a total of 1000 spatial points. After calculating the estimated compaction degree of the center point of each grid, the Kriging interpolation algorithm is used to fill in the blank areas between the grids to generate a complete preliminary compaction degree distribution map. Different compaction degree ranges are marked with different colors in the map. For example, 85% to 90% is the qualified range, marked with green; below 85% is marked with red; and above 90% is marked with yellow.

[0175] In step S500 of this embodiment, based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, a fused compaction distribution map of the area to be tested and a dynamic confidence score for each spatial point are calculated, including:

[0176] A Bayesian data fusion framework was adopted, and the spatial compaction degree estimates and their uncertainties provided by the preliminary compaction degree distribution map were used as observation likelihood functions.

[0177] The construction process compliance index and the process uniformity index are spatially interpolated and mapped to generate a probability map reflecting the prior quality distribution of construction, which serves as the prior distribution.

[0178] The prior distribution and the observed likelihood function are fused using Bayes' theorem to calculate the posterior probability distribution. The expected value map of this posterior probability distribution is then used as the fused compaction distribution map.

[0179] In this embodiment, the purpose of step S500 is to deeply integrate the preliminary compaction test data with the construction process quality assessment indicators to obtain a more realistic integrated compaction distribution map. By using a Bayesian data fusion framework, prior information corresponding to the construction process compliance index and process uniformity index is introduced to compensate for the limitations of single test data and reduce test uncertainty. Simultaneously, this allows the compaction assessment to be based not only on direct test results but also on the compliance and uniformity of the construction process, improving the accuracy and reliability of the final compaction distribution and providing more engineering-interpretive core data for subsequent quality judgment.

[0180] To achieve the above objectives, a Bayesian data fusion framework is first required, which uses the spatial compaction estimate and its uncertainty provided by the preliminary compaction distribution map as the observation likelihood function.

[0181] Based on the preliminary compaction distribution map, the compaction estimate of each point in space is extracted and its uncertainty is quantified. Together, they form the observation likelihood function required for Bayesian fusion, reflecting the credibility of the detection data itself.

[0182] The implementation logic is as follows: the compaction estimate of each spatial point in the preliminary compaction distribution map is the core observation value. Its uncertainty is calculated by deeply correcting the dispersion of the compaction estimate through each data source. The greater the dispersion, the higher the uncertainty, and the greater the variance of the observation likelihood function.

[0183] For example, at a spatial point in the area to be tested, the preliminary compaction estimate is 86.5%. The deeply corrected compaction estimates from the three data sources involved in the fusion are 85.8%, 86.7%, and 87.0%, respectively. The standard deviation of the dispersion index is calculated to be 0.61%. Therefore, the observation likelihood function for this point is set to a normal distribution with a mean of 86.5% and a variance of 0.61%, i.e., N(86.5, 0.61). 2 ).

[0184] Next, the construction process compliance index and the process uniformity index need to be spatially interpolated and mapped to generate a probability map reflecting the prior quality distribution of construction, which serves as the prior distribution.

[0185] This involves spatially interpolating and mapping the construction process compliance index and the process uniformity index to generate a probability map that reflects the prior quality distribution of construction. This map serves as the prior information for Bayesian fusion, reflecting the impact of the construction process on compaction quality.

[0186] The implementation logic is as follows: First, the construction process compliance index and process uniformity index have been calculated by region. For example, the construction process compliance index for a certain region is 93.4 points, and the process uniformity index is 0.97. A spatial interpolation algorithm is used to extend the discrete regional indices to every spatial point in the entire area to be inspected, obtaining the process compliance score and uniformity score for each spatial point. Then, the two scores are merged according to preset weights, such as process compliance accounting for 60% and uniformity accounting for 40%, to obtain the comprehensive construction quality score for each spatial point. Finally, a prior probability distribution is constructed based on the comprehensive score; the higher the comprehensive score, the higher the corresponding expected value of prior compaction and the smaller the variance.

[0187] For example, a spatial point, through interpolation, yields a process compliance score of 92.8 and a uniformity score of 0.96. The weighted composite score is calculated as follows: 92.8 × 0.6 + 0.96 × 100 × 0.4 = 55.68 + 38.4 = 94.08 points. Here, the uniformity index is converted to a percentage for calculation. According to the preset mapping rules, a composite score of 94 points corresponds to a priori expected compaction degree of 88% with a variance of 0.3%. Therefore, the prior distribution of this point is N(88, 0.3). 2 ).

[0188] Then, the prior distribution and the observed likelihood function are fused using Bayes' theorem to calculate the posterior probability distribution, and the expected value map of the posterior probability distribution is used as the fused compaction distribution map.

[0189] That is, by using Bayes' theorem, the prior distribution and the observed likelihood function are fused to obtain the posterior probability distribution of each spatial point, and then the fused compaction distribution map is constructed using its posterior expected value.

[0190] The logic of Bayes' theorem application: The core of Bayes' theorem is posterior probability ∝ prior probability × observed likelihood probability. For the prior and observed likelihood of a normal distribution, the mean and variance of the posterior distribution can be calculated analytically, where the mean of the distribution is the fused compaction estimate. Posterior mean = (prior variance × observed mean + observed variance × prior mean) / (prior variance + observed variance), posterior variance = (prior variance × observed variance) / (prior variance + observed variance). For example, continuing the above example, the observed likelihood function for a spatial point is N(86.5, 0.612), and the prior distribution is N(88, 0.32). Posterior mean = (0.3... 2 ×86.5+0.61 2 ×88) / (0.3 2 +0.61 2 =(0.09×86.5+0.3721×88) / (0.09+0.3721)=(7.785+32.7448) / 0.4621≈40.5298 / 0.4621≈87.71%, posterior variance =(0.3 2 ×0.61 2 ) / (0.3 2 +0.61 2 = (0.09 × 0.3721) / 0.4621 ≈ 0.033489 / 0.4621 ≈ 0.0725. Then the posterior distribution of this spatial point is N(87.71, 0.0725), and its posterior mean of 87.71% is the estimated value of the fusion compaction degree of this point.

[0191] Finally, the Bayesian fusion calculation is repeated for all spatial points in the area to be tested to obtain the fusion compaction estimate for each spatial point. Then, through visualization technology, color gradients are divided according to the compaction value range to generate a continuous fusion compaction distribution map, which clearly presents the compaction quality distribution of the entire area.

[0192] In step S500 of this embodiment, based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, a fused compaction distribution map of the area to be tested and a dynamic confidence score for each spatial point are calculated. The method further includes:

[0193] The first confidence level is obtained based on the dispersion of the estimated values ​​after depth correction from various data sources at the target spatial locations.

[0194] The construction process compliance index of the area where the target spatial point is located is used as the second confidence level;

[0195] The third confidence level is calculated based on the spatiotemporal rate of change of material moisture content in the region where the target spatial point is located.

[0196] The first confidence level, the second confidence level, and the third confidence level are fused based on preset weighting coefficients to calculate the dynamic confidence score for each spatial point.

[0197] In step S500 of this embodiment, the purpose of the above steps is to quantify the credibility of the fused compaction estimate for each spatial point from three dimensions: reliability of the test data, compliance of the construction process, and stability of the material state, thereby generating a dynamic confidence score. By comprehensively evaluating the confidence score from multiple factors, misjudgments of confidence caused by a single dimension are avoided, ensuring that subsequent quality assessments not only clarify whether the compaction is qualified but also determine the reliability of the assessment result. For areas with low confidence scores, targeted optimization instructions for testing strategies can be generated to further improve the overall rigor of the testing and its engineering application value.

[0198] To achieve the above objectives, it is first necessary to obtain the first confidence level based on the dispersion of the estimated values ​​after depth correction of each data source at the target spatial point.

[0199] That is, based on the degree of dispersion of the compaction estimate after depth correction of each data source at the target spatial location, the consistency and reliability of the test data itself are quantified. The smaller the degree of dispersion, the higher the first confidence level.

[0200] The calculation logic is as follows: First, extract the depth-corrected compaction estimate of all data sources participating in the fusion of the target spatial point, and calculate the discrete index of these estimates. The commonly used discrete index is the coefficient of variation, which is the ratio of the standard deviation to the mean. Then, establish the mapping relationship between the coefficient of variation and the first confidence level. The smaller the coefficient of variation, the higher the first confidence level. The first confidence level ranges from 0 to 1.

[0201] For example, for a target spatial point, the Rayleigh wave data depth-corrected compaction estimate is 86.2%, the vibration wheel response inversion compaction index is 85.7%, and the dielectric constant profile data is 85.5%. The calculated mean is (86.2 + 85.7 + 85.5) / 3 ≈ 85.8%, the standard deviation is ≈ 0.36%, and the coefficient of variation is 0.36% / 85.8% ≈ 0.0042. According to the preset mapping rule, when the coefficient of variation is ≤ 0.005, the first confidence level is 0.95; therefore, the first confidence level for this point is 0.95.

[0202] Next, the construction process compliance index of the area where the target spatial point is located is used as the second confidence level;

[0203] The construction process compliance index of the area where the target spatial point is located is directly converted into the second confidence level, which reflects the impact of the standardization of the construction process on the credibility of the compaction degree estimation result. The more compliant the construction process, the higher the second confidence level.

[0204] The conversion logic is as follows: the construction process compliance index ranges from 0 to 100, and it is linearly converted into a value of 0 to 1 as the second confidence level. The conversion formula is: second confidence level = construction process compliance index / 100.

[0205] For example, if the construction process compliance index of the area where the target spatial point is located is 93.4 points, the second confidence level can be calculated as 93.4 / 100 = 0.934.

[0206] Then, based on the spatiotemporal rate of change of material moisture content in the region where the target spatial point is located, the third confidence level is calculated.

[0207] That is, based on the spatiotemporal change rate of material moisture content in the region where the target spatial point is located, the degree of interference of material state stability on compaction estimation is assessed. The more gradual the change in moisture content, the higher the third confidence level.

[0208] The calculation logic is as follows: First, obtain the material moisture content data of the target spatial point at different times and locations through the material property parameter set, and calculate the spatiotemporal change rate of moisture content. The spatiotemporal change rate = (maximum moisture content - minimum moisture content) / time span. Then, establish a negative correlation mapping relationship between the change rate and the third confidence level, and set a change rate threshold. For example, when the change rate is ≤0.5% / hour, the third confidence level is 1. After exceeding the change rate threshold, the third confidence level decreases by 0.05 for every 0.1% / hour increase in the change rate, with a minimum of 0.5.

[0209] For example, in the area where the target spatial point is located, the water content data collected over 3 hours at different locations are 6.2%, 6.3%, 6.1%, and 6.4%, with a maximum water content of 6.4% and a minimum water content of 6.1%. The spatiotemporal change rate is (6.4% - 6.1%) / 3 = 0.1% / hour. This change rate is lower than the threshold of 0.5% / hour, therefore the third confidence level is 1.

[0210] Finally, the first confidence level, the second confidence level, and the third confidence level are fused based on preset weight coefficients to calculate the dynamic confidence score for each spatial point.

[0211] The first, second, and third confidence levels are weighted and fused according to preset weight coefficients to obtain the dynamic confidence score of each spatial point, which comprehensively reflects the credibility of the results under multiple dimensions.

[0212] The fusion logic is as follows: three confidence levels are preset with weight coefficients, which are calibrated based on engineering experience and historical data. For example, the weight of the first confidence level is 0.4, the weight of the second confidence level is 0.3, and the weight of the third confidence level is 0.3. The dynamic confidence score is calculated as: first confidence level × 0.4 + second confidence level × 0.3 + third confidence level × 0.3, with a value range of 0 to 1. The closer the score is to 1, the more reliable the result is.

[0213] For example, continuing the example above, the first confidence level is 0.95, the second confidence level is 0.934, and the third confidence level is 1. The dynamic confidence score is calculated as follows: 0.95 × 0.4 + 0.934 × 0.3 + 1 × 0.3 = 0.38 + 0.2802 + 0.3 = 0.9602, which means the dynamic confidence score for this spatial point is 0.96.

[0214] In step S600 of this embodiment, based on preset compaction thresholds and confidence thresholds, and combining the fused compaction distribution map with the dynamic confidence score, the compaction quality of the area to be detected is determined, and a detection strategy optimization instruction is generated for areas with confidence scores lower than the confidence threshold, including:

[0215] Pre-set the lower and upper thresholds for the acceptable compaction range, as well as the minimum confidence threshold required for the judgment result to be reliable;

[0216] Traverse the fused compaction distribution map, and for each spatial point in the map, determine whether its compaction estimate falls within the preset qualified range, and at the same time determine whether its dynamic confidence score reaches or exceeds the minimum confidence threshold.

[0217] If the estimated compaction degree of a spatial point is within the acceptable range and the dynamic confidence score meets the standard, then the compaction quality of the corresponding spatial point is deemed acceptable.

[0218] If the compaction estimate of a spatial point is within the acceptable range, but its dynamic confidence score does not meet the standard, the quality of the corresponding spatial point is determined to be unverified, and one or more detection strategy optimization instructions are automatically generated. The detection strategy optimization instructions include at least: suggesting the use of a destructive method of core drilling at the corresponding spatial point for result calibration and verification; suggesting the adjustment of the parameter settings of the non-destructive testing equipment and rescanning the corresponding area to be tested; and prompting the use of spatial statistical analysis in conjunction with the data of the surrounding qualified spatial points to assist in inference.

[0219] If the estimated compaction value of a spatial point is lower than the lower limit of the acceptable range or higher than the upper limit, the quality of the corresponding spatial point is directly determined to be unacceptable, and the corresponding spatial point is prominently marked in the visualization map output by the system. At the same time, a construction remedial instruction containing specific location information is generated.

[0220] In this embodiment, the purpose of step S600 is to accurately and hierarchically determine the compaction quality of the area to be tested in a thick water-stabilized base course based on the fusion compaction distribution map and dynamic confidence score. Simultaneously, it provides optimized testing solutions for areas with unreliable results and clarifies remedial directions for unqualified areas. This ensures the accuracy of qualified area determination while avoiding misjudgments due to insufficient confidence. Furthermore, it provides operable follow-up processing instructions for both the areas to be verified and unqualified areas, achieving a closed loop from testing data to quality conclusions and then to engineering treatment. This ensures that the testing results can directly guide engineering practice and improve the effectiveness of quality control for thick water-stabilized base courses.

[0221] To achieve the above objectives, it is first necessary to set the lower and upper thresholds of the acceptable compaction range, as well as the minimum confidence threshold required for the reliability of the judgment results.

[0222] That is, based on the engineering design requirements, construction quality standards and testing technical specifications of thick water-stabilized base courses, the lower limit threshold, upper limit threshold and minimum confidence threshold of the compaction degree qualified range are set in advance, which serve as the core basis for quality judgment.

[0223] The threshold setting logic is as follows: the acceptable compaction range must conform to industry standards and engineering design indicators. For example, the compaction requirement for thick water-stabilized base courses on highways is no less than 95%. Considering construction errors and testing deviations, the lower limit threshold is set at 95% and the upper limit threshold is set at 98% to avoid over-compaction leading to cracking of the base course. The minimum confidence threshold needs to balance testing reliability and engineering efficiency. Based on historical data verification, it is set at 0.85, meaning that the judgment result has sufficient credibility only when the dynamic confidence score is ≥0.85.

[0224] For example, the lower limit threshold for the acceptable compaction range is 95%, the upper limit threshold is 98%, and the lowest confidence threshold is 0.85.

[0225] Next, the fused compaction distribution map is traversed. For each spatial point in the map, it is determined whether its compaction estimate falls within the preset qualified range, and at the same time, it is determined whether its dynamic confidence score reaches or exceeds the minimum confidence threshold.

[0226] For each spatial point in the fusion compaction distribution map, the estimated fusion compaction value and the corresponding dynamic confidence score are extracted one by one. At the same time, two judgments are made against the preset dual thresholds: first, whether the compaction estimate is within the qualified range; and second, whether the dynamic confidence score reaches the minimum confidence threshold.

[0227] The implementation logic is as follows: the system automatically traverses all spatial points in the area to be detected, calls the fused compaction data and dynamic confidence data of each spatial point, and uses a conditional judgment algorithm to complete the double screening, providing a basis for subsequent grading judgment.

[0228] For example, the estimated compaction degree of a spatial point in the area to be tested is 96.3%, with a dynamic confidence score of 0.92; the estimated compaction degree of another spatial point is 95.5%, with a dynamic confidence score of 0.78; and yet another spatial point has an estimated compaction degree of 94.2%, with a dynamic confidence score of 0.88. Based on this, the first point's compaction degree is within the acceptable range and the confidence score meets the standard; the second point's compaction degree is acceptable but the confidence score does not meet the standard; and the third point's compaction degree is unacceptable but the confidence score meets the standard.

[0229] If the estimated compaction degree of a spatial point is within the acceptable range and the dynamic confidence score meets the standard, then the compaction quality of the corresponding spatial point is deemed acceptable.

[0230] If the compaction estimate of a spatial point is within the acceptable range, but its dynamic confidence score does not meet the standard, the quality of the corresponding spatial point is determined to be unverified, and one or more detection strategy optimization instructions are automatically generated. The detection strategy optimization instructions include at least: suggesting the use of a destructive method of core drilling at the corresponding spatial point for result calibration and verification; suggesting the adjustment of the parameter settings of the non-destructive testing equipment and rescanning the corresponding area to be tested; and prompting the use of spatial statistical analysis in conjunction with the data of the surrounding qualified spatial points to assist in inference.

[0231] If the estimated compaction value of a spatial point is lower than the lower limit of the acceptable range or higher than the upper limit, the quality of the corresponding spatial point is directly determined to be unacceptable, and the corresponding spatial point is prominently marked in the visualization map output by the system. At the same time, a construction remedial instruction containing specific location information is generated.

[0232] Based on the above dual judgment results, the compaction quality of each spatial point is divided into three levels: qualified, pending verification, and unqualified, and the judgment criteria and output format of different levels are clearly defined.

[0233] (1) Determined to be qualified

[0234] The judgment criteria are: the estimated value of spatial point fusion compaction is within the preset qualified range, that is, the lower limit threshold ≤ the estimated value of fusion compaction ≤ the upper limit threshold, and the dynamic confidence score ≥ the lowest confidence threshold.

[0235] For example, if the estimated compaction degree is 96.3%, which is within the acceptable range of 95% to 98%, and the dynamic confidence score is 0.92 ≥ 0.85, the compaction quality of this spatial point is determined to be acceptable, and the system marks the area in green on the visualization map.

[0236] 2) Determined to be verified

[0237] The judgment criteria are: the estimated compaction degree of spatial point fusion is within the acceptable range, but the dynamic confidence score is less than the minimum confidence threshold.

[0238] For example, if the estimated compaction degree is 95.5% within the acceptable range, but the dynamic confidence score is 0.78 < 0.85, the quality of this spatial point is determined to be unverifiable. The system marks this area in yellow on the visualization map and automatically generates inspection strategy optimization instructions. These instructions include suggestions to use a destructive method such as core drilling for result calibration and verification at this spatial point, suggestions to adjust the parameter settings of the non-destructive testing equipment and rescan the corresponding area to be inspected, and prompts to perform spatial statistical analysis based on data from surrounding qualified spatial points to aid in inference.

[0239] (3) Determined to be unqualified

[0240] The judgment criteria are: the estimated value of spatial point fusion compaction is less than the lower limit threshold of the qualified range, or the estimated value of spatial point fusion compaction is greater than the upper limit threshold of the qualified range.

[0241] For example, if the lower threshold of 94.2% < 95% for the estimated compaction degree, the spatial point is deemed unqualified; similarly, if the upper threshold of 98.5% > 98% for the estimated compaction degree, the spatial point is also deemed unqualified. The system prominently marks the unqualified areas in red on the visualization map and generates a construction remedial instruction containing specific location information. For example, the instruction might specify that a 5-meter radius around coordinates X118.7654 and Y32.0568 should be re-compacted, with the roller's vibration frequency maintained at 38Hz to 42Hz and the amplitude at 1.8mm to 2.2mm. A re-inspection should be performed after compaction is completed.

[0242] Example 2, as Figure 2 As shown, based on the same inventive concept as the non-destructive testing method for deep compaction of thick water-stabilized base courses provided in Embodiment 1, this embodiment of the invention also provides a non-destructive testing system for deep compaction of thick water-stabilized base courses, including:

[0243] The multi-source data acquisition module 11 is used to acquire a multi-source heterogeneous dataset of the area to be inspected using integrated testing equipment. The multi-source heterogeneous dataset includes at least a construction process parameter set, a material property parameter set, a multimodal nondestructive testing dataset, and a process image dataset.

[0244] The construction process evaluation module 12 is used to analyze and obtain the construction process compliance index and process uniformity index of the area to be inspected based on the construction process parameter set and the process image dataset.

[0245] The original data correction module 13 is used to preprocess the multimodal nondestructive testing dataset and perform depth correction on each data source in the multimodal nondestructive testing data based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source.

[0246] The multi-source data fusion module 14 is used to input the depth-corrected compaction degree estimation set of each data source into the adaptive weight fusion model to obtain a preliminary compaction degree distribution map. The adaptive weight fusion model dynamically adjusts the fusion weight of different data sources according to the depth of the detection point and the material property parameters.

[0247] The test result analysis module 15 is used to calculate the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point based on the preliminary compaction distribution map, the construction process compliance index and the process uniformity index.

[0248] The compaction quality determination module 16 is used to determine the compaction quality of the area to be tested based on a preset compaction degree threshold and confidence degree threshold, combined with the fused compaction degree distribution map and the dynamic confidence degree score, and to generate a detection strategy optimization instruction for areas with a confidence degree score lower than the confidence degree threshold.

[0249] Furthermore, the multi-source data acquisition module 11 includes the following execution steps:

[0250] An integrated detection equipment network consisting of construction parameter acquisition terminals, multi-sensor detection arrays, data communication interfaces, and high-definition camera equipment is configured at the construction site.

[0251] The construction parameter acquisition terminal collects the vibration frequency, amplitude, travel speed, number of compaction passes, and compaction trajectory coordinates obtained by the satellite positioning system in real time during the operation of the road roller, forming the construction process parameter set.

[0252] Through the data communication interface, the cement dosage, key parameters of the gradation curve, and material transportation time of each batch of mixture are obtained in real time from the production control system of the mixing plant, forming the set of material characteristic parameters;

[0253] The multi-sensor detection array synchronously collects Rayleigh wave data generated by the acceleration sensor array arranged on the surface of the water-stabilized base course, compaction index inverted by the vibration wheel response generated by the continuous compaction control system, and dielectric constant profile data obtained by multi-frequency ground-penetrating radar scanning, forming the multi-modal non-destructive testing dataset.

[0254] The high-definition camera equipment is used to collect high-definition texture images of the surface of the paved mixture and continuous video images of the roller compaction process, forming the process image dataset.

[0255] Furthermore, the construction process assessment module 12 includes the following execution steps:

[0256] The vibration frequency, amplitude, and travel speed collected in real time are compared with the pre-set process design specifications for the construction of thick water-stabilized base courses, and the vibration frequency compliance score, amplitude compliance score, and travel speed compliance score are calculated respectively.

[0257] Based on the rolling trajectory coordinates, the coverage integrity of the actual rolling trajectory over the designed rolling area is calculated, and the overlap width between adjacent rolling zones is analyzed to meet the theoretical requirements, thus obtaining the trajectory overlap rate.

[0258] Image analysis is performed on the high-definition texture image of the surface of the paved mixture to identify the uneven texture areas in the image caused by aggregate segregation, and the percentage of the area of ​​the uneven texture area to the total area of ​​the image is calculated as the segregation area percentage.

[0259] The construction process compliance index is calculated by weighting the vibration frequency compliance score, the amplitude compliance score, and the travel speed compliance score together with the trajectory overlap rate.

[0260] The value obtained by subtracting the percentage of the segregated area from 1 is defined as the process uniformity index.

[0261] Furthermore, the original data correction module 13 includes the following execution steps:

[0262] The method for constructing the effective detection depth attenuation coefficient includes: during construction of a representative test section, density sensing elements are buried at different preset depths, and the original signals of various non-destructive testing devices on the surface are collected simultaneously. A correlation model between the characteristic values ​​of various signals and the actual density of the buried points is established. The variation of the parameters of the correlation model with detection depth, material moisture content and gradation is analyzed, and the effective detection depth attenuation coefficient is obtained through function fitting.

[0263] For any detection data in the multimodal nondestructive testing dataset, based on the corresponding signal type and frequency, the corresponding effective detection depth attenuation coefficient function is called, and combined with the target detection depth of the current detection point, material moisture content and gradation parameters, the correction coefficient for converting the multimodal nondestructive testing data from the surface response value to the compaction estimate at the target depth is calculated. The depth of each data source in the multimodal nondestructive testing data is then corrected, and finally the depth-corrected compaction estimate set is obtained.

[0264] Furthermore, the multi-source data fusion module 14 includes the following execution steps:

[0265] Each nondestructive testing data source participating in the fusion is assigned a basic weight based on historical verification data, which is determined according to the average reliability level of the corresponding data source under ideal conditions;

[0266] Based on the current depth of the detection point, the basic weight of each data source is dynamically adjusted: for data sources whose physical principle sensitivity to the surface meets a preset standard, their weight is reduced as the depth increases according to a preset decay function; for data sources whose physical principle penetration capability to deep information meets a preset standard, their weight is adjusted within a specific depth range, and the adjusted weight is greater than or equal to the basic weight corresponding to the data source, wherein the specific depth range is a high-precision depth range set based on the physical principle characteristics of each data source detection method.

[0267] The weights are adjusted a second time based on the set of material property parameters: if, according to a preset standard, a certain data source shows that it is sensitive to the current material property parameter in historical data, then the fusion weight of the corresponding data source for that material property parameter is reduced accordingly.

[0268] The adjusted weights of each data source are normalized, and the depth-corrected compaction estimate of the corresponding data source is weighted and averaged to obtain the compaction estimate of each spatial point in the preliminary compaction distribution map.

[0269] Based on the estimated compaction degree, the preliminary compaction degree distribution map is generated.

[0270] Furthermore, the detection result analysis module 15 includes the following execution steps:

[0271] A Bayesian data fusion framework was adopted, and the spatial compaction estimate and its uncertainty provided by the preliminary compaction distribution map were used as the observation likelihood function.

[0272] The construction process compliance index and the process uniformity index are spatially interpolated and mapped to generate a probability map reflecting the prior quality distribution of construction, which serves as the prior distribution.

[0273] The prior distribution and the observed likelihood function are fused using Bayes' theorem to calculate the posterior probability distribution. The expected value map of this posterior probability distribution is then used as the fused compaction distribution map.

[0274] The first confidence level is obtained based on the dispersion of the estimated values ​​after depth correction from various data sources at the target spatial points.

[0275] The construction process compliance index of the area where the target spatial point is located is used as the second confidence level;

[0276] The third confidence level is calculated based on the spatiotemporal rate of change of material moisture content in the region where the target spatial point is located.

[0277] The first confidence level, the second confidence level, and the third confidence level are fused based on preset weighting coefficients to calculate the dynamic confidence score for each spatial point.

[0278] Furthermore, the compaction quality judgment module 16 includes the following execution steps:

[0279] Pre-set the lower and upper thresholds for the acceptable compaction range, as well as the minimum confidence threshold required for the judgment result to be reliable;

[0280] Traverse the fused compaction distribution map, and for each spatial point in the map, determine whether its compaction estimate falls within the preset qualified range, and at the same time determine whether its dynamic confidence score reaches or exceeds the minimum confidence threshold.

[0281] If the estimated compaction degree of a spatial point is within the acceptable range and the dynamic confidence score meets the standard, then the compaction quality of the corresponding spatial point is deemed acceptable.

[0282] If the compaction estimate of a spatial point is within the acceptable range, but its dynamic confidence score does not meet the standard, the quality of the corresponding spatial point is determined to be unverified, and one or more detection strategy optimization instructions are automatically generated. The detection strategy optimization instructions include at least: suggesting the use of a destructive method of core drilling at the corresponding spatial point for result calibration and verification; suggesting the adjustment of the parameter settings of the non-destructive testing equipment and rescanning the corresponding area to be tested; and prompting the use of spatial statistical analysis in conjunction with the data of the surrounding qualified spatial points to assist in inference.

[0283] If the estimated compaction value of a spatial point is lower than the lower limit of the acceptable range or higher than the upper limit, the quality of the corresponding spatial point is directly determined to be unacceptable, and the corresponding spatial point is prominently marked in the visualization map output by the system. At the same time, a construction remedial instruction containing specific location information is generated.

Claims

1. Non-destructive testing methods for deep compaction of thick water-stabilized base courses, including: The multi-source heterogeneous dataset of the area to be inspected is obtained using integrated inspection equipment. The multi-source heterogeneous dataset includes at least a construction process parameter set, a material property parameter set, a multimodal nondestructive testing dataset, and a process image dataset. Based on the construction process parameter set and the process image dataset, the construction process compliance index and process uniformity index of the area to be detected are obtained through analysis. The multimodal nondestructive testing dataset is preprocessed, and the depth of each data source in the multimodal nondestructive testing dataset is corrected based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source. The depth-corrected compaction degree estimation sets from each data source are input into the adaptive weight fusion model to obtain a preliminary compaction degree distribution map. The adaptive weight fusion model dynamically adjusts the fusion weights of different data sources based on the depth of the detection points and material property parameters. Based on the preliminary compaction distribution map, the construction process compliance index, and the process uniformity index, the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point are calculated. Based on preset compaction thresholds and confidence thresholds, and combining the fused compaction distribution map with the dynamic confidence score, the compaction quality of the area to be detected is determined, and detection strategy optimization instructions are generated for areas with confidence scores lower than the confidence threshold. Use integrated detection equipment to acquire multi-source heterogeneous datasets of the region to be detected, including: An integrated detection equipment network consisting of construction parameter acquisition terminals, multi-sensor detection arrays, data communication interfaces, and high-definition camera equipment is configured at the construction site. The construction parameter acquisition terminal collects the vibration frequency, amplitude, travel speed, number of compaction passes, and compaction trajectory coordinates obtained by the satellite positioning system in real time during the operation of the road roller, forming the construction process parameter set. Through the data communication interface, the cement dosage, key parameters of the gradation curve, and material transportation time of each batch of mixture are obtained in real time from the production control system of the mixing plant, forming the set of material characteristic parameters; The multi-sensor detection array synchronously collects Rayleigh wave data generated by the acceleration sensor array arranged on the surface of the water-stabilized base course, compaction index inverted by the vibration wheel response generated by the continuous compaction control system, and dielectric constant profile data obtained by multi-frequency ground-penetrating radar scanning, forming the multi-modal non-destructive testing dataset. The high-definition camera equipment is used to collect high-definition texture images of the surface of the paved mixture and continuous video images of the roller compaction process to form the process image dataset. The depth-corrected compaction degree estimates from each data source are input into the adaptive weighted fusion model to obtain a preliminary compaction degree distribution map, which includes: Each nondestructive testing data source participating in the fusion is assigned a basic weight based on historical verification data, which is determined according to the average reliability level of the corresponding data source under ideal conditions; Based on the current depth of the detection point, the basic weight of each data source is dynamically adjusted: for data sources whose physical principle sensitivity to the surface meets a preset standard, their weight is reduced as the depth increases according to a preset decay function; for data sources whose physical principle penetration capability to deep information meets a preset standard, their weight is adjusted within a specific depth range, and the adjusted weight is greater than or equal to the basic weight corresponding to the data source, wherein the specific depth range is a high-precision depth range set based on the physical principle characteristics of each data source detection method. The weights are adjusted a second time based on the set of material property parameters: if, according to a preset standard, a certain data source shows that it is sensitive to the current material property parameter in historical data, then the fusion weight of the corresponding data source for that material property parameter is reduced accordingly. The adjusted weights of each data source are normalized, and the depth-corrected compaction estimate of the corresponding data source is weighted and averaged to obtain the compaction estimate of each spatial point in the preliminary compaction distribution map. Based on the estimated compaction degree, the preliminary compaction degree distribution map is generated.

2. The non-destructive testing method for deep compaction degree of thick water-stabilized base courses according to claim 1, characterized in that, Based on the construction process parameter set and the process image dataset, the construction process compliance index and process uniformity index of the area to be inspected are analyzed and obtained, including: The vibration frequency, amplitude, and travel speed collected in real time are compared with the pre-set process design specifications for the construction of thick water-stabilized base courses, and the vibration frequency compliance score, amplitude compliance score, and travel speed compliance score are calculated respectively. Based on the rolling trajectory coordinates, the coverage integrity of the actual rolling trajectory over the designed rolling area is calculated, and the overlap width between adjacent rolling zones is analyzed to meet the theoretical requirements, thus obtaining the trajectory overlap rate. Image analysis is performed on the high-definition texture image of the surface of the paved mixture to identify the uneven texture areas in the image caused by aggregate segregation, and the percentage of the area of ​​the uneven texture area to the total area of ​​the image is calculated as the segregation area percentage. The construction process compliance index is calculated by weighting the vibration frequency compliance score, the amplitude compliance score, and the travel speed compliance score together with the trajectory overlap rate. The value obtained by subtracting the percentage of the segregated area from 1 is defined as the process uniformity index.

3. The non-destructive testing method for deep compaction degree of thick water-stabilized base courses according to claim 1, characterized in that, Based on a preset effective detection depth attenuation coefficient, depth correction is performed on each data source in the multimodal nondestructive testing data to obtain a depth-corrected compaction estimate set for each data source, including: The method for constructing the effective detection depth attenuation coefficient includes: during construction of a representative test section, density sensing elements are buried at different preset depths, and the original signals of various non-destructive testing devices on the surface are collected simultaneously. A correlation model between the characteristic values ​​of various signals and the actual density of the buried points is established. The variation of the parameters of the correlation model with detection depth, material moisture content and gradation is analyzed, and the effective detection depth attenuation coefficient is obtained through function fitting.

4. The non-destructive testing method for deep compaction degree of thick water-stabilized base courses according to claim 1, characterized in that, Based on a preset effective detection depth attenuation coefficient, depth correction is performed on each data source in the multimodal nondestructive testing data to obtain a depth-corrected compaction estimate set for each data source, and the method further includes: For any detection data in the multimodal nondestructive testing dataset, based on the corresponding signal type and frequency, the corresponding effective detection depth attenuation coefficient function is called, and combined with the target detection depth of the current detection point, material moisture content and gradation parameters, the correction coefficient for converting the multimodal nondestructive testing data from the surface response value to the compaction estimate at the target depth is calculated. The depth of each data source in the multimodal nondestructive testing data is then corrected, and finally the depth-corrected compaction estimate set is obtained.

5. The non-destructive testing method for deep compaction degree of thick water-stabilized base courses according to claim 1, characterized in that, Based on the preliminary compaction degree distribution map, the construction process compliance index, and the process uniformity index, a fused compaction degree distribution map of the area to be tested and a dynamic confidence score for each spatial point are calculated, including: A Bayesian data fusion framework was adopted, and the spatial compaction degree estimates and their uncertainties provided by the preliminary compaction degree distribution map were used as observation likelihood functions. The construction process compliance index and the process uniformity index are spatially interpolated and mapped to generate a probability map reflecting the prior quality distribution of construction, which serves as the prior distribution. The prior distribution and the observed likelihood function are fused using Bayes' theorem to calculate the posterior probability distribution. The expected value map of this posterior probability distribution is then used as the fused compaction distribution map.

6. The non-destructive testing method for deep compaction degree of thick water-stabilized base courses according to claim 1, characterized in that, Based on the preliminary compaction degree distribution map, the construction process compliance index, and the process uniformity index, a fused compaction degree distribution map of the area to be tested and a dynamic confidence score for each spatial point are calculated, including: The first confidence level is obtained based on the dispersion of the estimated values ​​after depth correction from various data sources at the target spatial locations. The construction process compliance index of the area where the target spatial point is located is used as the second confidence level; The third confidence level is calculated based on the spatiotemporal rate of change of material moisture content in the region where the target spatial point is located. The first confidence level, the second confidence level, and the third confidence level are fused based on preset weighting coefficients to calculate the dynamic confidence score for each spatial point.

7. The non-destructive testing method for deep compaction degree of thick water-stabilized base courses according to claim 1, characterized in that, Based on preset compaction degree thresholds and confidence thresholds, and combining the fused compaction degree distribution map with the dynamic confidence score, the compaction quality of the area to be inspected is determined, and inspection strategy optimization instructions are generated for areas with confidence scores lower than the confidence threshold, including: Pre-set the lower and upper thresholds for the acceptable compaction range, as well as the minimum confidence threshold required for the judgment result to be reliable; Traverse the fused compaction distribution map, and for each spatial point in the map, determine whether its compaction estimate falls within the preset qualified range, and at the same time determine whether its dynamic confidence score reaches or exceeds the minimum confidence threshold. If the estimated compaction degree of a spatial point is within the acceptable range and the dynamic confidence score meets the standard, then the compaction quality of the corresponding spatial point is deemed acceptable. If the compaction estimate of a spatial point is within the acceptable range, but its dynamic confidence score does not meet the standard, the quality of the corresponding spatial point is determined to be unverified, and one or more detection strategy optimization instructions are automatically generated. The detection strategy optimization instructions include at least: suggesting the use of a destructive method of core drilling at the corresponding spatial point for result calibration and verification; suggesting the adjustment of the parameter settings of the non-destructive testing equipment and rescanning the corresponding area to be tested; and prompting the use of spatial statistical analysis in conjunction with the data of the surrounding qualified spatial points to assist in inference. If the estimated compaction value of a spatial point is lower than the lower limit of the acceptable range or higher than the upper limit, the quality of the corresponding spatial point is directly determined to be unacceptable, and the corresponding spatial point is prominently marked in the visualization map output by the system. At the same time, a construction remedial instruction containing specific location information is generated.

8. A non-destructive testing system for deep compaction of thick water-stabilized base courses, characterized in that the system is used to implement the non-destructive testing method for deep compaction of thick water-stabilized base courses as described in any one of claims 1-7, the system comprising: The multi-source data acquisition module is used to acquire a multi-source heterogeneous dataset of the area to be inspected using integrated testing equipment. The multi-source heterogeneous dataset includes at least a construction process parameter set, a material property parameter set, a multimodal nondestructive testing dataset, and a process image dataset. The construction process evaluation module is used to analyze and obtain the construction process compliance index and process uniformity index of the area to be inspected based on the construction process parameter set and the process image dataset. The original data correction module is used to preprocess the multimodal nondestructive testing dataset and perform depth correction on each data source in the multimodal nondestructive testing data based on a preset effective detection depth attenuation coefficient to obtain a depth-corrected compaction estimate set for each data source. The multi-source data fusion module is used to input the depth-corrected compaction degree estimation set of each data source into the adaptive weight fusion model to obtain a preliminary compaction degree distribution map. The adaptive weight fusion model dynamically adjusts the fusion weights of different data sources according to the depth of the detection point and the material property parameters. The test result analysis module is used to calculate the fused compaction distribution map of the area to be tested and the dynamic confidence score of each spatial point based on the preliminary compaction distribution map, the construction process compliance index and the process uniformity index. The compaction quality determination module is used to determine the compaction quality of the area to be tested based on a preset compaction degree threshold and confidence degree threshold, combined with the fused compaction degree distribution map and the dynamic confidence degree score, and to generate detection strategy optimization instructions for areas with confidence degree scores lower than the confidence degree threshold.

Images