Maintenance decision method based on comprehensive conditions of internal and external damage of bridge deck asphalt pavement

Abstract

The application discloses a kind of maintenance decision-making methods based on the comprehensive condition of inside and outside damage of bridge deck asphalt pavement, comprising: step S1, cooperates UAV with GPR to complete bridge deck asphalt pavement multi-source data acquisition and pre-processing;Step S2, based on pre-processing data, carry out bridge three-dimensional reconstruction, quantize external apparent damage and calculate RQOI;Step S3, complete internal structure damage quantization by GPR signal analysis, fuse PCI, RQOI, RDI, TCI, bridgehead bump value index;Step S4, realize multi-source information nonlinear fusion and dimension reduction using KPCA, based on comprehensive evaluation score, complete bridge deck asphalt pavement maintenance priority ranking.The application realizes internal and external damage global quantization, multi-source information deep fusion and maintenance priority accurate sorting, improves the scientificity and economy of bridge deck pavement maintenance decision-making.

Description

Technical Field

[0001] This invention belongs to the field of bridge deck asphalt pavement maintenance assessment technology, specifically a maintenance decision-making method based on the comprehensive internal and external damage status of bridge deck asphalt pavement. Background Technology

[0002] Bridge asphalt pavement layers are subjected to multiple coupled effects over long periods, including vehicle loads, environmental temperature changes, and rainwater erosion. Their service condition directly impacts the bridge's driving safety, comfort, and durability. Current bridge deck maintenance decisions largely rely on conventional surface performance indicators such as the Pavement Condition Index (PCI), Ride Quality Index (RQI), and Rutting Depth Index (RDI). While these indicators can reflect local surface performance, they fail to accurately reveal the overall service condition of the bridge asphalt pavement layer, especially regarding multi-dimensional damage information such as macroscopic deformation, localized bumps, and abnormal internal structural thickness.

[0003] Existing multi-functional inspection vehicles typically conduct inspections along fixed wheel tracks, with results mainly reflecting the local conditions of a single or a few longitudinal sections. Their lateral coverage is limited, making it difficult to characterize the distribution of bridge deck smoothness and macroscopic deformation features at the full cross-sectional scale. While UAVs can quickly acquire large-scale bridge deck images and point cloud data, limitations in platform load and sensor capabilities mean they typically only reflect the surface condition and cannot reveal internal structural anomalies within the pavement layer. Ground-penetrating radar, although capable of non-destructive internal inspection, primarily focuses on thickness or interface information. Without collaborative analysis with bridge deck surface damage information, it cannot directly support maintenance prioritization decisions.

[0004] Furthermore, bridge deck smoothness testing typically relies on data from a single longitudinal section or a single wheel track, which cannot fully reflect the smoothness differences between different locations on the bridge deck in the transverse direction. Traditional methods cannot represent the entire cross-section state with a single longitudinal section. Therefore, a full-section smoothness index is constructed based on multiple transverse survey lines, and fuzzy logic is further used to establish the RQOI to comprehensively characterize the severity and scope of impact. At the same time, to address the issues of inconsistent data dimensions and correlations and redundancies between indices from different testing methods, kernel principal component analysis is introduced to calculate the comprehensive score and determine maintenance priorities.

[0005] Therefore, it is necessary to provide a method that can integrate surface inspection data, internal inspection data and routine road condition inspection data of bridge deck asphalt pavement, and realize comprehensive service status evaluation and maintenance priority determination of bridge deck asphalt pavement, so as to make up for the shortcomings of single index, single detection method and local evaluation method in the existing technology. Summary of the Invention

[0006] To address the aforementioned problems in existing technologies, this invention provides a maintenance decision-making method based on the comprehensive internal and external damage status of bridge deck asphalt pavement. This method solves the problems of traditional assessments relying solely on a single external indicator, ignoring internal hidden dangers, and having insufficient ranking accuracy. It achieves full-domain quantification of internal and external damage, deep fusion of multi-source information, and accurate ranking of maintenance priorities, thereby improving the scientific and economical nature of bridge deck pavement maintenance decisions.

[0007] The technical solution to achieve the above objectives is: A maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement includes: Step S1: Use UAV aerial photography and GPR to collect external appearance data and internal structure data of bridge deck asphalt pavement layer. Perform occlusion removal, noise removal and signal correction on the collected data to obtain standardized multi-source detection data, including standardized external appearance data and standardized internal structure data. Step S2: Construct a full-domain 3D model of the bridge deck based on standardized external appearance data, identify macroscopic deformation and bridge approach slab settlement, extract full-section smoothness data and calculate RQOI (Road Quality Oscillation Index) based on fuzzy logic to complete the full-domain quantification of external damage; Step S3: Based on the standardized internal structure data, analyze the pavement layer thickness distribution characteristics, calculate TCI, and integrate multiple source indicators such as PCI, RQOI, RDI, TCI, and bridge approach slab settlement value to complete the quantitative characterization of internal structure damage. Step S4: Standardize the multi-source indicators of internal and external damage, use KPCA (nonlinear dimensionality reduction algorithm) to perform nonlinear data fusion and feature dimensionality reduction, and complete the priority assessment and ranking of bridge deck asphalt pavement layer maintenance based on the comprehensive evaluation score.

[0008] Preferably, step S1 includes: Step S11: Use a UAV (unmanned aerial vehicle) equipped with a full-frame camera to conduct a full-area aerial photography of the bridge surface according to preset parameters to obtain high-resolution visible light images and three-dimensional point cloud data; Step S12: Use 1600MHz GPR (Ground Ground Radar) to scan the internal structure of the bridge deck lanes, obtain the pavement layer structure thickness reflection signal, and simultaneously collect the PCI, RQI, and RDI conventional index data of the road multi-functional inspection vehicle. Step S13: Use the YOLO-VEHICLE (a dedicated vehicle detection model optimized based on the YOLO framework) model to remove vehicles and shadow occlusions in the image, accurately identify driving vehicles and shadows in the UAV aerial image, and remove occlusion noise through whitening mask processing. Step S14: The ROR (Order Robust) denoising algorithm is used to filter the 3D point cloud data to remove redundant noise and outliers. Step S15: Perform time-domain correction, waveform smoothing, and noise suppression on the original GPR signal to eliminate environmental electromagnetic interference; Step S16: Finally, the processed appearance data, internal structure data, and conventional indicator data are unified in a single format to obtain a standardized multi-source detection dataset.

[0009] Preferably, step S2 includes: Step S21: Based on multi-view images, the SfM (Structure of Motion) algorithm is used to complete the reconstruction of the three-dimensional model of the entire bridge deck, and a relative digital elevation model is constructed to identify macroscopic uplift and subsidence deformation; Step S22: Based on the three-dimensional point cloud, locate the lane lines and expansion joints, extract longitudinal profile data and calculate the bridge approach slab settlement value, divide the bridge surface into equal parts laterally, and extract multi-section flatness data. Step S23: Using the severity and impact range of the flatness defect as input, the RQOI is calculated using the Mandani fuzzy inference system to complete the full-domain quantification of external damage.

[0010] Preferably, step S21 includes: Step S211: Based on the preprocessed multi-view UAV images, the SfM algorithm is used to perform feature matching and three-dimensional coordinate calculation to generate dense point cloud data. Step S212: Construct a high-precision 3D digital model of the entire bridge surface through point cloud optimization, mesh reconstruction, and texture mapping; Step S213: A normal bridge deck elevation model is constructed using a 2.5D polygon surface fitting method. The actual elevation and the fitted elevation are subtracted point by point to generate a relative digital elevation model of the bridge deck. Macroscopic uplift and large-scale deformation areas of subsidence are visually identified through a color distribution map.

[0011] Preferably, step S22 includes: Step S221: Based on the 3D point cloud, a generative adversarial network is used to perform pixel-level segmentation and localization of lane lines, and point cloud data of the longitudinal profiles on the left and right sides of the expansion joint are extracted with the expansion joint as the center. Step S222: Principal component analysis is used to determine the driving direction, the center of the expansion joint is located by peak detection method, and the elevation difference between the two sides is calculated to obtain the bridge approach slab value; Step S223: Divide the single lane into 30 survey line positions along the transverse direction of the bridge deck, extract the longitudinal profile elevation sequence at each position, calculate the international smoothness index of each survey line using a quarter-car model, and convert it into RQI (Ride Quality Index) value.

[0012] Preferably, step S23 includes: Step S231: Based on the Mandani fuzzy inference system, the membership relationship between input variables and output variables is established using triangular membership function and trapezoidal membership function, wherein the severity and range of RQI defects are used as input variables; The membership function formula for a triangle is as follows: ; In the formula, Describes the membership function of a triangle. and Let represent the lower and upper bounds of the membership function of a triangle, respectively. This represents a membership degree measurement value that is equal to 1. The formula for the trapezoidal membership function is as follows: ; In the formula, Represents the membership function of a trapezoid. and Let represent the lower and upper bounds of the trapezoidal membership function, respectively. and This represents the values ​​between 1 and 1 for the trapezoidal membership degree measurement. Step S232: The severity of RQI defects is divided into three levels: acceptable, problematic, and critical. The defect range is divided into three levels: local, moderate, and widespread. RQOI is calculated by defuzzification using the centroid method. The formula for calculating RQOI is as follows: ; In the formula, This indicates the precise output of RQOI. The aggregate membership function is calculated using the maximum aggregation method. A variable representing a possible output value; Step S233 comprehensively characterizes the driving quality and external damage degree of the entire cross section of the bridge deck asphalt pavement, and completes the full-area quantification of external damage.

[0013] Preferably, step S3 includes: Step S31: Analyze the GPR reflection signal, identify the reflection waveforms at the upper and lower interfaces of the pavement layer, calculate the actual thickness of the pavement layer at each measuring point based on the electromagnetic wave two-way travel time, and statistically analyze the mean, maximum, minimum, and thickness deviation rates of each bridge deck section to establish a thickness distribution database. Step S32: Calculate TCI based on the thickness distribution characteristics and deviation rate of the pavement layer. The TCI value is positively correlated with the thickness uniformity. The smaller the thickness deviation, the higher the TCI value, which indicates better internal structural integrity. Step S33: Integrate multiple source indicators such as PCI, RQOI, RDI, TCI, and bridge approach slab settlement value to construct a quantitative indicator set for internal structural damage. This set comprehensively reflects the abnormal thickness and hidden damage status of the internal structure of the asphalt pavement layer on the bridge deck, thus completing the quantitative characterization of the internal structural damage of the asphalt pavement layer on the bridge deck.

[0014] Preferably, step S4 includes: Step S41: The Z-Score method is used to normalize the five indicators, PCI, RQOI, RDI, TCI, and bridge approach slab slab value, to eliminate differences in dimensions and values. The formula is as follows: ; In the formula, Indicates the first The first evaluation object The value of each indicator, Indicates the number of samples. , Indicates the dimension of the indicator. ; After standardization, the standardized matrix is ​​obtained. ,Right now: ; ; ; In the formula, This represents the standardized indicator variable. Indicates the first The sample mean of each indicator, Indicates the first The variance of each indicator; Step S42: Select the Gaussian radial basis kernel function to construct the kernel matrix and complete the centering process. Perform eigenvalue decomposition on the kernel matrix and select the core principal components according to the cumulative variance contribution rate. Step S43: Calculate the principal component scores and comprehensive evaluation index of each bridge deck section, sort them from low to high, and determine the maintenance priority of the bridge deck asphalt pavement layer.

[0015] Preferably, in step S42, the process of constructing the kernel matrix using the Gaussian radial basis kernel function and completing the centering process is as follows: Build kernel matrix : ; In the formula, and They represent the first The and the first A standardized road surface sample vector, This represents a kernel function used to compute two sample vectors. and The similarity between them Represents an exponential function. The parameter representing the width of the Gaussian kernel; Kernel matrix By performing centralization processing, a centralized kernel matrix is ​​obtained. : ; In the formula, Represent a A matrix whose elements are all equal to 0. ; Kernel matrix Perform eigenvalue decomposition and solve for the eigenvalues. and the corresponding feature vector : ; Find 1 non-negative eigenvalue, arranged in descending order of numerical value: ; For the selected feature vectors Normalization is performed so that the magnitude of its projection vector in the feature space is 1, that is: ; In the formula, Indicates the transpose symbol. Representing the kernel matrix The One eigenvalue; Core principal components were selected based on a cumulative variance contribution rate of ≥90% to eliminate correlations and data redundancy among indicators and extract core features that can represent the comprehensive damage state of the asphalt pavement layer on the bridge deck.

[0016] Preferably, in step S44, the score of each bridge deck segment on the core principal component is calculated, and the comprehensive evaluation index is calculated by weighting the principal component variance contribution rate. The process is as follows: If before One nonlinear principal component The amount of information is respectively For any road surface sample In its first Scores on each principal component Calculated by the formula, that is: ; In the formula, Indicates the first The th normalized eigenvector One portion, This represents the computation of the centered kernel function. Indicates training samples; Information contribution rate of each principal component The evaluation calculation formula is as follows: ; In the formula, Indicates the first Principal Components The corresponding eigenvalues, Indicates the first Eigenvalues ​​of each principal component; A comprehensive evaluation function for calculating the performance of bridge deck asphalt pavement layers. ,Right now: ; in, The higher the value, the better the overall performance of the asphalt pavement, and vice versa; Sorted from low to high according to the comprehensive evaluation index, the lower the index, the more serious the comprehensive damage to the asphalt pavement layer of the bridge deck, and the higher the maintenance priority. Finally, a maintenance priority list for the bridge deck asphalt pavement layer is generated, enabling scientific maintenance decisions.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves comprehensive data collection of both internal and external aspects of bridge deck asphalt pavement through collaborative detection using UAV and GPR, overcoming the shortcomings of traditional detection methods such as insufficient coverage and missed detection of internal hazards. It employs fuzzy logic to construct a full-section RQOI, comprehensively considering the severity and impact range of defects to fully characterize external damage. KPCA enables nonlinear fusion of multi-source indicators, eliminating data redundancy and accurately identifying hidden defects. Based on the fusion results, maintenance priorities are ranked, correcting the biases of traditional single-indicator assessments, and providing core technical support for precise maintenance and optimized allocation of funds for bridge deck asphalt pavement. Attached Figure Description

[0018] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement layers according to the present invention; Figure 2 This is a flowchart of the multi-source data acquisition and preprocessing process in this invention; Figure 3 This is a flowchart of the global quantization of external damage in this invention; Figure 4 This is a flowchart of the process for constructing a relative digital elevation model to identify macroscopic uplift and subsidence deformation in this invention; Figure 5 This is a flowchart of the process for calculating bridge approach slab settlement and extracting multi-section flatness data in this invention; Figure 6 This is a flowchart of the calculation of RQOI using the Mandani fuzzy inference system in this invention; Figure 7 This is a flowchart of the quantitative characterization of internal structural damage in this invention; Figure 8 This is a flowchart of the process for evaluating and ranking the maintenance priority of bridge deck asphalt pavement based on comprehensive evaluation scores in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown, a maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement includes: Step S1: Use UAV aerial photography and GPR to collect external appearance data and internal structure data of bridge deck asphalt pavement layer. Perform occlusion removal, noise removal and signal correction on the collected data to obtain standardized multi-source detection data, including standardized external appearance data and standardized internal structure data.

[0021] like Figure 2 As shown, step S1 includes: Step S11: Using a UAV equipped with a full-frame camera, conduct full-area aerial photography of the bridge surface according to preset parameters to acquire high-resolution visible light images and 3D point cloud data. Specifically, a DJI Matrice M300 RTK UAV paired with a Zenmuse P1 full-frame visible light camera is used to form an aerial photography system, with flight path planning based on DJI's professional aerial surveying software. Fixed aerial photography parameters: flight altitude 20m, flight speed 5m / s, camera shooting angle -90°, forward overlap 75%, and lateral overlap 60%. Continuous aerial photography is conducted along the traffic direction of the bridge surface, automatically acquiring high-resolution visible light images and raw 3D point cloud data containing RTK positioning information. A single flight can complete 52,000m. 2 Bridge surface data collection ensures comprehensive coverage of the entire bridge surface without any blind spots.

[0022] Step S12: An internal scan of the bridge deck driveway is performed using a 1600MHz GPR to obtain the reflection signal of the pavement layer structure thickness. Simultaneously, PCI, RQI, and RDI conventional index data from the road surface multi-functional inspection vehicle are collected. Specifically, a GER1600A type 1600MHz GPR is used to perform non-destructive scanning of the bridge deck driveway in a ground coupling manner. The time window is set to 10ns, and the detection depth is 0.2-0.5m to obtain the electromagnetic wave signal reflected from the pavement layer structure interface. Simultaneously, conventional maintenance indicators of the bridge deck asphalt pavement layer, including PCI, RQI, and RDI, are collected through the road surface multi-functional inspection vehicle, and the original values ​​of each indicator and the detection point information are recorded.

[0023] Step S13: Use the YOLO-VEHICLE model to remove vehicles and shadows from the image, accurately identify vehicles and shadows in the UAV aerial image, and remove occlusion noise through whitening mask processing.

[0024] Step S14: The ROR denoising algorithm is used to filter the 3D point cloud data to remove redundant noise and outliers.

[0025] Step S15: Perform time-domain correction, waveform smoothing and noise suppression on the original GPR signal to eliminate environmental electromagnetic interference.

[0026] Step S16: Finally, the processed appearance data, internal structure data, and conventional indicator data are unified in a single format to obtain a standardized multi-source detection dataset.

[0027] Step S2: Construct a full-domain 3D model of the bridge deck based on standardized external appearance data, identify macroscopic deformation and bridge approach slab settlement, extract full-section flatness data and calculate RQOI based on fuzzy logic to complete the full-domain quantification of external damage.

[0028] like Figure 3 As shown, step S2 includes: Step S21: Based on multi-view images, the SfM algorithm is used to complete the reconstruction of the three-dimensional model of the entire bridge deck and construct a relative digital elevation model to identify macroscopic uplift and subsidence deformation.

[0029] like Figure 4 As shown, step S21 includes: Step S211: Based on the preprocessed multi-view UAV images, the SfM algorithm is used to perform feature matching and three-dimensional coordinate calculation to generate dense point cloud data. Step S212: Construct a high-precision 3D digital model of the entire bridge surface through point cloud optimization, mesh reconstruction, and texture mapping; Step S213: A normal bridge deck elevation model is constructed using a 2.5D polygon surface fitting method. The actual elevation and the fitted elevation are subtracted point by point to generate a relative digital elevation model of the bridge deck. Macroscopic uplift and large-scale deformation areas of subsidence are visually identified through a color distribution map.

[0030] Step S22: Based on the three-dimensional point cloud, locate the lane lines and expansion joints, extract longitudinal profile data and calculate the bridge approach slab settlement value, divide the bridge surface laterally into equal parts, and extract multi-section flatness data.

[0031] like Figure 5 As shown, step S22 includes: Step S221: Based on the 3D point cloud, a generative adversarial network is used to perform pixel-level segmentation and localization of lane lines, and point cloud data of the longitudinal profiles on the left and right sides of the expansion joint are extracted with the expansion joint as the center. Step S222: Principal component analysis is used to determine the driving direction, the center of the expansion joint is located by peak detection method, and the elevation difference between the two sides is calculated to obtain the bridge approach slab value. Step S223: Divide the single lane into 30 survey line positions along the transverse direction of the bridge deck, extract the longitudinal profile elevation sequence at each position, calculate the international smoothness index of each survey line using a quarter-car model, and convert it into an RQI value.

[0032] Step S23: Using the severity and impact range of the flatness defect as input, the RQOI is calculated using the Mandani fuzzy inference system to complete the full-domain quantification of external damage.

[0033] like Figure 6 As shown, step S23 includes: Step S231: Based on the Mandani fuzzy inference system, the membership relationship between input variables and output variables is established using triangular membership function and trapezoidal membership function, wherein the severity and range of RQI defects are used as input variables; The membership function formula for a triangle is as follows: ; In the formula, Describes the membership function of a triangle. and Let represent the lower and upper bounds of the membership function of a triangle, respectively. The triangular membership function represents a membership measure value equal to 1. It is often used in fuzzy logic because it is simple and easy to interpret, providing a linear increase and decrease in membership values ​​within the boundary. The formula for the trapezoidal membership function is as follows: ; In the formula, Represents the membership function of a trapezoid. and Let represent the lower and upper bounds of the trapezoidal membership function, respectively. and The trapezoidal membership function represents values ​​between 1 and 1. It offers more flexibility than triangular functions, allowing for flat regions with membership values ​​of 1, which is particularly effective when a range of values ​​can be considered equally valid. Step S232: The severity of RQI defects is divided into three levels: acceptable, problematic, and critical. The defect range is divided into three levels: local, moderate, and widespread. RQOI is calculated by defuzzification using the centroid method. The formula for calculating RQOI is as follows: ; In the formula, This indicates the precise output of RQOI. The aggregate membership function is calculated using the maximum aggregation method. The variable representing the possible output values ​​corresponds to the RQOI value in this invention; Step S233 comprehensively characterizes the driving quality and external damage degree of the entire cross section of the bridge deck asphalt pavement, and completes the full-area quantification of external damage.

[0034] Step S3: Based on the standardized internal structure data, analyze the pavement layer thickness distribution characteristics, calculate TCI, and integrate multiple source indicators such as PCI, RQOI, RDI, TCI, and bridge approach slab settlement value to complete the quantitative characterization of internal structure damage.

[0035] like Figure 7 As shown, step S3 includes: Step S31: Analyze the GPR reflection signal, identify the reflection waveforms at the upper and lower interfaces of the pavement layer, calculate the actual thickness of the pavement layer at each measuring point based on the electromagnetic wave two-way travel time, and statistically analyze the mean, maximum, minimum, and thickness deviation rates of each bridge deck section to establish a thickness distribution database.

[0036] Step S32: Calculate TCI based on the thickness distribution characteristics and deviation rate of the pavement layer. The TCI value is positively correlated with the thickness uniformity. The smaller the thickness deviation, the higher the TCI value, and the better the internal structural integrity.

[0037] Step S33: Integrate multiple source indicators such as PCI, RQOI, RDI, TCI, and bridge approach slab settlement value to construct a quantitative indicator set for internal structural damage. This set comprehensively reflects the hidden damage status of the bridge deck asphalt pavement layer, such as abnormal thickness and structural deterioration, and completes the quantitative characterization of the internal structural damage of the bridge deck asphalt pavement layer.

[0038] Step S4: Standardize the multi-source indicators of internal and external damage, use the KPCA algorithm for nonlinear data fusion and feature dimensionality reduction, and complete the priority assessment and ranking of bridge deck asphalt pavement layer maintenance based on the comprehensive evaluation score.

[0039] like Figure 8 As shown, step S4 includes: Step S41: The Z-Score method is used to normalize the five indicators, PCI, RQOI, RDI, TCI, and bridge approach slab settlement, to eliminate differences in dimensions and values.

[0040] The formula is as follows: ; In the formula, Indicates the first The first evaluation object The value of each indicator, Indicates the number of samples. , Indicates the dimension of the indicator. ; After standardization, the standardized matrix is ​​obtained. ,Right now: ; ; ; In the formula, This represents the standardized indicator variable. Indicates the first The sample mean of each indicator, Indicates the first The variance of each indicator; Step S42: Select the Gaussian radial basis kernel function to construct the kernel matrix and complete the centering process. Perform eigenvalue decomposition on the kernel matrix and select the core principal components according to the cumulative variance contribution rate.

[0041] In this embodiment, the Gaussian radial basis kernel function is used to construct the kernel matrix and perform the centering process as follows: Build kernel matrix : ; In the formula, and They represent the first The and the first A standardized road surface sample vector, This represents a kernel function used to compute two sample vectors. and The similarity between them Represents an exponential function. The parameter representing the width of the Gaussian kernel; Kernel matrix By performing centralization processing, a centralized kernel matrix is ​​obtained. : ; In the formula, Represent a A matrix whose elements are all equal to 0. ; Kernel matrix Perform eigenvalue decomposition and solve for the eigenvalues. and the corresponding feature vector : ; Find 1 non-negative eigenvalue, arranged in descending order of numerical value: ; Number of principal components Depending on the cumulative variance contribution rate, it is usually selected Each principal component contributes a cumulative rate of over 90%. For the selected feature vectors Normalization is performed so that the magnitude of its projection vector in the feature space is 1, that is: ; In the formula, Indicates the transpose symbol. Representing the kernel matrix The One eigenvalue; Core principal components were selected based on a cumulative variance contribution rate of ≥90% to eliminate correlations and data redundancy among indicators and extract core features that can represent the comprehensive damage state of the asphalt pavement layer on the bridge deck.

[0042] Step S43: Calculate the principal component scores and comprehensive evaluation index of each bridge deck section, sort them from low to high, and determine the maintenance priority of the bridge deck asphalt pavement layer.

[0043] In this embodiment, the score of each bridge deck segment on the core principal components is calculated, and the comprehensive evaluation index is calculated by weighting the principal component variance contribution rate. The process is as follows: If before One nonlinear principal component The information content (eigenvalues) are respectively For any road surface sample In its first Scores on each principal component Calculated by the formula, that is: ; In the formula, Indicates the first The th normalized eigenvector One portion, This represents the computation of the centered kernel function. Indicates training samples; Information contribution rate of each principal component The evaluation calculation formula is as follows: ; In the formula, Indicates the first Principal Components The corresponding eigenvalues, Indicates the first Eigenvalues ​​of each principal component; A comprehensive evaluation function for calculating the performance of bridge deck asphalt pavement layers. ,Right now: ; in, The higher the value, the better the overall performance of the asphalt pavement, and vice versa; Sorted from low to high according to the comprehensive evaluation index, the lower the index, the more serious the comprehensive damage to the asphalt pavement layer of the bridge deck, and the higher the maintenance priority. Finally, a maintenance priority list for the bridge deck asphalt pavement layer is generated, enabling scientific maintenance decisions.

[0044] Finally, it should be noted that the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement, characterized in that, include: Step S1: Use UAV aerial photography and GPR to collect external appearance data and internal structure data of bridge deck asphalt pavement layer. Perform occlusion removal, noise removal and signal correction on the collected data to obtain standardized multi-source detection data, including standardized external appearance data and standardized internal structure data. Step S2: Construct a full-domain 3D model of the bridge deck based on standardized external appearance data, identify macroscopic deformation and bridge approach slab settlement, extract full-section flatness data and calculate RQOI based on fuzzy logic to complete the full-domain quantification of external damage. Step S3: Based on the standardized internal structure data, analyze the pavement layer thickness distribution characteristics, calculate TCI, and integrate multiple source indicators such as PCI, RQOI, RDI, TCI, and bridge approach slab settlement value to complete the quantitative characterization of internal structure damage. Step S4: Standardize the multi-source indicators of internal and external damage, use the KPCA algorithm for nonlinear data fusion and feature dimensionality reduction, and complete the priority assessment and ranking of bridge deck asphalt pavement layer maintenance based on the comprehensive evaluation score.

2. The maintenance decision-making method based on the comprehensive internal and external damage status of bridge deck asphalt pavement as described in claim 1, characterized in that, Step S1 includes: Step S11: Use a UAV equipped with a full-frame camera to conduct a full-area aerial survey of the bridge surface according to preset parameters to obtain high-resolution visible light images and three-dimensional point cloud data; Step S12: Use 1600MHz GPR to perform an internal scan of the bridge deck roadway to obtain the pavement layer structure thickness reflection signal, and simultaneously collect the PCI, RQI, and RDI conventional index data of the road multi-functional inspection vehicle. Step S13: Use the YOLO-VEHICLE model to remove vehicles and shadows from the image, accurately identify the moving vehicles and shadows in the UAV aerial image, and remove occlusion noise through whitening mask processing. Step S14: The ROR denoising algorithm is used to filter the 3D point cloud data to remove redundant noise and outliers. Step S15: Perform time-domain correction, waveform smoothing, and noise suppression on the original GPR signal to eliminate environmental electromagnetic interference; Step S16: Finally, the processed appearance data, internal structure data, and conventional indicator data are unified in a single format to obtain a standardized multi-source detection dataset.

3. The maintenance decision-making method based on the comprehensive internal and external damage status of bridge deck asphalt pavement as described in claim 2, characterized in that, Step S2 includes: Step S21: Based on multi-view images, the SfM algorithm is used to complete the reconstruction of the three-dimensional model of the entire bridge deck and construct a relative digital elevation model to identify macroscopic uplift and subsidence deformation. Step S22: Based on the three-dimensional point cloud, locate the lane lines and expansion joints, extract longitudinal profile data and calculate the bridge approach slab settlement value, divide the bridge surface into equal parts laterally, and extract multi-section flatness data. Step S23: Using the severity and impact range of the flatness defect as input, the RQOI is calculated using the Mandani fuzzy inference system to complete the full-domain quantification of external damage.

4. The maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement as described in claim 3, characterized in that, Step S21 includes: Step S211: Based on the preprocessed multi-view UAV images, the SfM algorithm is used to perform feature matching and three-dimensional coordinate calculation to generate dense point cloud data. Step S212: Construct a high-precision 3D digital model of the entire bridge surface through point cloud optimization, mesh reconstruction, and texture mapping; Step S213: A normal bridge deck elevation model is constructed using a 2.5D polygon surface fitting method. The actual elevation and the fitted elevation are subtracted point by point to generate a relative digital elevation model of the bridge deck. Macroscopic uplift and large-scale deformation areas of subsidence are visually identified through a color distribution map.

5. The maintenance decision-making method based on the comprehensive internal and external damage status of bridge deck asphalt pavement as described in claim 3, characterized in that, Step S22 includes: Step S221: Based on the 3D point cloud, a generative adversarial network is used to perform pixel-level segmentation and localization of lane lines, and point cloud data of the longitudinal profiles on the left and right sides of the expansion joint are extracted with the expansion joint as the center. Step S222: Principal component analysis is used to determine the driving direction, the center of the expansion joint is located by peak detection method, and the elevation difference between the two sides is calculated to obtain the bridge approach slab value; Step S223: Divide the single lane into 30 survey line positions along the transverse direction of the bridge deck, extract the longitudinal profile elevation sequence at each position, calculate the international smoothness index of each survey line using a quarter-car model, and convert it into an RQI value.

6. The maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement as described in claim 3, characterized in that, Step S23 includes: Step S231: Based on the Mandani fuzzy inference system, the membership relationship between input variables and output variables is established using triangular membership function and trapezoidal membership function, wherein the severity and range of RQI defects are used as input variables; The membership function formula for a triangle is as follows: ； In the formula, Describes the membership function of a triangle. and Let represent the lower and upper bounds of the membership function of a triangle, respectively. This represents a membership degree measurement value that is equal to 1. The formula for the trapezoidal membership function is as follows: ； In the formula, Represents the membership function of a trapezoid. and Let represent the lower and upper bounds of the trapezoidal membership function, respectively. and This represents the values ​​between 1 and 1 for the trapezoidal membership degree measurement. Step S232: The severity of RQI defects is divided into three levels: acceptable, problematic, and critical. The defect range is divided into three levels: local, moderate, and widespread. RQOI is calculated by defuzzification using the centroid method. The formula for calculating RQOI is as follows: ； In the formula, This indicates the precise output of RQOI. The aggregate membership function is calculated using the maximum aggregation method. A variable representing a possible output value; Step S233 comprehensively characterizes the driving quality and external damage degree of the entire cross section of the bridge deck asphalt pavement, and completes the full-area quantification of external damage.

7. The maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement as described in claim 6, characterized in that, Step S3 includes: Step S31: Analyze the GPR reflection signal, identify the reflection waveforms at the upper and lower interfaces of the pavement layer, calculate the actual thickness of the pavement layer at each measuring point based on the electromagnetic wave two-way travel time, and statistically analyze the mean, maximum, minimum, and thickness deviation rates of each bridge deck section to establish a thickness distribution database. Step S32: Calculate TCI based on the thickness distribution characteristics and deviation rate of the pavement layer. The TCI value is positively correlated with the thickness uniformity. The smaller the thickness deviation, the higher the TCI value, which indicates better internal structural integrity. Step S33: Integrate multiple source indicators such as PCI, RQOI, RDI, TCI, and bridge approach slab settlement value to construct a quantitative indicator set for internal structural damage. This set comprehensively reflects the abnormal thickness and hidden damage status of the internal structure of the asphalt pavement layer on the bridge deck, thus completing the quantitative characterization of the internal structural damage of the asphalt pavement layer on the bridge deck.

8. The maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement as described in claim 7, characterized in that, Step S4 includes: Step S41: The Z-Score method is used to normalize the five indicators, PCI, RQOI, RDI, TCI, and bridge approach slab slab value, to eliminate differences in dimensions and values. The formula is as follows: ； In the formula, Indicates the first The first evaluation object The value of each indicator, Indicates the number of samples. , Indicates the dimension of the indicator. ; After standardization, the standardized matrix is ​​obtained. ,Right now: ； ； ； In the formula, This represents the standardized indicator variable. Indicates the first The sample mean of each indicator, Indicates the first The variance of each indicator; Step S42: Select the Gaussian radial basis kernel function to construct the kernel matrix and complete the centering process. Perform eigenvalue decomposition on the kernel matrix and select the core principal components according to the cumulative variance contribution rate. Step S43: Calculate the principal component scores and comprehensive evaluation index of each bridge deck section, sort them from low to high, and determine the maintenance priority of the bridge deck asphalt pavement layer.

9. A maintenance decision-making method based on the comprehensive condition of internal and external damage to bridge deck asphalt pavement as described in claim 8, characterized in that, In step S42, the process of constructing the kernel matrix and completing the centering process using the Gaussian radial basis kernel function is as follows: Build kernel matrix : ； In the formula, and They represent the first The and the first A standardized road surface sample vector, This represents a kernel function used to compute two sample vectors. and The similarity between them Represents an exponential function. The parameter representing the width of the Gaussian kernel; Kernel matrix By performing centralization processing, a centralized kernel matrix is ​​obtained. : ； In the formula, Represent a A matrix whose elements are all equal to 0. ; Kernel matrix Perform eigenvalue decomposition and solve for the eigenvalues. and the corresponding feature vector : ； Find 1 non-negative eigenvalue, arranged in descending order of numerical value: ； For the selected feature vectors Normalization is performed so that the magnitude of its projection vector in the feature space is 1, that is: ； In the formula, Indicates the transpose symbol. Representing the kernel matrix The One eigenvalue; Core principal components were selected based on a cumulative variance contribution rate of ≥90% to eliminate correlations and data redundancy among indicators and extract core features that can represent the comprehensive damage state of the asphalt pavement layer on the bridge deck.

10. A maintenance decision-making method based on the comprehensive internal and external damage status of bridge deck asphalt pavement as described in claim 9, characterized in that, In step S44, the score of each bridge deck segment on the core principal component is calculated, and the comprehensive evaluation index is calculated by weighting the principal component variance contribution rate. The process is as follows: If before One nonlinear principal component The amount of information is respectively For any road surface sample In its first Scores on each principal component Calculated by the formula, that is: ； In the formula, Indicates the first The th normalized eigenvector One portion, This represents the computation of the centered kernel function. Indicates training samples; Information contribution rate of each principal component The evaluation calculation formula is as follows: ； In the formula, Indicates the first Principal Components The corresponding eigenvalues, Indicates the first Eigenvalues ​​of each principal component; A comprehensive evaluation function for calculating the performance of bridge deck asphalt pavement layers. ,Right now: ； in, The higher the value, the better the overall performance of the asphalt pavement, and vice versa; Sorted from low to high according to the comprehensive evaluation index, the lower the index, the more serious the comprehensive damage to the asphalt pavement layer of the bridge deck, and the higher the maintenance priority. Finally, a maintenance priority list for the bridge deck asphalt pavement layer is generated, enabling scientific maintenance decisions.

Images