Bayesian algorithm coupled finite element method for automated analysis of road structure safety

Abstract

This invention discloses an automated road structure safety analysis method coupled with Bayesian algorithm and finite element method. It collects road surface crack images, multi-period elevation data, settlement monitoring data, and geological, geotechnical, design, and construction data to construct a comprehensive database. Through crack identification, elevation processing, coordinate transformation, segmentation, and mileage matching, a three-level index database of "road segment-section-measuring point" is established. Road segments are screened based on crack and settlement characteristics, and an initial finite element model is established to identify road segments with abnormal settlement. An Optuna tree-structured Parzen estimator is used for sampling, and the optimal parameters are automatically inverted by combining loss targets and engineering constraints. The backfill model is used to simulate the entire process from construction to long-term service, and the data is spatially compared with measured data to analyze the correlation and influencing factors between abnormal settlement and cracks, outputting a safety diagnosis and risk assessment. This method can be used for crack cause analysis, abnormal settlement risk identification, and operation and maintenance decision-making.

Description

Technical Field

[0001] This invention relates to the fields of geotechnical engineering, road engineering, and road defect detection and structural safety analysis. Specifically, it relates to a method for automated analysis of road structural safety using a Bayesian algorithm coupled with finite element method, which is particularly suitable for analyzing the causes of road surface cracks, identifying abnormal settlement risks, diagnosing road structural safety, and providing decision support for road operation and maintenance. Technical Background

[0002] As roads age, they and their subgrade structures may experience cumulative deformation to varying degrees due to long-term traffic loads, environmental effects, and the evolution of existing subgrade material properties. On one hand, environmental factors such as continuous rainfall and changes in groundwater conditions can alter the water content and mechanical properties of the subgrade soil. On the other hand, highway widening and reconstruction, subgrade splicing, and localized construction disturbances can also change the stress and deformation state of the existing subgrade, leading to uneven settlement and surface cracks in some road sections. Therefore, a method is needed to rapidly analyze the deformation state of the internal road structure by combining data on surface defects, settlement monitoring, and existing engineering data during routine road inspections, operational safety assessments, and the analysis of existing road conditions during reconstruction and expansion projects.

[0003] In areas with deep soft soil, soft soil typically exhibits high water content, high compressibility, low permeability, and characteristics of long-term consolidation and creep. Under the influence of road embankment, preloading, and long-term operational loads, soft soil foundations may experience continuous settlement. When geological conditions, soft soil treatment methods, or subgrade treatment parameters differ across different road sections, longitudinal or transverse differential deformation may also occur, which is related to the formation and development of transverse, longitudinal, or diagonal cracks on the road surface. Therefore, how to combine readily available information on road surface cracks and settlement deformation to analyze the internal deformation state of the road and its main engineering influencing factors is a technical problem that needs to be solved in the road operation and maintenance process.

[0004] Current road structure condition analysis typically employs techniques such as pavement surface defect detection, settlement monitoring, or finite element numerical analysis. While pavement surface crack detection can reveal the location, type, and geometric characteristics of cracks, it's difficult to determine whether cracks are related to differential settlement of the subgrade based solely on their appearance. Settlement monitoring can reflect the deformation development process at specific monitoring points, but monitoring points, crack locations, geological conditions, road design information, and soft soil treatment design and construction parameters are usually stored separately, making it difficult to establish a unified spatial correspondence.

[0005] For finite element analysis, the constitutive parameters of the soil and some subgrade treatment parameters in the model are usually determined based on survey, test, or design data. However, due to factors such as the spatial variability of the soil, construction deviations, and changes in long-term service performance, some parameters are uncertain. In traditional analysis processes, engineers usually need to repeatedly adjust the model parameters and re-perform finite element calculations, and then manually determine the parameter combinations based on the differences between the simulated settlement results and the measured data. This process results in low efficiency for parameter adjustment and model calibration.

[0006] Furthermore, there is a lack of a unified data association and feedback mechanism between existing crack detection data, settlement monitoring data, and finite element calculation results, making it difficult to form an integrated technical process from road surface defect identification, abnormal settlement screening, finite element parameter inversion to road structure safety analysis.

[0007] Therefore, existing technologies still have room for improvement in terms of unified spatial correlation of road surface cracks, multi-stage elevation, settlement monitoring, geological survey and road design and construction data, automatic inversion of uncertain parameters of finite element models, and joint analysis between finite element structural response and measured cracks. Summary of the Invention

[0008] This invention aims to solve the technical problems in existing road structure safety analysis processes, such as the difficulty in effectively correlating multi-source detection with engineering data, the reliance on manual experience for crack risk assessment, the need for repeated manual adjustment of finite element model parameters, and the difficulty in establishing a quantitative correspondence between crack appearance characteristics and roadbed internal deformation mechanisms. It provides a method for automated analysis of road structure safety by coupling Bayesian algorithm with finite element method.

[0009] According to a first aspect of the embodiments of this application, a method for automated analysis of road structure safety coupled with a Bayesian algorithm and finite element method is provided, comprising: Collect and integrate road surface crack images, road surface elevation data at different detection periods, settlement monitoring data during construction and operation, geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, and store them according to unified data fields to build a full-dimensional original database; Data cleaning, coordinate transformation, error correction, and spatial registration are performed on the road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database. Data from different sources are unified to the geodetic coordinate system used in road design. Crack locations, elevation measurement points, and settlement monitoring points are matched with the road design horizontal alignment and mileage station numbers. Using the road mileage station numbers as a unified spatial index, a three-level indexed structured database of "road segment - cross section - measurement point" is constructed. Based on the aforementioned three-level indexed structured database, candidate road sections are selected for analysis according to the type of road surface cracks, the spatial location of cracks, and the corresponding settlement and deformation characteristics. The geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate road sections are used to establish an initial finite element calculation model for the candidate road sections. Parameters from the geological survey and design / construction data are used as the initial parameters for the finite element model. Abnormal settlement road sections are identified by combining measured settlement and deformation data with the initial finite element calculation results, and a correlation sample set between crack characteristics, settlement and deformation characteristics, geological conditions, and subgrade treatment parameters is constructed. Based on the measured settlement and deformation data in the associated sample set, a comprehensive loss objective function is constructed to evaluate the difference between the finite element simulation deformation data and the measured deformation data. Based on geological survey data and roadbed treatment design and construction parameters, the prior search interval and physical constraints of the parameters to be inverted are determined. The tree structure Parzen estimator sampling strategy under the Optuna framework is used to automatically iteratively search for the parameters to be inverted in the finite element model. The comprehensive loss objective function is used as the basis for parameter optimization to obtain the parameter combination with the best fit between the finite element simulation deformation data and the measured deformation data. The optimal parameter combination is backfilled into the initial finite element calculation model to simulate the subgrade filling, preloading, operation, and long-term service process, obtaining deformation, stress, and excess pore water pressure response data of the subgrade and road structure. The response data is spatially correlated and compared with the location, type, and distribution characteristics of measured cracks on the road surface to determine the correspondence between abnormal settlement and road surface cracks, assess the safety status of the road structure, identify the main influencing factors related to the formation of road surface cracks, and output the road structure safety diagnosis results and risk assessment results.

[0010] According to a second aspect of the embodiments of this application, a system for automated analysis of road structure safety coupled with a Bayesian algorithm and finite element method is provided, comprising: The data acquisition and raw database construction module is used to collect and integrate road surface crack images, road surface elevation data at different detection periods, settlement monitoring data during construction and operation, geological survey data, indoor geotechnical physical and mechanical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, and store them according to unified data fields to build a full-dimensional raw database. The spatiotemporal registration and structured database construction module is used to perform data cleaning, coordinate transformation, error correction, and spatial registration on the road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database. It unifies data from different sources to the geodetic coordinate system used in road design, and matches crack locations, elevation measurement points, and settlement monitoring points with the road design horizontal alignment and mileage station numbers. Using the road mileage station numbers as a unified spatial index, it constructs a three-level indexed structured database of "road segment - cross section - measurement point". The candidate analysis segment screening and initial finite element analysis module is used to screen candidate analysis segments based on the three-level indexed structured database, according to the type of road surface cracks, the spatial location of cracks, and the corresponding settlement deformation characteristics. It then calls upon the geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate analysis segments to establish an initial finite element calculation model for the candidate analysis segments. Parameters from the geological survey and design / construction data are used as initial parameters for the finite element model calculation. Finally, it combines measured settlement deformation data with the initial finite element calculation results to identify abnormal settlement segments and constructs a correlation sample set between crack characteristics, settlement deformation characteristics, geological conditions, and subgrade treatment parameters. The Bayesian optimization parameter inversion module is used to construct a comprehensive loss objective function based on the measured settlement deformation data in the associated sample set to evaluate the difference between the finite element simulation deformation data and the measured deformation data. It also determines the prior search interval and physical constraints of the parameters to be inverted based on geological survey data and roadbed treatment design and construction parameters. The module employs a tree-structured Parzen estimator sampling strategy under the Optuna framework to automatically iteratively search for the parameters to be inverted in the finite element model, using the comprehensive loss objective function as the basis for parameter optimization, to obtain the parameter combination with the optimal fit between the finite element simulation deformation data and the measured deformation data. The finite element calibration and road structure safety analysis module is used to backfill the optimal parameter combination into the initial finite element calculation model, simulate the roadbed filling, preloading, operation and long-term service process, and obtain the deformation, stress and excess pore water pressure response data of the roadbed and road structure; spatially correlate and compare the response data with the location, type and distribution characteristics of the measured cracks on the road surface to determine the correspondence between abnormal settlement and road surface cracks, judge the safety status of the road structure, identify the main influencing factors related to the formation of road surface cracks, and output the road structure safety diagnosis results and risk assessment results.

[0011] According to a third aspect of the embodiments of this application, an electronic device is provided, comprising: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method as described in the first aspect.

[0012] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the method described in the first aspect.

[0013] As can be seen from the above embodiments, compared with the prior art, the present invention, based on its original combination of technologies and parameter constraints, has the following outstanding beneficial effects: First, by uniformly storing road surface cracks, multi-stage road surface elevations, settlement monitoring, geological surveys, road design, roadbed treatment design, construction and as-built data, and establishing a three-level index relationship of "road section - cross section - measuring point" based on road mileage station numbers, road engineering data from different sources can be correlated under a unified spatial benchmark, providing a data foundation for joint analysis of cracks, settlement deformation, geological conditions and soft soil treatment structures.

[0014] Second, by screening candidate road sections based on crack characteristics and settlement deformation characteristics, the amount of computation generated by performing finite element modeling and parameter inversion on all road sections can be reduced, allowing subsequent numerical analysis to focus on road sections with abnormal deformation risks.

[0015] Third, by calling up geological survey, indoor geotechnical test, road design, subgrade treatment design, construction and as-built data to automatically establish an initial finite element calculation model, the finite element analysis is based on the actual engineering geometry, material parameters and construction stage, reducing the deviation caused by relying entirely on manual experience to set model parameters.

[0016] Fourth, by adopting the tree-structured Parzen estimator sampling strategy under the Optuna framework, a closed loop is formed, which includes candidate parameter generation, automatic finite element calculation, extraction of simulated settlement results, calculation of comprehensive loss value, and subsequent sampling strategy updates. The engineering physical constraints of soil parameters and roadbed treatment structure parameters are used to limit the parameter search range, thereby improving the automation level and engineering rationality of finite element model parameter inversion.

[0017] Fifth, by backfilling the optimal parameter combination obtained from the inversion into the initial finite element calculation model and simulating the road construction, preloading, operation and long-term service process, the simulation results can be analyzed in correspondence with the measured settlement and spatial distribution of cracks, thereby providing a quantitative basis for judging the main causes of road cracks, analyzing the safety status of road structures and subsequent risk assessments.

[0018] Sixth, the method can use road surface crack and elevation data obtained by vehicle-mounted detection to conduct road structure condition analysis, which can reduce the workload of manual point-to-point inspection and repeated manual adjustment of finite element parameters, and improve the automation of data processing and numerical analysis in the road structure safety analysis process.

[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0021] Figure 1 This is a flowchart illustrating a method for automated analysis of road structure safety using a Bayesian algorithm coupled with finite element method, based on an exemplary embodiment.

[0022] Figure 2 This is a road surface crack acquisition and calibration map shown according to an exemplary embodiment, wherein (a) is a map of the overall crack distribution of the route, and (b) is a detailed calibration map of the crack locations.

[0023] Figure 3 This is a diagram illustrating a three-level indexed database structure according to an exemplary embodiment.

[0024] Figure 4 This is a flowchart illustrating the spatiotemporal registration of crack location calibration data according to an exemplary embodiment.

[0025] Figure 5 This is a schematic diagram of a road surface treatment section and a road longitudinal profile according to an exemplary embodiment.

[0026] Figure 6 This is a road surface elevation variation diagram illustrated according to an exemplary embodiment.

[0027] Figure 7 This is a diagram illustrating the relationship between cracks and treatment methods according to an exemplary embodiment. Figure 8 This is a flowchart illustrating crack-settlement correlation analysis according to an exemplary embodiment.

[0028] Figure 9 This is a cross-sectional settlement curve of a road section during the construction period, illustrated according to an exemplary embodiment.

[0029] Figure 10 This is a longitudinal section settlement curve of a road section during the construction period, illustrated according to an exemplary embodiment.

[0030] Figure 11This is a road surface cross-section settlement feature extraction map illustrated according to an exemplary embodiment.

[0031] Figure 12 This is a flowchart illustrating a Bayesian optimization process according to an exemplary embodiment.

[0032] Figure 13 This is a comparison chart of simulated settlement values ​​and measured settlement values ​​under different pile equivalent parameter conditions according to an exemplary embodiment.

[0033] Figure 14 This is a comparison chart of calculated and observed settlement values ​​at the center of a road, according to an exemplary embodiment.

[0034] Figure 15 This is a diagram illustrating the results of a finite element analysis according to an exemplary embodiment.

[0035] Figure 16 This is a comparison diagram of road surface cracks and finite element simulation deformation space, according to an exemplary embodiment.

[0036] Figure 17 This is a system block diagram illustrating an automated analysis of road structure safety using a Bayesian algorithm coupled with finite element method, according to an exemplary embodiment. Detailed Implementation

[0037] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of systems and methods consistent with some aspects of this application.

[0038] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0039] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0040] The technical solution of the present invention will be described in detail below with reference to a specific implementation case. This embodiment takes the section K202+700-K226+000 of the SY04 section of a bridge and connecting road project in a certain province as the actual measurement object. This section is a soft soil subgrade with a soft soil depth of up to 24.5m. A composite foundation treatment of prestressed pipe piles and cement mixing piles is adopted. The prestressed pipe piles are 25m long and the pile spacing is 2.3m. The cement mixing piles are 0.5m in diameter and the pile spacing is 1.7-1.9m. The construction adopts a five-level surcharge scheme. The total construction time is 390 days. During the operation period, the average daily traffic volume is large, and there are obvious road surface cracks and settlement phenomena. This embodiment is used to specifically illustrate the implementation process of the present invention and is not intended to limit the scope of protection of the present invention.

[0041] Figure 1 This is a flowchart illustrating an automated method for road structure safety analysis using a Bayesian algorithm coupled with finite element method, according to an exemplary embodiment. Figure 1 As shown, the method may include the following steps: S1: Collect and integrate road surface crack images, road surface elevation data at different detection periods, settlement monitoring data during construction and operation, geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, and store them according to unified data fields to construct a full-dimensional original database; specifically including the following sub-steps: S11: Establish a structured data table to store road engineering data from different sources, and set the data source, road mileage station number, spatial coordinates, collection time, data type and data value fields according to the data category.

[0042] Specifically, in this embodiment, a DataFrame structured data table is created using the Pandas library in the Python environment. The DataFrame organizes data according to rows and columns, where each row corresponds to a road measurement point, a crack record, or an engineering parameter record, and each column is set as a corresponding data attribute field.

[0043] Depending on the type of data stored, at least the following fields should be set: data source, road mileage station number, collection time, longitude, latitude, plane coordinates, elevation, crack type, crack length, crack width, soil layer number, soil layer thickness, soil physical and mechanical parameters, soft soil treatment method, pile length, pile diameter, pile spacing, embankment filling height, construction stage, and settlement monitoring value.

[0044] For data whose road mileage markers or design coordinates cannot be determined at present, the original timestamp, latitude and longitude, and data source identifier are retained in this step, and coordinate transformation and mileage matching are performed in the subsequent step S2.

[0045] Using the above data structure, vehicle detection data, geological survey data, design and construction data, and historical monitoring data can be stored according to unified data fields, enabling data from different sources to be correlated through spatial location, road mileage, and time information, providing a unified data interface for the subsequent establishment of a three-level index database and automatic finite element modeling.

[0046] S12: An on-board detection system integrating a high-definition camera, GNSS receiver, and IMU inertial measurement unit is used to collect road surface crack images and road surface elevation data with timestamps and spatial location information along the target road, and write the road surface crack images, road surface elevation data, corresponding collection time, and spatial location information into the structured data table.

[0047] Specifically, for the soft soil subgrade inspection section in this embodiment, a high-definition camera, a GNSS receiver, a high-precision IMU inertial measurement unit, and a laser ranging device are integrated into the inspection vehicle.

[0048] In this embodiment, the image resolution of the high-definition camera is set to no less than 1080P, and the image acquisition frame rate is set to 20Hz; the sampling frequency of the GNSS receiver and the IMU inertial measurement unit is set to 20Hz. All devices adopt a unified time reference and can be synchronized through PPS second pulse signals, so that the road surface images, vehicle spatial positions, vehicle attitudes, and road surface elevation measurement results acquired at the same time can establish a correspondence.

[0049] As the vehicle travels along the target road section, a high-definition camera continuously captures images of the road surface; a GNSS receiver simultaneously records the latitude and longitude information of the vehicle's location; an IMU inertial measurement unit records changes in the vehicle's attitude; and a laser rangefinder measures the relative distance between the vehicle and the road surface to reduce the impact of vertical vibrations of the vehicle body on the road surface elevation acquisition results.

[0050] Each data acquisition device records the acquisition time according to a unified time base. The road surface images, elevation measurements and spatial locations acquired at the same sampling time are bound together by timestamps to form a data record that includes at least the acquisition time, latitude and longitude, road surface image and road surface elevation. This data is then written into the structured data table established in S11.

[0051] For the same road segment, road surface elevation is repeatedly collected at different inspection periods. By recording the inspection date and collection time, the elevation data from different inspection periods can be matched in subsequent steps according to the same road location, providing a data basis for obtaining the changes in road surface elevation over time.

[0052] S13: Collect settlement monitoring data, geological survey data, indoor geotechnical physical and mechanical test data, road design data, subgrade treatment design and construction data, and as-built data corresponding to the target road during the construction and operation periods. Extract soil layer distribution and soil physical and mechanical parameters, road horizontal and vertical alignment and mileage information, subgrade cross-section and structural dimensions, soft soil treatment form and treatment range, pile geometry and material parameters, embankment graded filling and preloading construction parameters, and actual engineering parameters after completion. Record the geodetic coordinate system used in the road design data. Write the extracted data into the structured data table according to the corresponding road mileage, cross-section location, and data category to form a full-dimensional original database for subsequent spatiotemporal registration, finite element modeling, and parameter inversion.

[0053] Specifically, for the target road section in this embodiment, geological survey reports, indoor geotechnical test reports, road construction design documents, soft soil foundation treatment design documents, construction records, as-built data, and settlement monitoring data during the construction and operation periods are retrieved.

[0054] Information such as borehole locations, soil layering, thickness of each soil layer, and groundwater level along the road are extracted from the geological survey report and used to establish the geological structure in the subsequent finite element model.

[0055] Soil parameters such as void ratio, compression coefficient, compression index, rebound index, creep parameter, shear strength, and permeability coefficient are extracted from indoor geophysical test data and used as the basis for determining the initial material parameters for subsequent finite element analysis and the search range for Bayesian optimization parameters.

[0056] The road horizontal curve, longitudinal profile design elevation, road mileage station, roadbed cross-sectional dimensions, and embankment design height are extracted from the road design data. The road horizontal curve, longitudinal profile design elevation, and mileage station information are used to match the vehicle-mounted detection data to the road design alignment, while the roadbed cross-sectional dimensions and embankment design height are used to establish the geometric structure of the finite element model.

[0057] The soft soil treatment methods and treatment ranges for different sections, as well as the pile length, pile diameter, pile spacing, and pile material parameters of prestressed pipe piles and cement mixing piles, are extracted from the subgrade treatment design and construction data. At the same time, the graded filling height of the embankment, the loading time of each grade, and the preloading duration are extracted to determine the subgrade treatment structural parameters and construction stage parameters in the finite element model.

[0058] Further information such as the actual construction length of the piles, the actual treatment range, the actual filling elevation, and the actual construction time are extracted from the as-built data. When there are differences between the actual construction parameters recorded in the as-built data and the design data, the actual as-built parameters are recorded as the corresponding engineering parameters to provide basic engineering data for subsequent finite element modeling and parameter inversion.

[0059] The road mileage, cross-sectional location, monitoring date, and cumulative settlement value of the monitoring points are extracted from the settlement monitoring data during the construction and operation periods and saved in chronological order of monitoring time. This data is used to construct the full-cycle settlement time series data of the road and also serves as the measured comparison data for finite element calculation results and Bayesian parameter inversion.

[0060] Simultaneously, the coordinate reference system used in the road design data is read, and the coordinate system type is stored as a database attribute field; the coordinate reference system serves as the target coordinate benchmark for coordinate transformation and spatial registration in the subsequent step S23.

[0061] Finally, the above-mentioned engineering data and monitoring data are written into the structured data table established in S11 according to the corresponding engineering location, road mileage, cross-section location, data category and time information, and together with the vehicle detection data obtained in S12, they form a full-dimensional original database.

[0062] S2: The road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database are cleaned, coordinate transformed, error-corrected, and spatially registered. Data from different sources are unified to the geodetic coordinate system used in road design. Crack locations, elevation measurement points, and settlement monitoring points are matched with the road design horizontal alignment and mileage markers. Using road mileage markers as a unified spatial index, a three-level indexed structured database of "road segment - cross-section - measurement point" is constructed. Specifically, this includes the following steps: S21: Perform image preprocessing and crack feature recognition on the road surface crack image. Use an edge recognition algorithm combined with morphological operations to remove image noise and extract crack contours to obtain crack length, crack width, crack morphology, acquisition time and spatial location information. The crack morphology includes at least longitudinal cracks, transverse cracks and diagonal cracks to form a crack feature dataset with spatial location information.

[0063] Specifically, in this embodiment, the continuous road surface images acquired in S12 are read sequentially according to the timestamps, and the images are converted to grayscale, enhanced, and noise processed. On this basis, the grayscale change regions of the road surface are extracted by the edge recognition algorithm, and then the discrete crack pixel regions are connected by morphological operations, and isolated noise regions are removed to obtain the continuous crack contour.

[0064] Based on the identified crack contours, the crack length, width, and extension direction are extracted. According to the extension direction of the crack relative to the road driving direction, the cracks are divided into transverse cracks, longitudinal cracks, and diagonal cracks.

[0065] Further, based on the acquisition timestamp corresponding to the crack image, the GNSS spatial location information corresponding to the timestamp is called to bind the crack identification result with the latitude and longitude location, forming a crack feature record that includes at least crack number, crack type, crack length, crack width, acquisition time, longitude and latitude fields, and writes it into the structured database.

[0066] In the road section under inspection in this embodiment, the location and morphology information of each crack are obtained through the above method, and the distribution of different types of cracks along the route is statistically analyzed. By binding crack features with spatial location, subsequent steps can reverse-engineer the corresponding road mileage, soft soil treatment method, and settlement deformation data based on the crack location, avoiding isolated judgments based solely on crack images.

[0067] It should be noted that this step is mainly used to identify and spatially mark the cracks themselves. The relationship between cracks and soft soil treatment methods and settlement anomalies will be analyzed in the subsequent S3 step, and the cause of cracks will not be directly determined in this step.

[0068] The results of road surface crack identification and feature extraction in this embodiment are as follows: Figure 2 As shown, where Figure 2 (a) represents the overall spatial distribution of different types of cracks along the target route. Figure 2 (b) indicates the calibration information such as mileage, type, length, risk level and collection time of the selected crack location.

[0069] S22: Outlier removal, filtering and noise reduction, and system error correction are performed on the road surface elevation data collected at different detection periods to obtain continuous road surface elevation curves for each detection period; based on the elevation changes of the same road location at different detection periods, the corresponding road surface elevation changes are obtained, and road deformation time series data are formed by combining the settlement monitoring data during the construction and operation periods.

[0070] Specifically, the raw road surface elevation data collected in S12 are arranged in order of collection time and spatial location.

[0071] First, outlier detection is performed on the raw elevation data. Obvious outlier measurement points caused by abnormal GNSS signals, instantaneous vehicle vibration, sensor lock-off, or abnormal data transmission are identified and removed.

[0072] Subsequently, a filtering algorithm is used to suppress high-frequency fluctuations in the continuous elevation data. In this embodiment, Kalman filtering and / or wavelet filtering can be used to smooth the elevation data, reducing the impact of measurement noise while maintaining the overall trend of the road's longitudinal elevation.

[0073] After outlier removal and filtering, continuous road surface elevation curves are generated for each detection period according to the detection time.

[0074] When multiple periods of monitoring data exist for the same road location, the elevation measurements from different periods are correlated, and the elevation change of the same road location over time is calculated to describe the vertical deformation of the road between different monitoring periods.

[0075] Simultaneously, the settlement monitoring data entered in S13 during the construction and operation periods are called up, and the settlement monitoring dates, monitoring point locations, and cumulative settlement amounts are organized in chronological order to obtain the settlement time sequence records corresponding to the monitoring points.

[0076] This results in road deformation time-series data that simultaneously includes changes in vehicle road surface elevation and existing settlement monitoring data, providing a data foundation for subsequent extraction of cumulative settlement, settlement gradient, and settlement rate.

[0077] The purpose of adopting the above method is to transform the original elevation measurement values ​​into continuous data that can reflect the actual deformation development of the road, and reduce the impact of random noise and abnormal measurement values ​​on the subsequent finite element inversion.

[0078] S23: Based on the coordinate reference system used in the road design data, perform coordinate transformation on the spatial location of the crack obtained in S21 and the spatial location of the elevation measurement point obtained in S22, and transform the vehicle detection data from the acquisition coordinate system to the geodetic coordinate system used in the road design; at the same time, extract the horizontal curve and longitudinal profile design elevation data of the road design, so that the crack measurement point, elevation measurement point and road design data are under a unified spatial coordinate reference.

[0079] Specifically, the coordinate reference system used for the target road segment is read from the road design data stored in S13, and the horizontal curve data corresponding to the design route is extracted. and longitudinal profile design elevation data .

[0080] The original spatial locations of crack points and elevation measurement points acquired by the vehicle-mounted system are stored in latitude and longitude format. A coordinate transformation program is then used to convert the original latitude and longitude coordinates into plane coordinates in the geodetic coordinate system used in road design.

[0081] In this embodiment, coordinate transformation can be performed using the PyProj library in the Python environment.

[0082] Let the original position and elevation of a certain detection point obtained by the vehicle-mounted detection system be represented as:

[0083] After coordinate transformation, it can be represented as:

[0084] in, and These represent the original longitude and latitude coordinates collected by the vehicle-mounted detection system, respectively. This indicates the original elevation of the corresponding measuring point; and These represent the plane coordinates after transformation to the geodetic coordinate system used in road design. Maintain the elevation value of the corresponding measuring point.

[0085] The locations of cracks and road surface elevation measurement points were processed in the manner described above to ensure the road design horizontal curves were achieved. Longitudinal section data Crack data and elevation detection data are in the same coordinate system.

[0086] The coordinate transformation described above eliminates spatial discrepancies caused by different coordinate representations used in vehicle-mounted GNSS data and engineering design data, providing a unified spatial benchmark for subsequent road mileage matching and soft soil treatment section association. The coordinate transformation and spatiotemporal registration process between crack locations and elevation measurement points and road design data is as follows: Figure 4 As shown.

[0087] S24: Determine the locations of the bridgeheads and tails in the target road based on the road design data, and divide the road into segments using adjacent bridgeheads or tails as road segment boundaries; perform planar position calibration on the detection points within each segment, and use the as-built measured elevation or design elevation at the bridgehead or tail as a stable elevation benchmark, and perform segmented adjustment correction on the road surface elevation data within each segment to obtain detection point data that completes the planar and elevation position correction.

[0088] Specifically, after completing the coordinate transformation, the bridgehead and tail positions corresponding to the start and end points of the bridge are extracted from the road design data, and the roads between adjacent bridgeheads or tails are divided into relatively independent data subsets.

[0089] For each road subset, the coordinates of the transformed detection points in planar coordinates are compared with the road design horizontal curve. The coordinates are compared, and the overall planar offset of the detection point relative to the design alignment is determined based on the control positions of the bridgehead and tail and the road design centerline. The overall planar offset is applied to the coordinates of the detection point within the road subset to obtain the coordinates after planar position calibration.

[0090] Let the detection points after coordinate transformation be represented as:

[0091] The planar position calibration is represented as follows:

[0092] in, and The plane coordinates of the detection points after the road design alignment has been calibrated.

[0093] After completing the horizontal position calibration, the bridgehead or bridge tail positions at both ends of the same road segment are used as the elevation control positions, and the road surface elevation within the road segment is adjusted linearly in segments.

[0094] In this embodiment, the elevation control position of the current segment starting point is used as the zero adjustment benchmark, and linear allocation correction is implemented for each measuring point within the segment based on the closure difference between the measured elevation of the segment ending point and the control elevation.

[0095] Let the distance from a certain detection point to the starting point of the current road segment be... The total length of this segment is The closure error between the measured elevation at the end of the road section and the control elevation is: The elevation adjustment for that measuring point is:

[0096] The adjusted elevation is:

[0097] in: : The distance of the measuring point from the starting point of the current segment; : Current total segment length; : Current segment endpoint elevation closure error; Measurement point Elevation adjustment amount at the location; : Measured elevation after filtering; Elevation after piecewise adjustment.

[0098] By correcting each detection point in the entire sub-section using the above method, the coordinates of the detection points with calibrated horizontal and vertical positions are finally obtained:

[0099] in, This indicates the corrected elevation after the elevation adjustment is completed.

[0100] In this embodiment, the bridgehead or bridge tail is selected as the control position for road data segmentation and elevation calibration. The purpose is to use the existing clear engineering control positions along the route to limit the plane position error and elevation system deviation accumulated during the long-distance road detection process to local road sections between adjacent bridges, thereby reducing the impact of the continuous accumulation of errors along the entire road.

[0101] S25: Match the locations of the corrected cracks, elevation measurement points, and settlement monitoring points with the road design horizontal alignment to determine the corresponding road mileage station numbers; use the road mileage station numbers as a unified spatial index to link and integrate crack characteristic data, road surface elevation change data, settlement monitoring data during construction and operation, geological survey data, road design data, roadbed treatment design and construction parameters, and as-built data to form a three-level indexed structured database organized according to "road section - cross section - measurement point".

[0102] Specifically, in this embodiment, the corrected plane coordinates of each crack point and elevation measuring point obtained in S24 are ( , ) and road design horizontal curve Perform spatial matching.

[0103] For each detection point, the route position with the smallest spatial distance to the detection point is searched in the road design alignment, and the road mileage station number of that position is obtained according to the correspondence of the design route. In this way, the detection data originally represented by spatial coordinates is converted into a road mileage expression that can be directly used in the project.

[0104] For example, once a crack measuring point is matched to the road design line K225+850, the crack type, crack length, crack width, and spatial location of that crack are uniformly associated with the K225+850 mileage.

[0105] Using the same method, the elevation measurement points, settlement monitoring points during construction, geological boreholes, and soft soil treatment areas at the corresponding locations are mapped to the road mileage system.

[0106] After mileage matching is completed, data from different sources are linked using road mileage markers as the main spatial association field.

[0107] in: The first-level "road segment" is divided according to the adjacent completed and measured bridgeheads, bridge tails, and engineering segment ranges. It is used to record the overall road design conditions, soft soil treatment methods, geological conditions, and distribution of defects of the road segment. The second-level "section" is set according to the road mileage marker and is used to record the roadbed structure, stratum structure, settlement status and crack conditions at a specific road cross-section location; The third level of "measuring points" is used to store specific crack locations, elevation measuring points, settlement monitoring points, and corresponding time series data.

[0108] For example, for a specific road section requiring soft soil treatment, a data index can be established according to the following relationship: Section A → K225+800 cross section → left elevation measuring point, road center elevation measuring point, right elevation measuring point and crack points near this cross section; Section A → K225+850 cross section → left elevation measuring point, road center elevation measuring point, right elevation measuring point and crack points near this cross section.

[0109] The corresponding road sections and cross-sections are associated with the thickness of the soft soil layer, soil physical and mechanical parameters, pile type, pile length, pile diameter, pile spacing, construction stage, and historical settlement data.

[0110] In this embodiment, the Pandas library in the Python 3.12 environment can be used to establish the relationship between the above data, and the road mileage markers can be used as index fields to perform data query and call; Dynamo and Civil 3D can also be combined to perform three-dimensional visualization of road horizontal curves, longitudinal profiles and measured points to check the spatial correspondence between measured data and design data.

[0111] The purpose of adopting the "road segment-section-measuring point" three-level index structure is to transform the heterogeneous data that was originally stored separately in crack images, vehicle elevation data, settlement monitoring tables, geological survey reports and road design documents into a data structure with a unified road spatial location relationship.

[0112] Through this three-level index structure, subsequent steps can retrieve elevation change data, historical settlement data, geological parameters, and soft soil treatment design and construction parameters of the corresponding road segment and adjacent sections based on the road mileage where the target crack is located. This provides structured data input for candidate analysis segment screening, automatic finite element model establishment, and Bayesian parameter search.

[0113] The organization of the "road segment-section-measuring point" three-level indexed structured database is as follows: Figure 3 As shown.

[0114] S3: Based on the three-level indexed structured database, candidate road sections are selected for analysis according to the type of road surface cracks, the spatial location of cracks, and the corresponding settlement deformation characteristics; the geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate road sections are called to establish an initial finite element calculation model for the candidate road sections, and the parameters in the geological survey and design and construction data are used as the initial parameters of the finite element model for calculation; abnormal settlement road sections are determined by combining the measured settlement deformation data and the initial finite element calculation results, and a correlation sample set between crack characteristics, settlement deformation characteristics, geological conditions, and subgrade treatment parameters is constructed; specifically, the following steps are included: S31: Based on the crack type, crack spatial coordinates, road mileage markers, and corresponding road surface elevation change data and settlement monitoring data in the three-level indexed structured database, candidate analysis road sections with preset crack characteristics are selected; wherein, the candidate analysis road sections include road sections with diagonal cracks, road sections with transverse through cracks, and road sections with longitudinal cracks; for road sections with diagonal cracks or longitudinal cracks, a predetermined distance is selected before and after the location of the corresponding crack as the center. The scope is defined as the finite element analysis scope; for road sections with transverse through cracks, the segment containing the transverse through crack is defined as the finite element analysis scope.

[0115] Specifically, based on the three-level indexed structured database constructed according to S25, the crack data is first used as the retrieval entry point. The road section and cross-section where the crack is located are located according to the road mileage station corresponding to the crack. The data on the change of road surface elevation, settlement monitoring data during construction period and settlement monitoring data during operation period are called up.

[0116] like Figure 7 As shown, the spatial distribution of cracks is superimposed on the road soft soil treatment section to obtain the soft soil treatment method, treatment range, and boundary position between different treatment methods corresponding to the location of the cracks.

[0117] For diagonal and longitudinal cracks, since they may reflect local differential deformation, a predetermined distance is extended longitudinally forward and backward from the location of the crack. This forms a local analysis section.

[0118] The predetermined distance The predetermined distance is used to ensure that the finite element analysis range covers the crack location and representative deformation areas on both sides of the crack. The range of the finite element analysis is determined based on the longitudinal influence range of the crack along the road, the layout range of the monitoring sections near the crack, and the engineering boundary of the soft soil treatment section where the crack is located. The selected finite element analysis range should at least cover the crack location, the deformation area near the crack, and the area of ​​soft soil treatment structure changes that may be related to the crack.

[0119] In this embodiment, the predetermined distance Take 20m; when the distance between monitoring sections adjacent to the crack is greater than 20m, or the transition zone of the soft soil treatment method exceeds this range, Extend to a distance that can simultaneously cover at least two monitoring sections on both sides of the crack and the transition area of ​​the complete treatment method.

[0120] For transverse through cracks, since the cracks are continuously distributed along the cross-section of the road, the entire corresponding road segment obtained by dividing S24 according to the bridgehead, bridge tail or engineering segment boundary is used as the analysis scope, so as to analyze the differential deformation of the road in both the transverse and longitudinal directions at the same time.

[0121] By using the above screening method, the entire route is narrowed down to candidate road sections that require finite element analysis, thus avoiding the need to perform high-computation finite element analysis on all road sections without obvious characteristic cracks, thereby reducing the computational scale of subsequent parameter inversion.

[0122] S32: For the candidate analysis road segment, retrieve the corresponding geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data from the three-level indexed structured database; establish a geometric model of soil layers and subgrade based on the geological survey data and road design data; use the indoor geotechnical parameters as the initial parameters of the soil constitutive model; determine the pile structure parameters, construction stage parameters, and load parameters based on the subgrade treatment design and construction parameters and as-built data; automatically establish the initial finite element calculation model of the candidate analysis road segment and perform calculations.

[0123] Specifically, after determining the candidate road segments for analysis, the system automatically retrieves data on geological boreholes, soil layering, soil layer thickness, groundwater level, indoor geotechnical test parameters, road cross-sectional dimensions, embankment height, soft soil treatment methods, and construction stages from the three-level indexed structured database based on the corresponding road mileage.

[0124] The road section in this embodiment corresponds to a soft soil subgrade with a soft soil layer depth of approximately 24.5m. Prestressed pipe piles and cement mixing piles are used for soft soil foundation treatment.

[0125] Establish roadbed cross-section and stratum geometric model based on road design and as-built data; establish soil layer model according to the actual stratification location of each soil layer based on geological survey data; establish prestressed pipe pile treatment area and cement mixing pile treatment area according to the roadbed treatment range.

[0126] For prestressed concrete pipe piles, input the pile length, pile spacing, and pile material parameters according to the design and as-built data; for cement mixing piles, input the pile diameter, pile spacing, treatment depth, and equivalent material parameters.

[0127] In this embodiment, the prestressed pipe piles are designed with a pile length of 25m and a pile spacing of approximately 2.3m; the cement mixing piles have a pile diameter of approximately 0.5m and a pile spacing of approximately 1.7 to 1.9m.

[0128] For soil materials, the initial material parameters of the finite element model are set based on the compression, rebound, creep, shear strength, and permeability parameters obtained from indoor geotechnical tests. These parameters are used to establish the initial calculation model and are not the final parameters after Bayesian inversion.

[0129] The construction phases were further set up according to the five-level surcharge process in the construction data, and corresponding construction conditions were established based on the filling height, loading time, and preloading duration of each level. The total construction phase time in this embodiment is approximately 390 days.

[0130] After completing the model geometry, initial material parameters, boundary conditions, and construction stage settings, the initial finite element calculation model of the candidate analysis road segment is automatically established through the Plaxis script interface according to the model parameters called from the database. Consolidation and deformation calculations are then initiated to obtain the roadbed settlement calculation results under the initial parameter conditions of engineering survey, design, construction, and as-built data.

[0131] In this embodiment, a two-dimensional plane strain model of the road cross section or longitudinal section is established according to the spatial direction of the crack to be analyzed: the bottom edge of the model simultaneously constrains horizontal and vertical displacement, the two side boundaries constrain horizontal displacement, and the ground surface is set as a free boundary; the groundwater level is set according to the geological survey data; local mesh densification is implemented in the road surface, pile body and the transition area between the two treatment methods, the maximum unit side length in the densified area is not greater than half of the distance between adjacent piles, and medium density mesh is used in other areas; the drainage boundary in the consolidation calculation is set according to the on-site groundwater conditions and drainage board layout.

[0132] The purpose of adopting the above method is to use the actual engineering parameters provided by engineering survey, design and as-built data as the initial physical model for subsequent Bayesian parameter inversion, so that the parameter search is based on the real stratum structure, roadbed structure and construction process, rather than directly fitting the settlement curve through mathematical parameters without engineering constraints.

[0133] S33: Call the settlement monitoring data of the construction and operation periods corresponding to the candidate analysis road section, extract the cumulative settlement, settlement gradient and settlement rate of the filling period, preloading period and operation period, and extract the longitudinal and cross-sectional deformation distribution of the crack location and its adjacent range; compare the measured settlement deformation data with the initial finite element calculation results and the road longitudinal section design elevation, determine the settlement state of the candidate analysis road section according to the preset settlement judgment conditions, and classify it into abnormal settlement road section or normal settlement road section.

[0134] Specifically, such as Figure 8 As shown, based on the road mileage range corresponding to the candidate analysis section, the monitoring dates and settlement values ​​of all settlement monitoring points in the section are retrieved from the three-level indexed structured database, and settlement time series curves are organized according to the filling period, preloading period and operation period respectively.

[0135] For the Each measuring point, at two adjacent monitoring times and The settling rate between them can be expressed as:

[0136] in, and These respectively indicate that the measuring point is at and The cumulative settlement at any given time.

[0137] For adjacent measuring points and Its spatial settlement gradient can be expressed as:

[0138] in, and These represent the cumulative settlement at the two measuring points. It indicates the actual distance between two measuring points along the longitudinal or transverse direction of the road.

[0139] The cumulative settlement, settlement rate, and settlement gradient of the candidate road sections at different construction and operation stages were obtained through the above calculations.

[0140] like Figure 9 As shown, settlement curves of the cross section during the construction period are plotted based on settlement data from measuring points at different transverse locations on the same mileage section, in order to identify whether there is significant uneven settlement in the cross section of the road.

[0141] like Figure 10 As shown, longitudinal settlement curves during the construction period are plotted based on settlement data at different mileage locations along the road, which are used to identify longitudinal differential deformation in different soft soil treatment sections and transitional locations of treatment methods.

[0142] like Figure 11 As shown, the transverse settlement characteristics of the section where the crack is located and its adjacent sections are further extracted, so that the location of the crack can be spatially correlated with the shape of the settlement trough, the settlement gradient and the range of differential deformation at the corresponding location.

[0143] The measured settlement curves were then compared with the settlement calculation results obtained from the initial finite element model of S32, and simultaneously compared with the road design longitudinal section elevation. The design lines reflected are compared.

[0144] The initial finite element calculation results are mainly used to determine whether the physical model established according to the initial parameters of engineering survey, design and completion can reflect the basic settlement development trend of the candidate analysis section, and to provide benchmark calculation results for subsequent S4 to determine the parameters to be inverted and to compare the model fitting effect before and after parameter inversion.

[0145] Settlement status determination conditions are established based on the settlement control requirements adopted in advance for the target road project. The settlement status determination conditions may include at least one or more of the following: cumulative settlement control value, settlement rate control value, and differential deformation or settlement gradient control value between adjacent measuring points.

[0146] For candidate road segments, the measured cumulative settlement, settlement rate, and settlement gradient of the segment are compared with the corresponding control values. When at least one of these values ​​meets the anomaly criteria used for the target road project, the corresponding road segment is marked as an abnormal settlement segment; when none of the settlement deformation indicators meet the anomaly criteria, the corresponding road segment is marked as a normal settlement segment.

[0147] The control values ​​are obtained from the design control requirements, construction and operation monitoring and early warning requirements, or road maintenance and management requirements adopted by the target road project, and are pre-stored in a structured database as judgment parameters.

[0148] By simultaneously analyzing cumulative settlement, settlement rate, and settlement gradient, anomaly judgments can be avoided based solely on a single cumulative settlement value. For example, for road sections with large cumulative settlement but relatively uniform settlement at all measuring points, the longitudinal deformation differences may be small; however, for road sections with less than the largest cumulative settlement, significant differential deformation may still occur in the road structure if the settlement gradient between adjacent sections is large.

[0149] S34: Correlate the crack type, crack morphology, crack spatial distribution, and crack density in the abnormal settlement road sections with the corresponding cumulative settlement, settlement gradient, settlement rate, geological conditions, soft soil treatment methods, and roadbed treatment design and construction parameters to construct an associated sample set of abnormal settlement road sections; at the same time, select normal settlement road sections with the same or similar geological conditions and road structure conditions to construct a control sample set, providing input data for subsequent Bayesian optimization parameter inversion and crack cause analysis.

[0150] Specifically, using road mileage markers as the association field, crack records in abnormal settlement road sections are associated with the corresponding longitudinal section settlement, cross section settlement, cumulative settlement, settlement gradient, and settlement rate.

[0151] Furthermore, geological survey and soft soil treatment data are retrieved based on the same mileage range, so that each anomaly sample simultaneously includes crack characteristics, settlement and deformation characteristics, soil conditions, and soft soil treatment structural parameters.

[0152] For example, for transverse crack samples located in the cement mixing pile treatment area, record the cumulative settlement, longitudinal settlement gradient, cement mixing pile diameter, pile spacing, treatment depth, and corresponding soft soil layer parameters at the corresponding location of the crack.

[0153] For longitudinal or oblique crack samples located near the transition point between prestressed pipe piles and cement mixing piles, record the distance of the crack from the soft soil treatment boundary, as well as the differences in pile parameters and settlement deformation on both sides of the boundary.

[0154] At the same time, road sections with the same or similar soil layer distribution, soft soil layer thickness, road cross-section structure, soft soil treatment method and construction stage conditions as the abnormal settlement road sections were selected from the three-level index structured database and were determined to be in a normal settlement state by S33 as normal control samples.

[0155] Among multiple normal road sections that meet the above conditions, priority should be given to selecting road sections that are adjacent to the abnormal settlement road sections in terms of road space location or have similar engineering conditions, so as to reduce the impact of differences in geological conditions, road structure and construction conditions on subsequent comparative analysis.

[0156] The final result is a crack-settlement correlation sample set that includes at least the following fields: road mileage, crack type, crack geometry, cumulative settlement, settlement gradient, settlement rate, soil parameters, soft soil treatment method, and pile parameters.

[0157] It should be noted that the sample labels established in this step are used to characterize objective states such as "abnormal settlement", "normal settlement", "soft soil treatment type" and "transition position of treatment method", and will not directly mark the cracks as "differential settlement induced" or "stiffness mutation induced" in this step.

[0158] The ultimate cause of the crack needs to be determined by parameter inversion in S4 and full-process calculation using the optimal parameter combination backfill finite element model in S5. The determination is then verified by comparing the simulated deformation, stress response, and measured crack spatial location and settlement deformation characteristics.

[0159] The above-mentioned association method can provide the Bayesian optimization algorithm with the actual settlement target data and engineering parameter search basis corresponding to the abnormal settlement road section. On the other hand, by setting a normal road section control sample, the bias caused by judging the cause based solely on the characteristics of the abnormal cracks can be reduced, thereby improving the reliability of subsequent crack cause analysis.

[0160] S4: Based on the measured settlement deformation data in the associated sample set, a comprehensive loss objective function is constructed to evaluate the difference between the finite element simulation deformation data and the measured deformation data. The prior search interval and physical constraints of the parameters to be inverted are determined based on geological survey data and roadbed treatment design and construction parameters. The tree-structured Parzen estimator sampling strategy under the Optuna framework is used to automatically iteratively search for the parameters to be inverted in the finite element model. The comprehensive loss objective function is used as the basis for parameter optimization to obtain the parameter combination with the optimal fit between the finite element simulation deformation data and the measured deformation data. Specifically, this includes the following sub-steps: S41: Obtain measured settlement deformation data and finite element simulation deformation data for the same road location and the same analysis period in the abnormal settlement road section, perform fitting evaluation on the two, and construct a comprehensive loss objective function through multiple evaluation indicators.

[0161] Specifically, in this embodiment, based on the abnormal settlement road section associated sample set established in S34, calculation nodes corresponding to the spatial positions of actual settlement monitoring points or road surface elevation detection points in the surface nodes of the finite element model are selected, and finite element calculation results are extracted according to the construction stage or operation time corresponding to the measured data.

[0162] Let the measured settlement sequence be:

[0163] The corresponding finite element simulation settlement sequence is:

[0164] in, Indicates the first The first monitoring location or the first The measured settlement value at each monitoring time. This represents the finite element simulation settlement value at the same location or at the same time.

[0165] To avoid evaluation bias caused by using only a single error index, this embodiment uses the coefficient of determination. The simulation results are comprehensively evaluated using structural similarity (SSIM), root mean square error (RMSE), dynamic time warping distance (DTW), maximum absolute error (MaxAE), and mean absolute error (MAE).

[0166] in: Used to evaluate the degree to which finite element calculation results explain the overall variation law of measured settlement; SSIM is used to evaluate the similarity between simulated and measured settlement curves in terms of overall shape and local variation characteristics. RMSE is used to evaluate the overall deviation between simulated and measured values ​​and to improve the impact of large error points on the objective function. DTW is used to evaluate the temporal similarity between simulated and measured settlement development processes when there is a time shift. MaxAE is used to evaluate the maximum simulation deviation at a single most unfavorable location or most unfavorable moment; MAE is used to evaluate the overall mean absolute deviation of simulation results.

[0167] 1. Unified conversion of evaluation indicators because SSIM is an indicator where a larger value indicates a better fit, while RMSE, DTW, MaxAE, and MAE are indicators where a smaller value indicates a better fit. Therefore, we first convert all evaluation indicators into loss components where "a smaller value indicates a better fit".

[0168] for The corresponding loss component can be defined as:

[0169] For SSIM with values ​​normalized to 0 to 1, the following definition can be made:

[0170] For error indices with the dimension of settlement, a characteristic scale of measured settlement data is introduced. Perform dimensionless transformation.

[0171] The To measure the settlement characteristics, this embodiment follows... Calculation, where This is the lower positive limit value; when the settlement value is entered in millimeters, Use 1 mm; when inputting in meters, use 0.001 m to avoid numerical anomalies when the measured settlement change is small. Then:

[0172]

[0173]

[0174] For DTW, considering that its cumulative distance is affected by the matching path length, the DTW distance is further divided by the optimal matching path length of the dynamic time warping. and measured settlement characteristic scale :

[0175] This transforms evaluation indicators with different dimensions and numerical ranges into dimensionless loss components that can participate in weighted calculations.

[0176] 2. Comprehensive Loss Function The comprehensive loss objective function is defined as:

[0177] in:

[0178] All weights are greater than or equal to 0.

[0179] This embodiment can be implemented using: ; ; ; ; ; .

[0180] at this time, The combined weight of SSIM is 0.70. The overall settlement trend and curve shape are used as the main criteria for optimization. At the same time, RMSE, DTW, MaxAE and MAE are used to constrain the overall error, time series offset, local maximum error and average error.

[0181] The weights are not limited to the fixed values ​​mentioned above and can be adjusted according to the road service stage and monitoring objectives.

[0182] For roads with a short operating history and where the soft soil is still in a significant consolidation or creep development stage, it can improve... Corresponding weights are assigned to enhance the fit to the overall settlement trend; For long-term operational roads that require focused identification of local abnormal deformations, the corresponding weight of MaxAE can be increased; For time series data under conditions of continuous rainfall or high-frequency monitoring during the flood season, the corresponding weights of RMSE and DTW can be increased to enhance the identification of short-term subsidence changes and subsidence development time shifts.

[0183] When adjusting the weights of each evaluation indicator according to the road service stage and monitoring objectives, the sum of the adjusted weights of each evaluation indicator remains at 1, and the coefficient of determination... The sum of the weights corresponding to the structural similarity SSIM is not less than 0.70.

[0184] Therefore, this step is not about finding the single parameter with the smallest error, but rather about finding a combination of finite element parameters that can simultaneously and well explain the settlement amplitude, spatial morphology, and temporal development patterns.

[0185] Table 1: Evaluation Indicators and Weights for the Comprehensive Loss Objective Function

[0186] S42: Based on geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, determine the parameters to be inverted in the finite element model and their prior search intervals.

[0187] Specifically, engineering parameters that can be clearly determined through as-built data and have high reliability can be directly input into the finite element model as determined parameters; parameters that have greater uncertainty due to the spatial variability of soil, construction deviations, or changes in long-term service performance are input into the Optuna optimization module as parameters to be inverted.

[0188] In this embodiment, when using the Soft Soil Creep model to describe long-term soft soil deformation, the compression index will be corrected. Corrected rebound index and corrected creep index As constitutive parameters of soil.

[0189] Among them, it is determined based on the results of geological surveys and indoor geotechnical tests. The initial reference value, and establish a system around that initial reference value. The search range.

[0190] To prevent Optuna from generating parameter combinations that do not conform to the physical properties of soft soil during the search process, a proportional parameter is introduced. and And set: ; ; ; ; Therefore, Optuna does not need to be searched entirely independently. , and Three parameters, and searchable , and Then calculate based on the above relationships. and .

[0191] Table 2: Bayesian Parameter Search Table for SSC Soil Model

[0192] The above method can reduce the combination of parameters that have no physical meaning, while retaining the degree of freedom to adjust parameters under different soft soil conditions.

[0193] For roadbed treatment structures, the structural parameters that need to be inverted are determined based on the design data and as-built data.

[0194] The candidate search range given in this embodiment includes:

[0195] For pile lengths and pile spacings that can be reliably determined from as-built data, they can be used as fixed parameters; for situations where there are uncertainties in construction quality, actual pile length, or equivalent stress performance, the corresponding parameters can be set as parameters to be inverted.

[0196] This yields the parameter vector: ; The parameters actually included in the parameter vector can be determined based on the geological conditions of the abnormal settlement section, the form of soft soil treatment, and existing engineering data.

[0197] It is important to emphasize that the parameter search range is not arbitrarily given, but is determined by the engineering data collected by S1 and the spatial correspondence established by S2, so that the Bayesian parameter search is always limited to the parameter space with engineering significance.

[0198] S43: The tree-structured Parzen estimator sampling strategy under the Optuna framework is adopted to generate candidate parameters in the parameter space established in S42 and automatically call the Plaxis finite element model for calculation.

[0199] This embodiment sets the optimization direction to minimize the overall loss, uses a fixed random seed of 2026, and sets the first 20 trials as the starting sampling. Each group of candidate parameters uses an independent working directory and a unique trial number. If the candidate parameters do not meet the physical constraints or the finite element calculation does not converge normally, the failure status is recorded and the penalty loss value is returned. This avoids the parameters from being involved in updating the optimal sample distribution.

[0200] The Plaxis script executes in a fixed order: "load initial model - write candidate parameters - update material and pile properties - start calculation - read vertical displacement of nodes corresponding to measured points - form simulated settlement sequence in the same time order - calculate and return comprehensive loss value". The test number, candidate parameters, calculation status, simulated settlement sequence and comprehensive loss value are written to the iteration record table.

[0201] Specifically, such as Figure 12 As shown, each parameter search iteration includes at least the following process: Optuna generates the first [parameter] based on the current parameter search space. Group candidate parameters ; Check based on physical constraints Does it meet the constraints of soil parameters and subgrade treatment parameters? For candidate parameters that meet the constraints, the corresponding parameters are automatically written into the finite element model established by S32 through the Plaxis script interface; Keep the soil geometry, road structure, boundary conditions, and construction stage settings of the finite element model unchanged, and only update the parameters that need to be inverted this time; Initiate finite element calculation; After the calculation is completed, the simulated settlement value is automatically read from the corresponding location of the measured settlement monitoring point and the corresponding construction stage or operation time. Forming a simulated settlement sequence ; Calculate using the comprehensive loss objective function established by S41: ; Candidate parameters and corresponding comprehensive loss value Return to the Optuna optimization module.

[0202] After completing several initial trials, the tree-structured Parzen estimator divides the parameter samples into better samples and other samples according to the objective function performance based on the obtained candidate parameters and comprehensive loss values, and establishes parameter probability distributions for each sample, thereby determining the candidate parameters with greater optimization potential.

[0203] Therefore, the first After the next iteration, the following is formed: Candidate parameters →Plaxis Finite Element Calculation→Simulation of Settlement →Comprehensive Loss →Optuna updates sampling strategy→Generates the next set of candidate parameters .

[0204] Repeat the above process to gradually reduce the overall loss value.

[0205] With the above coupling method, engineers no longer need to manually modify Plaxis material parameters, recalculate and manually compare settlement curves one by one. Instead, the parameter search module, finite element calculation module and objective function evaluation module form an automatic closed loop.

[0206] S44: Record the parameter combinations and comprehensive loss values ​​for each iteration in real time, and determine whether the parameter search has converged based on the changes in the comprehensive loss objective function.

[0207] Specifically, for the first In the next iteration, the following was recorded:

[0208] And compare it with the current historical best comprehensive loss value.

[0209] when When the overall loss value is less than the historical best value, Update to the current optimal parameter combination.

[0210] In this embodiment, the maximum number of iterations Take 200 as the target value for overall loss. Take 0.05, for the number of consecutive times with no significant improvement. Set the threshold to 20, the minimum improvement threshold. Pick The above values ​​are executable settings for this embodiment and can be adjusted according to the data scale and computing resources.

[0211] The parameter search ends when any of the following termination conditions are met: Reaching the preset maximum number of iterations ; The current optimal overall loss value has reached the preset target value. ; continuous The improvement in the optimal comprehensive loss value in each iteration is less than .

[0212] After the parameter search is completed, extract the parameter combination with the minimum overall loss value from all completed experiments:

[0213] and output The corresponding soil constitutive parameters, subgrade treatment structural parameters, comprehensive loss value, individual fitting evaluation indexes, and complete parameter iteration records.

[0214] like Figure 13 As shown, during the parameter search process, the simulated settlement curves under different candidate pile equivalent parameter conditions are output and compared with the measured settlement points at the same monitoring time.

[0215] Figure 13 (a) Figure 13 (b) and Figure 13 (c) represent the equivalent parameters of the pile body. , and Simulated settlement values ​​and measured settlement values ​​under the given conditions; the comprehensive loss objective function calculates the loss value based on the trend, shape and error between each candidate curve and the measured data, and updates the subsequent parameter sampling distribution accordingly.

[0216] like Figure 14As shown, after backfilling the finite element model with the determined optimal parameter combination, the calculated settlement value at the road center is consistent with the observed settlement value in the main development stage, which is used to verify the fitting effect of the inversion parameters on the settlement time history.

[0217] The final optimal parameter combination is passed to the subsequent S5 and backfilled into the initial finite element calculation model established in S32. It is used to carry out the calculation of the entire process from the road construction period to the long-term service period, as well as the analysis of the main causes of cracks and the safety status of the road structure.

[0218] S5: Backfill the optimal parameter combination into the initial finite element calculation model established in S3 to simulate the roadbed filling, preloading, operation, and long-term service processes, and obtain the deformation, stress, and excess pore water pressure response data of the roadbed and road structure; spatially correlate and compare the response data with the location, type, and distribution characteristics of measured cracks on the road surface to determine the correspondence between abnormal settlement and road surface cracks, judge the safety status of the road structure, identify the main influencing factors related to the formation of road surface cracks, and output the road structure safety diagnosis results and risk assessment results; specifically including the following sub-steps: S51: Write the optimal parameter combination obtained in S4 into the corresponding parameters to be inverted in the initial finite element calculation model established in S32, forming a road finite element calculation model after inversion calibration based on measured settlement data.

[0219] Specifically, S32 has completed the setting of soil structure, road subgrade structure, soft soil treatment form, calculation boundary conditions and construction stage of the target road section based on the geological survey data, indoor geotechnical test data, road design data, subgrade treatment design and construction parameters and as-built data corresponding to the candidate analysis road section, and has completed the initial finite element calculation using the initial parameters in the engineering data.

[0220] In S4, the optimal parameter combination that minimizes the overall loss is obtained through automatic iteration between the tree-structured Parzen estimator sampling strategy under the Optuna framework and the finite element calculation model. .

[0221] This step combines the optimal parameters. Write the corresponding material parameters or roadbed treatment structure parameters into the initial finite element calculation model. For model parameters that can be reliably determined from design data, construction records, or as-built data and are not used as objects to be inverted, keep their original values ​​unchanged.

[0222] For the soft soil layer in this embodiment, the Soft Soil Creep (SSC) constitutive model is used to describe the long-term consolidation and creep deformation of the soft soil, and the corrected compression index obtained by S4 inversion is used. Corrected rebound index and corrected creep index The optimal parameters are then written into the corresponding soft soil material model.

[0223] For soil layers that are determined to be described by the Hardening Soil (HS) constitutive model based on engineering data, the corresponding material parameters to be inverted are written into the corresponding soil material model according to the optimal parameter combination obtained in S4.

[0224] For road sections using piles for soft soil foundation treatment, when the relevant calculation parameters of the piles are set as parameters to be inverted, the corresponding optimal parameters obtained in S4 are written into the pile calculation unit in the finite element model; for pile parameters that can be reliably determined from the as-built data, the corresponding as-built parameters are directly used.

[0225] In this embodiment, the pile body is simulated using embedded pile body units; for road sections with drainage boards, drainage lines are used to simulate the drainage structure; the pavement water-stabilized layer is set as a Mohr-Coulomb material model or a linear material model according to the actual calculation model used; the asphalt layer adopts a preset pavement structure material model.

[0226] By calling the Plaxis finite element calculation interface through a Python script, the optimal parameter combination is automatically written into the corresponding material model or structural element, and saved to form a finite element calculation model after inversion and calibration based on measured settlement data.

[0227] It should be noted that S32 establishes a finite element calculation model based on the initial engineering parameters in geological survey, design and construction, and as-built data, which is used to provide the basic physical model required for Bayesian parameter inversion; S51 uses the optimal parameter combination obtained in S4 to update the uncertain parameters, so that the finite element model can better reflect the measured settlement development law of the target road section, and provide a calibrated calculation model for the structural response calculation of the entire process from the road construction period to the long-term service period.

[0228] S52: Based on the actual construction and service process of the target road, the subgrade grading, preloading, operation and long-term service stages are simulated in the finite element calculation model after inversion calibration, and consolidation and stress-deformation calculations are performed to obtain the road and subgrade structure response corresponding to each stage.

[0229] Specifically, based on the roadbed construction data and as-built data stored in S13, a finite element calculation stage is established according to the actual construction time sequence of the road.

[0230] For the five-stage surcharge construction process adopted in this embodiment, the embankment filling height, corresponding filling load and loading duration for each stage are set according to the construction records; after the staged filling is completed, the preloading stage and corresponding preloading duration are set according to the construction data; after the preloading stage is completed, the road operation stage and long-term service calculation stage are further set.

[0231] During the operation phase, equivalent loads corresponding to the road's operating status can be set based on the target road's design traffic conditions or actual operating data, so that the operating conditions in the finite element model correspond to the actual service conditions of the target road.

[0232] Soil consolidation and stress-deformation calculations are performed in each calculation stage to obtain calculation results of changes in road and subgrade structures over construction and service time.

[0233] The calculation results include at least: Vertical deformation of the road surface and subgrade; Longitudinal and transverse differential deformation of the road; Stress distribution in soil and road structures; Excess pore water pressure and its change over time; Pile deformation; Simulated settlement data corresponding to different road mileages and different cross-sectional locations.

[0234] The finite element calculation results obtained at different construction and service stages are saved in the manner of "road mileage - cross section position - calculation time", and a spatial and temporal correspondence is established with the three-level index structured database of "road segment - cross section - measuring point" established by S2.

[0235] For example, for a certain road mileage location, the simulated road surface settlement value at that location can be extracted during the roadbed graded filling stage, preloading stage, initial operation stage, and long-term service stage, and a corresponding relationship can be established with the settlement monitoring data during the construction period, the settlement monitoring data during the operation period, and the road surface elevation change data of the same road mileage location.

[0236] By adopting the above-mentioned full-process simulation method, we can not only obtain the final settlement value of the target road at a certain calculation time, but also obtain the development law of deformation, stress and excess pore water pressure over time during the road's process from graded filling, preloading and consolidation to operation and long-term service. This provides a calculation basis for subsequent judgment on whether there is a corresponding relationship between the current road surface cracks and the long-term settlement and deformation of the roadbed.

[0237] S53: Based on the road mileage station and cross-sectional spatial location, spatially match the finite element simulation results obtained in S52 with the measured location, type, and distribution range of road surface cracks. Extract the road surface settlement, settlement gradient, structural deformation, stress, excess pore water pressure, and pile deformation characteristics corresponding to each crack location and adjacent area. Based on the consistency between the simulated deformation characteristics and the measured settlement and crack distribution, determine whether the road surface cracks are related to abnormal subgrade settlement. Combine the geological conditions at the location of the cracks and the subgrade treatment parameters to identify the main influencing factors of abnormal settlement.

[0238] Specifically, using the unified spatial index of road mileage established by S2, the road mileage station number and cross-sectional position corresponding to each measured crack are mapped to the inverted and calibrated finite element calculation model formed by S51.

[0239] For the For a measured crack, based on its corresponding road mileage and cross-sectional location, structural response data of the crack location and its adjacent area are extracted from the finite element analysis results. The structural response data includes at least: The simulated cumulative settlement at this location; Longitudinal settlement gradient of the road; Lateral differential settlement of the road; Deformation distribution of roads and roadbed structures; Stress response of soil and road structures; Changes in excess pore water pressure; Pile deformation; And the corresponding soft soil treatment methods and geological conditions at this location.

[0240] Subsequently, the above finite element simulation results were compared with the measured cumulative settlement, settlement rate, settlement gradient, and crack spatial distribution extracted from S3.

[0241] The differential deformation concentration areas calculated by the finite element model and the measured crack concentration areas are spatially compared according to road mileage and cross-sectional location, and a comprehensive analysis is carried out in combination with the fitting results between the finite element simulated settlement and the measured settlement obtained by S4.

[0242] When the finite element model after inversion calibration can reproduce the development law of the measured settlement well, and when the simulated differential deformation concentration area and the measured crack concentration area have a clear spatial correspondence in terms of road mileage and cross-sectional location, it is determined that the corresponding cracks are strongly correlated with the abnormal settlement of the roadbed.

[0243] Further, based on the road mileage where the crack is located, the geological conditions, soft soil treatment methods, and roadbed treatment design and construction parameters of the corresponding location are retrieved from the three-level indexed structured database, and the differences in settlement and structural deformation of different soft soil treatment sections and transition points of soft soil treatment methods are compared.

[0244] In this embodiment, the following can be compared respectively: Cumulative settlement and settlement gradient within the cement mixing pile treatment area; Cumulative settlement and settlement gradient within the prestressed pipe pile treatment area; The settlement difference between the two sides of the transition between the cement mixing pile treatment area and the prestressed pipe pile treatment area; Spatial distance and correspondence between the locations of concentrated crack distribution and the boundaries of different soft soil treatment methods; Variations in soil conditions, pile parameters, and other roadbed treatment parameters corresponding to different soft soil treatment sections.

[0245] When the locations of the measured crack concentration, the measured differential settlement concentration, and the finite element simulation differential deformation concentration areas show a high degree of consistency in spatial location, further analysis of the geological conditions and roadbed treatment parameter changes at the corresponding locations can help identify engineering influencing factors that are highly correlated with abnormal settlement.

[0246] For example, by comparing the simulated and measured settlement distributions of cement mixing pile treatment areas, prestressed pipe pile treatment areas, and transition areas between the two treatment methods, the correspondence between different soft soil treatment methods and their parameter changes and the concentrated distribution of cracks and differential settlement can be analyzed, and the geological or subgrade treatment factors that have a significant impact on the abnormal settlement of the target road section can be identified.

[0247] It should be noted that the determination of the main causes of cracks in this step is not based solely on directly assigning a cause label according to the crack type. Instead, it is based on the finite element model being calibrated by inversion of measured settlement data and able to reflect the actual deformation law of the target road section. Then, it is combined with a comprehensive analysis of the spatial correspondence between measured cracks, measured settlement, simulated deformation, geological conditions, and soft soil treatment parameters.

[0248] like Figure 15 As shown, Figure 15 This represents the road and subgrade structural response results obtained using a finite element model calibrated through parametric inversion. These results can be used to identify areas of relatively concentrated deformation within the target analysis section and to extract simulated settlement, stress, and other structural response data for the corresponding road mileage and cross-sectional location.

[0249] like Figure 16 As shown, Figure 16 This represents the correspondence analysis results between the finite element method (FEM) simulation deformation results and the spatial distribution of road surface cracks. By spatially comparing the concentrated areas of differential deformation obtained from the FEM simulation with the measured locations of road surface cracks, the correlation between abnormal settlement and road surface cracks can be analyzed, and a calculation basis can be provided for identifying the main influencing factors of road surface cracks.

[0250] S54: Based on the crack-deformation correspondence, main influencing factors of abnormal settlement, and finite element calculation results of road structure obtained in S53, generate standardized road structure safety diagnosis results including the location of abnormal settlement road sections, crack characteristics, settlement and deformation characteristics, finite element calculation results, main causes of cracks, road structure safety status, and risk assessment information.

[0251] Specifically, for each abnormal settlement road segment identified by S33 and further analyzed by S5, a corresponding road structure safety analysis record is generated.

[0252] The road structure safety analysis record includes at least: (1) Basic information of the road section, including the road mileage range, road structure, main geological conditions and soft soil treatment methods; (2) Crack characteristic information, including the road mileage where the crack is located, the cross-sectional location, the crack type, the crack length, the crack width, and the crack distribution range; (3) Measured settlement deformation information, including cumulative settlement, settlement rate, longitudinal settlement gradient, lateral differential settlement and corresponding settlement time sequence characteristics; (4) Bayesian parameter inversion information, including the optimal parameter combination The corresponding comprehensive loss value, various fitting evaluation indicators, and the fitting results between finite element simulation settlement and measured settlement; (5) Finite element structural response information, including road and subgrade structural deformation, stress, excess pore water pressure, pile deformation, and the spatial distribution and temporal variation of the corresponding structural response; (6) Results of crack-abnormal settlement correlation analysis, including the spatial correspondence between crack location and simulated differential deformation concentration area, whether crack is significantly associated with abnormal settlement, and geological conditions or roadbed treatment engineering parameters that are highly associated with abnormal settlement. (7) Road structure safety status and risk assessment information.

[0253] For the risk status of road structure, corresponding risk judgment conditions are set according to the cumulative settlement, differential settlement, settlement rate and other road structure safety control indicators adopted in advance for the target road project, and the judgment conditions are stored in the structured database as risk analysis parameters in advance.

[0254] When conducting risk analysis, the measured settlement deformation index of the road section with abnormal settlement, the finite element structural response after inversion calibration, and the results of crack-abnormal settlement correlation analysis are compared with the risk judgment conditions to generate corresponding road structure risk assessment information.

[0255] When the target road project has established risk level classification rules, the road structure risk assessment results can be mapped to the corresponding risk level according to the preset risk level judgment rules; when the target road project only specifies the corresponding structural safety control indicators, the analysis results of whether each structural safety indicator meets the corresponding control conditions are output.

[0256] Taking an abnormal settlement section in this embodiment as an example, the generated road structure safety diagnosis results may include: Transverse, longitudinal, or diagonal cracks were detected in the target road section. Based on multiple periods of road surface elevation data and historical settlement monitoring data during the construction and operation periods, differential deformation was found near the cracks. The finite element calculation model, after Bayesian parameter inversion calibration, can well reflect the development law of the measured settlement of the road section. The concentrated area of ​​differential deformation obtained by the finite element calculation has a spatial correspondence with the concentrated distribution location of the measured cracks. Further, combined with the geological conditions at the location of the cracks, the soft soil treatment method, and the changes in engineering parameters of adjacent sections, the main engineering influencing factors with a high degree of correlation with the abnormal settlement were identified. The above-mentioned measured deformation, finite element calculation results, and analysis results of the main influencing factors were jointly incorporated into the road structure safety diagnosis and risk assessment.

[0257] Through the above methods, starting from road surface crack and settlement monitoring data, and through multi-source engineering data association, finite element parameter inversion, full-process numerical simulation, and crack-deformation space comparison, the road structure safety diagnosis results and risk assessment information corresponding to the target abnormal settlement road section are formed.

[0258] The generated road structure safety diagnosis results can serve as the basis for analysis in road operation and maintenance, further on-site verification, intensified monitoring, and the formulation of subsequent maintenance and treatment plans.

[0259] Corresponding to the aforementioned method embodiment of automated analysis of road structure safety using Bayesian algorithm coupled with finite element method, this application also provides a system for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method.

[0260] Figure 17 This is a system block diagram illustrating an automated analysis of road structure safety using a Bayesian algorithm coupled with finite element methods, according to an exemplary embodiment. Figure 17 As shown, the system includes a data acquisition and raw database construction module 1, a spatiotemporal registration and structured database construction module 2, a candidate analysis road segment screening and initial finite element analysis module 3, a Bayesian optimization parameter inversion module 4, and a finite element calibration and road structure safety analysis module 5, which sequentially perform data interaction.

[0261] The data acquisition and raw database construction module 1 is used to collect and integrate road surface crack images, road surface elevation data at different detection periods, settlement monitoring data during construction and operation, geological survey data, indoor geotechnical physical and mechanical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, and store them according to unified data fields to build a full-dimensional raw database.

[0262] Specifically, the data acquisition and raw database construction module 1 can acquire road surface crack images and road surface elevation data with acquisition time and spatial location information through a vehicle-mounted detection system that integrates a high-definition camera, a GNSS receiver and an IMU inertial measurement unit, and write the vehicle-mounted detection data, construction period and operation period settlement monitoring data and existing road engineering data into a structured data table according to unified data fields.

[0263] The existing data for the road engineering project includes at least geological survey data, indoor geotechnical test data, road design data, subgrade treatment design and construction data, and as-built data. The structured data table records at least the data source, road mileage marker, spatial coordinates, acquisition time, data type, and data value, enabling detection data, monitoring data, and engineering data from different sources to be stored and retrieved using a unified data structure.

[0264] The spatiotemporal registration and structured database construction module 2 is used to perform data cleaning, coordinate transformation, error correction, and spatial registration on the road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database. It unifies data from different sources to the geodetic coordinate system used in road design, and matches the crack locations, elevation measurement points, and settlement monitoring points with the road design horizontal alignment and mileage station numbers. Using the road mileage station numbers as a unified spatial index, a three-level indexed structured database of "road segment - cross section - measurement point" is constructed.

[0265] Specifically, the spatiotemporal registration and structured database construction module 2 performs image preprocessing and crack feature recognition on road surface crack images, extracting crack type, crack length, crack width, and crack spatial location; and performs outlier removal, filtering and noise reduction, and error correction on road surface elevation data from different detection periods to form road surface elevation change data and road deformation time series data.

[0266] The spatiotemporal registration and structured database construction module 2 also converts the vehicle-mounted detection data to the geodetic coordinate system used in the road design based on the coordinate reference system used in the road design data, and segments the road detection data according to the bridgehead, bridge tail or engineering segment boundary, and performs plane position calibration and elevation adjustment correction on the detection points.

[0267] After completing the spatial location correction, the crack location, elevation measurement point and settlement monitoring point are matched with the road design plane alignment to determine the road mileage station number corresponding to each data point. Then, according to the road mileage station number, the crack characteristics, elevation changes, settlement monitoring, geological conditions, road design and subgrade treatment design and construction data are linked to form a three-level indexed structured database organized according to "road section - cross section - measurement point".

[0268] The candidate analysis segment screening and initial finite element analysis module 3 is used to screen candidate analysis segments based on the three-level indexed structured database, according to the type of road surface cracks, the spatial location of cracks, and the corresponding settlement deformation characteristics; it calls the geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate analysis segments to establish an initial finite element calculation model for the candidate analysis segments, and uses the parameters in the geological survey and design and construction data as the initial parameters for the finite element model for calculation; it combines the measured settlement deformation data with the initial finite element calculation results to determine abnormal settlement segments, and constructs a correlation sample set between crack characteristics, settlement deformation characteristics, geological conditions, and subgrade treatment parameters.

[0269] Specifically, the candidate analysis section screening and initial finite element analysis module 3 retrieves road surface elevation change data and settlement monitoring data during construction and operation from the three-level indexed structured database based on crack type, crack spatial location and corresponding road mileage, and determines the candidate analysis sections that need to be analyzed by finite element analysis according to preset crack characteristics.

[0270] For the candidate road sections, an initial finite element calculation model is established based on the corresponding stratum distribution, road cross-sectional dimensions, soil physical and mechanical parameters, soft soil treatment methods, pile parameters, construction stages, and load parameters. The initial parameters from engineering survey, design, construction, and as-built data are then used for calculation.

[0271] The candidate analysis segment screening and initial finite element analysis module 3 further extracts the cumulative settlement, settlement rate, settlement gradient, and longitudinal and lateral deformation distribution of the candidate analysis segments, and classifies the candidate analysis segments into abnormal settlement segments or normal settlement segments according to the preset settlement judgment conditions.

[0272] For road sections with abnormal settlement, crack characteristics, settlement deformation characteristics, geological conditions, and roadbed treatment parameters are correlated to construct a correlation sample set for abnormal settlement road sections. At the same time, normal settlement road sections with the same or similar engineering conditions are selected to construct a control sample set for normal settlement road sections, providing measured target data and engineering parameter basis for subsequent Bayesian parameter inversion.

[0273] The Bayesian optimization parameter inversion module 4 is used to construct a comprehensive loss objective function based on the measured settlement deformation data in the associated sample set to evaluate the difference between the finite element simulation deformation data and the measured deformation data. Based on the geological survey data and the roadbed treatment design and construction parameters, it determines the prior search interval and physical constraints of the parameters to be inverted. The tree structure Parzen estimator sampling strategy under the Optuna framework is used to automatically iteratively search for the parameters to be inverted in the finite element model. The comprehensive loss objective function is used as the basis for parameter optimization to obtain the parameter combination with the best fit between the finite element simulation deformation data and the measured deformation data.

[0274] Specifically, the Bayesian optimization parameter inversion module 4 acquires measured settlement data and finite element simulation settlement data corresponding to the same road location and the same analysis period, and uses at least the coefficient of determination. The simulation results are evaluated using multiple metrics, including Dynamic Time Warped Distance (DTW).

[0275] In some embodiments, the evaluation metrics further include one or more of structural similarity (SSIM), root mean square error (RMSE), maximum absolute error (MaxAE), and mean absolute error (MAE). Each evaluation metric is converted into a dimensionless loss component, where smaller values ​​indicate better fitting, and then weighted according to preset weights to form a comprehensive loss objective function.

[0276] The Bayesian optimization parameter inversion module 4 determines the uncertain parameters to be inverted and the search range corresponding to each parameter based on geological surveys, indoor geotechnical tests, and subgrade treatment design and construction data. It also establishes physical constraint conditions based on the physical relationships between soil parameters and the engineering constraints between subgrade treatment structural parameters.

[0277] During each parameter search iteration, the tree-structured Parzen estimator sampling strategy under the Optuna framework generates candidate parameters within the preset search space. For candidate parameters that meet the physical constraints, they are automatically written into the initial finite element calculation model established by module 3 through the finite element script interface and the finite element calculation is started. After the calculation is completed, simulated settlement data at the same road location and the same analysis period as the measured settlement data are automatically extracted, and the comprehensive loss value corresponding to the current candidate parameter is calculated according to the comprehensive loss objective function.

[0278] The current candidate parameters and their corresponding comprehensive loss values ​​are fed back to the Optuna optimization module. Based on the multiple sets of candidate parameters and comprehensive loss values ​​that have been completed, the subsequent parameter sampling distribution is updated. The processes of candidate parameter generation, finite element calculation, simulation result extraction, and comprehensive loss value calculation are repeated until the comprehensive loss objective function meets the preset convergence condition or reaches the preset maximum number of iterations, thereby obtaining the optimal parameter combination with the minimum comprehensive loss value.

[0279] The finite element calibration and road structure safety analysis module 5 is used to backfill the optimal parameter combination into the initial finite element calculation model, simulate the roadbed filling, preloading, operation and long-term service process, and obtain the deformation, stress and excess pore water pressure response data of the roadbed and road structure; spatially correlate and compare the response data with the location, type and distribution characteristics of the measured cracks on the road surface to determine the correspondence between abnormal settlement and road surface cracks, judge the safety status of the road structure, identify the main influencing factors related to the formation of road surface cracks, and output the road structure safety diagnosis results and risk assessment results.

[0280] Specifically, while maintaining the model geometry, road structure, boundary conditions, and construction stage settings determined by engineering data, the optimal parameter combination is written into the corresponding soil material parameters or subgrade treatment structure parameters to form a finite element calculation model calibrated by measured settlement data.

[0281] Based on the actual construction and service process of the road, the finite element calculation model after inversion and calibration is used to simulate the stages of graded filling, preloading, operation and long-term service of the roadbed in sequence, and to perform consolidation and stress-deformation calculations to obtain the calculation results of vertical deformation, longitudinal and transverse differential deformation, stress, excess pore water pressure and pile deformation of the road and roadbed structure.

[0282] The Finite Element Calibration and Road Structure Safety Analysis Module 5 spatially matches the finite element calculation results with the measured location, type, and distribution range of road surface cracks according to the road mileage station and cross-sectional location. It also extracts the simulated settlement, settlement gradient, structural deformation, stress, excess pore water pressure, and pile deformation characteristics corresponding to the crack location and adjacent areas.

[0283] Based on the consistency between the deformation characteristics simulated by finite element method and the measured settlement deformation data and the spatial distribution of cracks, it is determined whether the surface cracks are significantly related to the abnormal settlement of the subgrade. Furthermore, by combining the geological conditions corresponding to the crack location, the soft soil treatment method and the subgrade treatment design and construction parameters, the main engineering influencing factors with a high degree of correlation with abnormal settlement are identified.

[0284] After completing the above analysis, the Finite Element Calibration and Road Structure Safety Analysis Module 5 generates standardized road structure safety diagnosis results, including the location of abnormal settlement road sections, crack characteristics, measured settlement deformation characteristics, optimal parameter combination, finite element structural response, crack-abnormal settlement correlation analysis results, main influencing factors, road structure safety status, and risk assessment information.

[0285] The specific methods of operation performed by each module in the above system embodiments have been described in detail in the aforementioned method embodiments S1 to S5. The same or corresponding technical content can be referred to the aforementioned method embodiments, and will not be repeated here.

[0286] It should be noted that the aforementioned data acquisition and raw database construction module 1, spatiotemporal registration and structured database construction module 2, candidate analysis road segment screening and initial finite element analysis module 3, Bayesian optimization parameter inversion module 4, and finite element calibration and road structure safety analysis module 5 can be implemented through software programs, hardware devices, or a combination of software and hardware. Each module can be implemented as an independent functional unit, or it can be integrated into one or more computing devices according to the actual system deployment requirements; this application does not impose any limitations on this.

[0287] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0288] Accordingly, this application also provides an electronic device, including one or more processors and a memory, the memory being used to store one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors enable the above-described method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method.

[0289] Accordingly, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for automated analysis of road structure safety using a Bayesian algorithm coupled with finite element method.

[0290] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.

[0291] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A method for automated analysis of road structure safety using a Bayesian algorithm coupled with finite element method, characterized in that, include: S1: Collect and integrate road surface crack images, road surface elevation data at different detection periods, settlement monitoring data during construction and operation, geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, store them according to unified data fields, and build a full-dimensional original database; S2: Perform data cleaning, coordinate transformation, error correction, and spatial registration on the road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database. Unify the data from different sources to the geodetic coordinate system used in the road design, and match the crack locations, elevation measurement points, and settlement monitoring points with the road design horizontal alignment and mileage station numbers. Using the road mileage station numbers as a unified spatial index, construct a three-level indexed structured database of "road segment - cross section - measurement point". S3: Based on the three-level indexed structured database, candidate road sections are selected for analysis according to the type of road surface cracks, the spatial location of cracks, and the corresponding settlement deformation characteristics; the geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate road sections are called to establish an initial finite element calculation model for the candidate road sections, and the parameters in the geological survey and design and construction data are used as the initial parameters of the finite element model for calculation; the abnormal settlement road sections are determined by combining the measured settlement deformation data and the initial finite element calculation results, and a correlation sample set between crack characteristics, settlement deformation characteristics, geological conditions, and subgrade treatment parameters is constructed; S4: Based on the measured settlement and deformation data in the associated sample set, a comprehensive loss objective function is constructed to evaluate the difference between the finite element simulation deformation data and the measured deformation data. Based on the geological survey data and the roadbed treatment design and construction parameters, the prior search interval and physical constraints of the parameters to be inverted are determined. The tree structure Parzen estimator sampling strategy under the Optuna framework is used to automatically iteratively search for the parameters to be inverted in the finite element model. The comprehensive loss objective function is used as the basis for parameter optimization to obtain the parameter combination with the best fit between the finite element simulation deformation data and the measured deformation data. S5: Backfill the optimal parameter combination into the initial finite element calculation model to simulate the roadbed filling, preloading, operation and long-term service process, and obtain the deformation, stress and excess pore water pressure response data of the roadbed and road structure; spatially correlate and compare the response data with the location, type and distribution characteristics of the measured cracks on the road surface to determine the correspondence between abnormal settlement and road surface cracks, judge the safety status of the road structure, identify the main influencing factors related to the formation of road surface cracks, and output the road structure safety diagnosis results and risk assessment results.

2. The method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method according to claim 1, characterized in that, The process involves collecting and integrating road surface crack images, road surface elevation data from different detection periods, settlement monitoring data from the construction and operation periods, geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data to construct a comprehensive original database, specifically including: S11: Establish a structured data table for storing road engineering data from different sources, and set fields for data source, road mileage station number, spatial coordinates, collection time, data type and data value according to data category; each row corresponds to one data record, and each column corresponds to one attribute field, so as to realize standardized storage, index management, related query and unified access of data from different sources; S12: An on-board detection system integrating a high-definition camera, GNSS receiver and IMU inertial measurement unit is used to collect road surface crack images and road surface elevation data with timestamps and spatial location information along the target road, and write the road surface crack images, road surface elevation data, corresponding collection time and spatial location information into the structured data table; S13: Collect settlement monitoring data, geological survey data, indoor geotechnical physical and mechanical test data, road design data, subgrade treatment design and construction data, and as-built data corresponding to the target road during the construction and operation periods. Extract soil layer distribution and soil physical and mechanical parameters, road horizontal and vertical alignment and mileage information, subgrade cross-section and structural dimensions, soft soil treatment form and treatment range, pile geometry and material parameters, embankment graded filling and preloading construction parameters, and actual engineering parameters after completion. Record the geodetic coordinate system used in the road design data. Write the extracted data into the structured data table according to the corresponding road mileage, cross-section location, and data category to form a full-dimensional original database for subsequent spatiotemporal registration, finite element modeling, and parameter inversion.

3. The method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method according to claim 1, characterized in that, The road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database are cleaned, coordinate transformed, error corrected, and spatially registered. Data from different sources are unified to the geodetic coordinate system used in road design. Crack locations, elevation measurement points, and settlement monitoring points are matched with the road design horizontal alignment and mileage markers. Using road mileage markers as a unified spatial index, a three-level indexed structured database of "road segment - cross-section - measurement point" is constructed, specifically including: S21: Perform image preprocessing and crack feature recognition on the road surface crack image. Use an edge recognition algorithm combined with morphological operations to remove image noise and extract crack contours to obtain crack length, crack width, crack morphology, acquisition time and spatial location information. The crack morphology includes at least longitudinal cracks, transverse cracks and diagonal cracks to form a crack feature dataset with spatial location information. S22: Outlier removal, filtering and noise reduction, and system error correction are performed on the road surface elevation data collected at different detection periods to obtain continuous road surface elevation curves for each detection period; based on the elevation changes of the same road location at different detection periods, the corresponding road surface elevation change is obtained, and road deformation time series data is formed by combining the settlement monitoring data during the construction and operation periods. S23: Based on the coordinate reference system used in the road design data, perform coordinate transformation on the spatial location of the crack obtained in S21 and the spatial location of the elevation measurement point obtained in S22, and transform the vehicle detection data from the acquisition coordinate system to the geodetic coordinate system used in the road design; at the same time, extract the horizontal curve and longitudinal profile design elevation data of the road design, so that the crack measurement point, elevation measurement point and road design data are under a unified spatial coordinate reference. S24: Determine the locations of the bridgeheads and tails in the target road based on the road design data, and divide the road into segments using adjacent bridgeheads or tails as road segment boundaries; perform plane position calibration on the detection points in each segment, and use the stable elevation position of the bridgehead or tail as the elevation benchmark point; perform segmented adjustment correction on the road surface elevation data in each segment to obtain detection point data that completes the plane position and elevation position correction; S25: Match the locations of the corrected cracks, elevation measurement points, and settlement monitoring points with the road design horizontal alignment to determine the corresponding road mileage station numbers; use the road mileage station numbers as a unified spatial index to link and integrate crack characteristic data, road surface elevation change data, settlement monitoring data during construction and operation, geological survey data, road design data, roadbed treatment design and construction parameters, and as-built data to form a three-level indexed structured database organized according to "road section - cross section - measurement point".

4. The method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method according to claim 1, characterized in that, Based on the aforementioned three-level indexed structured database, candidate road sections are selected for analysis according to the type of road surface cracks, their spatial location, and corresponding settlement and deformation characteristics. The geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate road sections are used to establish an initial finite element calculation model for each candidate road section. Parameters from the geological survey and design / construction data are used as the initial parameters for the finite element model. Abnormal settlement road sections are identified by combining measured settlement and deformation data with the initial finite element calculation results. A correlation sample set between crack characteristics, settlement and deformation characteristics, geological conditions, and subgrade treatment parameters is constructed, specifically including: S31: Based on the crack type, crack spatial coordinates, road mileage markers, and corresponding road surface elevation change data and settlement monitoring data in the three-level indexed structured database, candidate analysis road sections with preset crack characteristics are selected; wherein, the candidate analysis road sections include road sections with diagonal cracks, road sections with transverse through cracks, and road sections with longitudinal cracks; for road sections with diagonal cracks or longitudinal cracks, a predetermined distance is selected before and after the location of the corresponding crack as the center. The scope is defined as the finite element analysis scope; for road sections with transverse through cracks, the segment where the transverse through crack is located is defined as the finite element analysis scope. S32: For the candidate analysis road segment, retrieve the corresponding geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data from the three-level indexed structured database; establish a geometric model of soil layers and subgrade based on the geological survey data and road design data; use the indoor geotechnical parameters as the initial parameters of the soil constitutive model; determine the pile structure parameters, construction stage parameters, and load parameters based on the subgrade treatment design and construction parameters and as-built data; automatically establish the initial finite element calculation model of the candidate analysis road segment and perform calculations. S33: Call the settlement monitoring data of the construction period and operation period corresponding to the candidate analysis road section, extract the cumulative settlement, settlement gradient and settlement rate of the filling period, preloading period and operation period, and extract the longitudinal and cross-sectional deformation distribution of the crack location and its adjacent range; compare the measured settlement deformation data with the initial finite element calculation results and the road longitudinal section design elevation, determine the settlement state of the candidate analysis road section according to the preset settlement judgment conditions, and classify it into abnormal settlement road section or normal settlement road section; S34: Correlate the crack type, crack morphology, crack spatial distribution, and crack density in the abnormal settlement road section with the corresponding cumulative settlement, settlement gradient, settlement rate, geological conditions, soft soil treatment method, and roadbed treatment design and construction parameters to construct an associated sample set of abnormal settlement road sections.

5. The method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method according to claim 1, characterized in that, Based on the measured settlement and deformation data in the associated sample set, a comprehensive loss objective function is constructed to evaluate the difference between the finite element simulation deformation data and the measured deformation data. The prior search interval and physical constraints of the parameters to be inverted are determined based on geological survey data and roadbed treatment design and construction parameters. The tree-structured Parzen estimator sampling strategy under the Optuna framework is used to automatically iteratively search for the parameters to be inverted in the finite element model. Using the comprehensive loss objective function as the basis for parameter optimization, the optimal parameter combination for fitting the finite element simulation deformation data and the measured deformation data is obtained, specifically including: S41: Obtain measured settlement deformation data and finite element simulation deformation data for the same road location and analysis period in the abnormal settlement road section, using at least the coefficient of determination. The evaluation metrics of dynamic time warp distance (DTW) are used to evaluate the goodness of fit between the two, and the different evaluation metrics are dimensionless and weighted to establish a comprehensive loss objective function with the goal of minimizing the comprehensive loss value. S42: Based on the geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the abnormal settlement road section, determine the parameters to be inverted in the finite element model and the prior search interval of each parameter to be inverted, and establish parameter constraint conditions based on the physical relationship between soil parameters and the engineering constraint relationship of the subgrade treatment structure. S43: Using the tree-structured Parzen estimator sampling strategy under the Optuna framework, a set of candidate parameters satisfying physical constraints is generated within the prior search interval. The candidate parameters are automatically written into the finite element calculation model of the corresponding abnormal settlement section and the finite element calculation is started. Simulated settlement data corresponding to the measured settlement data and the corresponding analysis period are extracted. The comprehensive loss value corresponding to the current candidate parameter is calculated according to the comprehensive loss objective function. The candidate parameter and its comprehensive loss value are returned to the Optuna optimization module. The sampling distribution of subsequent parameters is updated according to the multiple sets of candidate parameters and their corresponding comprehensive loss values. The process of candidate parameter generation, finite element calculation, simulation result extraction and comprehensive loss value calculation is repeated. S44: Record the candidate parameters and comprehensive loss value corresponding to each iteration in real time. When the comprehensive loss objective function meets the preset convergence condition or reaches the preset maximum number of iterations, terminate the parameter search and determine the parameter combination with the smallest comprehensive loss value from the completed candidate parameter combinations as the optimal parameter combination.

6. The method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method according to claim 5, characterized in that: The comprehensive loss objective function, excluding the coefficient of determination In addition to Dynamic Time Warped Distance (DTW), it also includes Structural Similarity (SSIM), and may further include one or more evaluation metrics among Root Mean Square Error (RMSE), Maximum Absolute Error (MaxAE), and Mean Absolute Error (MAE). The evaluation index reflecting the better the fit is converted into a loss term. The dimensional error evaluation index is made dimensionless according to the characteristic scale of the measured settlement data, and then weighted and summed according to the preset weights to obtain the comprehensive loss value. The sum of the weights of each evaluation indicator is 1, and the coefficient of determination is... The sum of the weights corresponding to the structural similarity SSIM is not less than 0.70; For soft soil layers described using the Soft Soil Creep constitutive model, the corrected rebound index is... With the corrected compression index satisfy: ； in, The value range is 0.08 to 0.15; Corrected creep index With the corrected compression index satisfy: ； in, The value range is 0.03 to 0.07; By and As an independent parameter to be searched, or determined according to the above proportional relationship. and The search range is determined so that the parameters of the generated Soft Soil Creep constitutive model meet the preset engineering physical constraints.

7. The method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method according to claim 1, characterized in that, The optimal parameter combination is backfilled into the initial finite element calculation model established in S3 to simulate the subgrade filling, preloading, operation, and long-term service processes, obtaining deformation, stress, and excess pore water pressure response data of the subgrade and road structure. The response data is spatially correlated and compared with the location, type, and distribution characteristics of measured surface cracks to determine the correspondence between abnormal settlement and surface cracks, assess the road structure safety status, identify the main influencing factors related to surface crack formation, and output road structure safety diagnosis and risk assessment results, specifically including: S51: Replace the corresponding parameters to be inverted in the initial finite element calculation model with the optimal parameter combination to form a road finite element calculation model after inversion calibration based on measured settlement data; wherein, the corresponding soil constitutive model, pile unit, drainage structure unit and pavement structure material model are set according to the soil layer and road structure type of the target road section. S52: Based on the actual construction and service process of the road, the roadbed is simulated in sequence in the road finite element calculation model after inversion calibration, including the stages of graded filling, preloading, operation and long-term service. Consolidation and stress deformation calculations are performed to obtain the calculation results of vertical deformation, horizontal deformation, stress, excess pore water pressure and pile deformation of the roadbed and road structure corresponding to each calculation stage. S53: Based on the road mileage station and cross-sectional spatial location, spatially match the finite element simulation results obtained in S52 with the measured location, type, and distribution range of road surface cracks, extract the road surface settlement, settlement gradient, structural deformation, stress, and pile deformation characteristics corresponding to each crack location and adjacent area; based on the consistency between the simulated deformation characteristics and the measured settlement and crack distribution, determine whether the road surface cracks are related to abnormal subgrade settlement, and determine the main influencing factors of abnormal settlement by combining the geological conditions of the crack location and the subgrade treatment parameters; S54: Based on the crack-deformation correspondence, main influencing factors of abnormal settlement, and road structure calculation results obtained in S53, generate standardized road structure safety diagnosis results including the location of abnormal settlement road sections, crack characteristics, settlement and deformation characteristics, finite element calculation results, main causes of cracks, road structure safety status, and risk assessment information.

8. A system for automated analysis of road structure safety using a Bayesian algorithm coupled with finite element method, characterized in that, include: The data acquisition and raw database construction module is used to collect and integrate road surface crack images, road surface elevation data at different detection periods, settlement monitoring data during construction and operation, geological survey data, indoor geotechnical physical and mechanical parameters, road design data, subgrade treatment design and construction parameters, and as-built data, and store them according to unified data fields to build a full-dimensional raw database. The spatiotemporal registration and structured database construction module is used to perform data cleaning, coordinate transformation, error correction, and spatial registration on the road surface crack images, road surface elevation data, and settlement monitoring data in the full-dimensional original database. It unifies data from different sources to the geodetic coordinate system used in road design and matches crack locations, elevation measurement points, and settlement monitoring points with the road design horizontal alignment and mileage station numbers. Using the road mileage station numbers as a unified spatial index, a three-level indexed structured database of "road segment - cross section - measurement point" is constructed. The candidate analysis segment screening and initial finite element analysis module is used to screen candidate analysis segments based on the three-level indexed structured database, according to the type of road surface cracks, the spatial location of cracks, and the corresponding settlement deformation characteristics. It then calls upon the geological survey data, indoor geotechnical parameters, road design data, subgrade treatment design and construction parameters, and as-built data corresponding to the candidate analysis segments to establish an initial finite element calculation model for the candidate analysis segments. Parameters from the geological survey and design / construction data are used as initial parameters for the finite element model calculation. Finally, it combines measured settlement deformation data with the initial finite element calculation results to identify abnormal settlement segments and constructs a correlation sample set between crack characteristics, settlement deformation characteristics, geological conditions, and subgrade treatment parameters. The Bayesian optimization parameter inversion module is used to construct a comprehensive loss objective function based on the measured settlement deformation data in the associated sample set to evaluate the difference between the finite element simulation deformation data and the measured deformation data. It also determines the prior search interval and physical constraints of the parameters to be inverted based on geological survey data and roadbed treatment design and construction parameters. The module employs a tree-structured Parzen estimator sampling strategy under the Optuna framework to automatically iteratively search for the parameters to be inverted in the finite element model, using the comprehensive loss objective function as the basis for parameter optimization, to obtain the parameter combination with the optimal fit between the finite element simulation deformation data and the measured deformation data. The finite element calibration and road structure safety analysis module is used to backfill the optimal parameter combination into the initial finite element calculation model, simulate the roadbed filling, preloading, operation and long-term service process, and obtain the deformation, stress and excess pore water pressure response data of the roadbed and road structure; spatially correlate and compare the response data with the location, type and distribution characteristics of the measured cracks on the road surface to determine the correspondence between abnormal settlement and road surface cracks, judge the safety status of the road structure, identify the main influencing factors related to the formation of road surface cracks, and output the road structure safety diagnosis results and risk assessment results.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for automated analysis of road structure safety using Bayesian algorithm coupled with finite element method as described in any one of claims 1 to 7.

Images