Highway pavement intelligent maintenance scientific decision-making system and decision-making method thereof

Abstract

The invention provides a highway pavement intelligent maintenance scientific decision-making system and a decision-making method thereof. A data collection and preprocessing module collects multi-source data in a public network and a highway management intranet, and performs cleaning, integration and dimensionality reduction by using a specific algorithm. And building an automatic road detection system based on a Python Django framework, determining an index weight by using a rough set theory, calculating PQI and MQI values, and generating a road technical condition evaluation report. And the pavement performance prediction module integrates multi-source data and predicts pavement performance by adopting a composite kernel function and a Gaussian process regression model. And the maintenance decision module determines the maintenance priority through ANP, generates a maintenance demand list in combination with a hidden Markov model, and generates a fund distribution scheme by using a cuckoo search algorithm. And the maintenance benefit evaluation module evaluates the maintenance benefit by adopting a gray ideal solution and a super-efficiency DEA model. And finally, utilizing a WebGL-based 3D visualization technology and an Echarts-based interactive query analysis function to realize road condition information display and auxiliary decision making.

Description

Technical Field

[0001] The present invention relates to the field of intelligent highway pavement maintenance, and in particular to a scientific decision-making system and a decision-making method for intelligent highway pavement maintenance. Background Art

[0002] As an important part of the national transportation infrastructure, the road condition of highways is directly related to the safety, comfort and efficiency of transportation. However, with the increase in the service life of highways and the continuous growth of traffic volume, various road surface defects such as cracks, potholes, rutting, etc. will appear, which seriously affect the performance and service quality of the road surface. The traditional highway pavement maintenance decision-making method mainly relies on manual experience and regular inspection, which has problems such as low inspection efficiency, poor data accuracy, and lack of scientific decision-making.

[0003] On the one hand, manual inspection is limited by the professional level and work experience of inspectors, and it is difficult to fully and accurately grasp the actual condition of the road surface. Moreover, the cycle of manual inspection is long, and it is impossible to detect the development and changes of road diseases in time, resulting in the lag of maintenance decision-making. On the other hand, traditional maintenance decisions often only consider a single factor, such as the degree of road damage, while ignoring the comprehensive influence of multiple factors such as traffic volume, climate conditions, and maintenance costs, making the formulation of maintenance plans unreasonable and difficult to achieve efficient use of maintenance funds and maximize the improvement of road performance.

[0004] In addition, with the rapid development of information technology, a large amount of pavement inspection data, traffic data, climate data, etc. are constantly accumulated, but there is a lack of effective data analysis and processing methods, and these data have not been able to fully play their role in maintenance decision-making. At the same time, the traditional way of displaying maintenance information is relatively single, and it is difficult for decision makers to intuitively understand the overall condition and maintenance needs of the pavement, which is not conducive to making scientific and reasonable decisions. Therefore, there is an urgent need for an intelligent and scientific decision-making method for intelligent maintenance of highway pavements to improve maintenance efficiency, reduce maintenance costs, and improve pavement performance and service quality. Summary of the invention

[0005] The main purpose of the present invention is to provide a scientific decision-making system and a decision-making method for intelligent maintenance of highway pavement, so as to solve the problem that the traditional maintenance information display method is relatively single, making it difficult for decision makers to intuitively understand the overall condition and maintenance needs of the road surface, which is not conducive to making scientific and reasonable decisions.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a scientific decision-making method for intelligent maintenance of highway pavement, the method comprising: S1, collecting multi-source data of basic data, auxiliary basic data and maintenance decision data on the public network and the highway management intranet; pre-processing the above multi-source data; S2. Build an automated road detection system, input the road damage, flatness, rutting, and lateral force coefficient of the road section that needs maintenance into the automated road detection system, and the automated road detection system generates a highway technical condition assessment report based on the PQI and MQI values; S3, based on the highway technical condition assessment report in step S2, combined with historical test data, traffic volume data, and climate data, a Gaussian process regression model is used to predict road performance, and a road performance prediction curve for the next 1-5 years is generated; S4, according to the pavement performance prediction curve in step S3, a prediction curve, maintenance standards, and capital budget data are obtained, and the prediction curve, maintenance standards, and capital budget data are used to determine the maintenance priorities of different road sections by using ANP; Combined with the priorities determined by ANP, the hidden Markov model is applied to predict the changing trend of maintenance needs and generate a maintenance needs list; S5. Based on the maintenance demand list in step S4, a cuckoo search algorithm is used to generate a funding allocation plan to obtain maintenance plans, implementation costs, and expected effect data; S6. Input the maintenance plan, implementation cost, and expected effect data into the analysis system to evaluate the improvement of pavement performance before and after maintenance, the economic and social benefits of maintenance, and generate a maintenance benefit analysis report; S7, using WebGL-based 3D visualization technology to display road condition distribution, disease location, and maintenance plan information for the highway technical condition assessment report, pavement performance prediction curve, maintenance demand list, and funding allocation plan in S2-S5; Then use the interactive query and statistical analysis functions based on Echarts. In the preferred solution, a density-based spatial clustering algorithm is used to clean and integrate the multi-source data to remove noise data; a local linear embedding algorithm is used to perform data dimensionality reduction processing; Among them, basic data include: map GIS data, road network information data, highway pavement and subgrade data, and pavement technical condition assessment data; Auxiliary basic data include: traffic flow data, meteorological data, geological environment data, and administrative management data; Maintenance decision data: maintenance material category data, maintenance equipment data, maintenance method data, and maintenance project unit price data.

[0007] In the preferred solution, the Python-based Django framework is used to build an automated road detection system; Use the jango-admin startproject command to create a Django project, and then execute the pythonmanage.py startapp command to create a specific application; Configure database connection information in the project configuration file, and select MySQL or SQLite database; define the database model for road detection data and indicator calculation, including road damage, flatness, rutting, lateral force coefficient detection data, and PQI, MQI indicator data, and execute python manage.py makemigrations and pythonmanage.py migrate commands to create the database table corresponding to the database model; Write Python scripts to collect data on road damage, flatness, rutting, and lateral force coefficients for sections that need maintenance; use Django's REST framework to develop an API interface to receive the collected data and store it in a database; Write Python code in Django application to implement rough set theory algorithm and build information system with collected data ,in is a collection of objects, ,in is the set of conditional attributes, i.e., various detection indicators. is the decision attribute set, i.e., the overall performance of the highway. is a set of attribute values, is the information function; Attribute reduction is performed by calculating the difference matrix, the formula is: ( ); Calculating attribute importance ,in , ; Finally, according to Determine the weight of each indicator; The collected raw data of PCI, RQI, RDI, and PBI that constitute the PQI are standardized again; Among them, PCI is the pavement damage index, RQI is the pavement ride quality index, RDI is the pavement rutting depth index, and PBI is the pavement skid resistance index; According to Calculate the PQI value; For MQI, the original data of its constituent indicators PQI, BCI, TCI, and SCI are subjected to the same standardization process, and the rough set theory is used again to determine the weight of each indicator. Calculate the MQI value; Among them, PQI is the pavement performance index, MQI is the highway technical condition index; Based on the calculated PQI and MQI values, combined with the pre-set evaluation grade standards, a highway technical condition evaluation report including basic information of the road section, indicator calculation results, evaluation grade, analysis and maintenance suggestions is generated in the Django application; Use Matplotlib and Echarts libraries to visualize the assessment results; use the Django template system to develop a Web interface to view data and assessment report information; Use the Django testing framework to test each functional module of the system to ensure that data collection, calculation, and report generation functions are normal; Deploy the system to a cloud server, using Nginx as the web server and Gunicorn as the application server.

[0008] In the preferred embodiment, the specific steps of step S3 are: S31. Collect the PQI, MQI and sub-item index values ​​in the highway technical condition assessment report of step S2, and integrate them with the collected historical test data, traffic volume data, and climate data; use the method based on local outlier factors to detect outliers. , through the formula Calculate the local reachability density using the formula Calculate the local outlier factor when Process outliers when they are greater than the set threshold; in yes of The nearest neighbor sample set, , It is a sample To its The distance to the nearest neighbor, It is a sample and The Euclidean distance between S32, using composite kernel function , for the training data set , calculate the covariance matrix , whose elements , considering the observation noise Get the covariance matrix with noise ; S33. Define the log-likelihood function for Gaussian process regression , using the Bayesian optimization algorithm to find the optimal hyperparameters ; S34. Generate forecast input data based on the time span of the next 1-5 years and the trend of historical data changes , calculate the covariance matrix between the prediction input and the training data and the covariance matrix of the prediction input itself , through the formula Calculate the predicted mean using the formula Calculate forecast covariance; S35. Based on the prediction covariance matrix The confidence interval of the predicted value is calculated, and the predicted mean and confidence interval are plotted over time to form a pavement performance prediction curve for the next 1-5 years.

[0009] In the preferred embodiment, the specific steps of step S4 are: S41, collect the pavement performance prediction curve data, maintenance standards and capital budget data of step S3; use the method based on the interquartile range to process the outliers, and use the formula Determine the outlier judgment interval , linear interpolation method is used to correct abnormal values; min-max normalization formula is used to normalize the road performance prediction curve data; S42. Construct a network structure model including pavement performance, traffic impact, and economic cost element sets, analyze the interdependence between elements; construct a judgment matrix through expert scoring method ; Solve Get the eigenvectors and combine them into an unweighted supermatrix ; Determine the element set association weight matrix , calculate the weighted supermatrix ;right Perform power calculation Get the limit supermatrix , determine maintenance priorities; S43. Define the maintenance requirement status set and the observation set ; Initialize the state transition probability matrix , observation probability matrix and the initial state probability vector ; Using the Baum-Welch algorithm, the forward probability recursive formula And the backward probability recursion formula Train the model and update the parameters ; Using the Viterbi algorithm, through the recursive formula Find the most likely sequence of states; S44. Combine the priorities determined by ANP and the trends predicted by the hidden Markov model, generate a maintenance demand list including road section information, maintenance time, maintenance demand status, and estimated required funds according to the road section number and time sequence, and calculate the maintenance demand list according to the capital budget formula. Adjust and screen maintenance items.

[0010] In the preferred embodiment, the specific steps of step S5 are: S51, extracting the road section maintenance demand information and estimated required funds from the maintenance demand list of step S4 and maintenance priorities , clarify the total amount of funds budget and the annual funding allocation ratio ; Define the objective function ,in: The maintenance benefit function is defined as ; The maintenance cost function is defined as: ; Setting Constraints and priority penalty ; S52, in the cuckoo search algorithm step, a solution with the largest objective function value is selected as the optimal solution; Generate a funding allocation plan based on the optimal solution, determine the maintenance section and allocated funds, clarify the maintenance plan; and calculate the total implementation cost and expected effect data .

[0011] In the preferred solution, in step S6, the maintenance plan, implementation cost, and expected effect data are input into the analysis system, and the analysis system uses the grey ideal solution to evaluate the improvement of the pavement performance before and after maintenance; The analysis system then uses the super-efficiency-based DEA model to evaluate the economic and social benefits of maintenance and generate a maintenance benefit analysis report.

[0012] In the preferred embodiment, the specific steps of step S6 are: S61, collecting the maintenance plan, implementation cost, and expected effect data in steps S51 and S52 and entering them into the analysis system; standardizing the original data matrix to eliminate the dimensional effect; S62, determining a standardized data matrix and calculating a road performance improvement index; S63. Use the super-efficiency DEA model to evaluate economic and social benefits: Each maintenance plan is used as a decision-making unit to determine the input implementation cost and output economic and social benefit indicators; a super-efficiency DEA model is established, and the objective function is ; The constraints are: , , , and solve for the super-efficiency value ; in For the DMU Input indicator values, For the DMU Output indicator value, and are the weights of output and input indicators respectively, is a non-Archimedean infinitesimal; S64. Report includes cover page and table of contents, project overview, and pavement performance improvement assessment , economic and social benefit evaluation , comprehensive evaluation and suggestions, and appendix.

[0013] In the preferred solution, the highway technical condition assessment report, pavement performance prediction curve, maintenance demand list and funding allocation plan of steps S2-S5 are integrated; the geographical location information of the road section is converted into geographical coordinates suitable for WebGL rendering; Use the WebGL library to create a 3D scene and load the geographic basemap; create a 3D model of the road segment based on the geographical location of the road segment, using the formula Mark the location of the disease with a specific geometric shape; in is the base size, is the standardized disease severity data, is the scaling factor, which is used to control the range of disease mark size; According to the maintenance needs list and funding allocation plan, maintenance measures are represented by different colored lines or textures, and the formula Use animation to show maintenance progress; in and are the starting color and the ending color respectively. progress is the progress of the maintenance plan; Bind the integrated data with Echarts to create bar charts, line charts, and pie charts; add interactive events and query input boxes to the charts to implement interactive queries; provide summation and average statistical analysis functions, and display the statistical results in new charts or tables; set the data update cycle to implement dynamic data updates; Using the formula Calculate comprehensive benefit indicators; in , and is the weight coefficient, which is determined by expert scoring or historical data statistics. is the improvement value of pavement performance PQI, cost is the maintenance cost, It is the quantitative value of social benefits; Provide decision-makers with suggestions on maintenance plan adjustment and fund allocation optimization based on PQI and MQI indicators.

[0014] The preferred solution includes a data collection and preprocessing module, a road automation detection system, a pavement performance prediction module, a maintenance decision module, a maintenance benefit evaluation module, and a visualization and analysis module; The data collection and preprocessing module collects multi-source data of basic data, auxiliary basic data and maintenance decision data from the public network and highway management intranet, and uses density-based spatial clustering algorithm for cleaning and integration, and adopts local linear embedding algorithm for data dimension reduction processing; The road automation detection system is built with the Python-based Django framework, which uses the rough set theory algorithm to determine the detection indicator weights and calculate the PQI and MQI values ​​for data collection, indicator calculation, report generation and display, system testing and deployment functions; The pavement performance prediction module is used to integrate relevant data, process outliers and construct derived features. It uses a composite kernel function and Gaussian process regression model combined with a Bayesian optimization algorithm to predict pavement performance and generate a pavement performance prediction curve for the next 1 to 5 years. The maintenance decision module is used to determine maintenance priorities, predict the trend of maintenance demand changes to generate a maintenance demand list, use the cuckoo search algorithm to generate a funding allocation plan, and determine the maintenance plan, implementation cost, and expected effect data; The maintenance benefit evaluation module uses the grey ideal solution and the super-efficiency DEA model to evaluate the improvement of pavement performance before and after maintenance, economic benefits and social benefits, and generates a maintenance benefit analysis report; The visualization and analysis module uses WebGL-based 3D visualization technology to display road condition distribution, disease location, and maintenance plan information. It uses Echarts-based interactive query and statistical analysis functions to support dynamic data updates and calculate comprehensive benefit indicators to assist decision-making.

[0015] The present invention provides a scientific decision-making system for intelligent maintenance of highway pavement and its decision-making method. At the data processing and analysis level, it can efficiently integrate multi-source data such as highway technical condition assessment reports, pavement performance prediction curves, maintenance needs lists and funding allocation plans, and process them through advanced algorithms to improve data accuracy and availability and tap the potential value of data.

[0016] In terms of maintenance decision-making, rough set theory is used to determine indicator weights, combined with the Gaussian process regression model to predict road performance, maintenance priorities are determined based on ANP, and then the cuckoo search algorithm is used to optimize fund allocation, so that decisions take into account multiple factors and are more scientific and reasonable, thereby improving the efficiency of the use of maintenance funds and accurately investing resources in the sections that need the most maintenance.

[0017] In terms of evaluation and report generation, the maintenance benefits are evaluated through the grey ideal solution and super-efficiency DEA model, and a comprehensive maintenance benefit analysis report is generated to provide strong data support and clear evaluation results for decision-making.

[0018] In the field of visualization and interactive analysis, 3D visualization technology based on WebGL can intuitively display the distribution of road conditions, the location of defects and maintenance plans. Combined with the interactive query and statistical analysis functions of Echarts, it supports dynamic data updates, making it easier for decision makers to intuitively grasp the road conditions, obtain detailed information in a timely manner and conduct statistical analysis, so as to make scientific decisions quickly.

[0019] This method can improve the intelligence and scientific level of highway pavement maintenance, reduce maintenance costs, improve pavement performance and service quality, ensure road traffic safety and smoothness, and promote highway maintenance management to develop in an efficient and precise direction. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 This is a diagram of a scientific decision-making system for intelligent maintenance of a highway pavement according to the present invention; Figure 2 It is a data collection and preprocessing module diagram of the present invention; Figure 3 It is a diagram of the road automation detection system of the present invention; Figure 4 It is a road performance prediction module diagram of the present invention; Figure 5 It is the maintenance decision module diagram of the present invention; Figure 6 It is a maintenance benefit evaluation module diagram of the present invention; Figure 7 It is a visualization and analysis module diagram of the present invention. DETAILED DESCRIPTION

[0021] Example 1 like Figure 1-7 As shown, a scientific decision-making method for intelligent maintenance of highway pavement includes: S1. Collect multi-source data of basic data, auxiliary basic data and maintenance decision data on the public network and highway management intranet; pre-process the multi-source data; S2. Build an automated road detection system, input the road damage, flatness, rutting, and lateral force coefficient of the road section that needs maintenance into the automated road detection system, and the automated road detection system generates a highway technical condition assessment report based on the PQI and MQI values; S3, based on the highway technical condition assessment report in step S2, combined with historical test data, traffic volume data, and climate data, a Gaussian process regression model is used to predict road performance, and a road performance prediction curve for the next 1-5 years is generated; S4, according to the pavement performance prediction curve in step S3, a prediction curve, maintenance standards, and capital budget data are obtained, and the prediction curve, maintenance standards, and capital budget data are used to determine the maintenance priorities of different road sections by using ANP; Combined with the priorities determined by ANP, the hidden Markov model is applied to predict the changing trend of maintenance needs and generate a maintenance needs list; S5. Based on the maintenance demand list in step S4, a cuckoo search algorithm is used to generate a funding allocation plan to obtain maintenance plans, implementation costs, and expected effect data; S6. Input the maintenance plan, implementation cost, expected effect and other data into the analysis system to evaluate the improvement of pavement performance before and after maintenance, the economic and social benefits of maintenance, and generate a maintenance benefit analysis report; S7, using WebGL-based 3D visualization technology to display road condition distribution, disease location, and maintenance plan information for the highway technical condition assessment report, pavement performance prediction curve, maintenance demand list, and funding allocation plan in S2-S5; Then use the interactive query and statistical analysis functions based on Echarts to support dynamic data updates and assist decision-making.

[0022] In the preferred solution, a density-based spatial clustering algorithm is used to clean and integrate the multi-source data to remove noise data; a local linear embedding algorithm is used to perform data dimensionality reduction processing; Among them, basic data include: map GIS data, road network information data, highway pavement and subgrade data, and pavement technical condition assessment data; Auxiliary basic data include: traffic flow data, meteorological data, geological environment data, and administrative management data; Maintenance decision data: maintenance material category data, maintenance equipment data, maintenance method data, and maintenance project unit price data.

[0023] Comprehensively collect necessary basic data including map GIS data, road network information data, highway pavement and subgrade data, pavement technical condition assessment data, auxiliary basic data such as traffic flow data, meteorological data, geological environment data, administrative management data, and maintenance decision-making related data such as maintenance material category data, maintenance equipment data, maintenance method data, and maintenance project unit price data. Use density-based spatial clustering algorithm (DBSCAN) to clean and integrate multi-source heterogeneous data, identify and remove noise data such as outliers, and achieve effective data integration. Use local linear embedding algorithm (LLE) to reduce data dimensionality, reduce redundant data dimensions while retaining the local geometric structure of the data, reduce data complexity, improve subsequent calculation efficiency, and provide high-quality data support for subsequent decision-making.

[0024] Example 2 In conjunction with Example 1, further explanation is given, a road automation detection system is built using the Python-based Django framework; Use the django-admin startproject command to create a Django project, and then execute the pythonmanage.py startapp command to create a specific application; Configure database connection information in the project configuration file. You can choose MySQL or SQLite database; define the database model related to road detection data and indicator calculation. The database model includes road damage, flatness, rutting, lateral force coefficient detection data, and PQI, MQI indicator data. Execute python manage.py makemigrations and python manage.py migrate commands to create the database table corresponding to the database model; Write Python scripts to collect data on road damage, flatness, rutting, and lateral force coefficients for sections that need maintenance; use Django's REST framework to develop an API interface to receive the collected data and store it in a database; Write Python code in Django application to implement rough set theory algorithm and build information system with collected data ,in is a collection of objects, ,in is the set of conditional attributes, i.e., various detection indicators. is the decision attribute set, i.e., the overall performance of the highway. is a set of attribute values, is the information function; Attribute reduction is performed by calculating the difference matrix, the formula is: ( ); Calculating attribute importance ,in , ; Finally, according to Determine the weight of each indicator; Standardize the collected raw data of PCI, RQI, RDI, PBI and other indicators that constitute PQI; Among them, PCI is the pavement damage index, RQI is the pavement ride quality index, RDI is the pavement rutting depth index, and PBI is the pavement skid resistance index; According to Calculate the PQI value; for MQI, the original data of its constituent indicators PQI, BCI, TCI, and SCI are subjected to the same standardization process, and the rough set theory is used again to determine the weight of each indicator, and then Calculate the MQI value; Based on the calculated PQI and MQI values, combined with the pre-set evaluation grade standards, a highway technical condition evaluation report including basic information of the road section, indicator calculation results, evaluation grade, analysis and maintenance suggestions is generated in the Django application; Use Matplotlib and Echarts libraries to visualize the assessment results; use the Django template system to develop a Web interface to view data and assessment report information; Use the Django testing framework to test the various functional modules of the system to ensure that data collection, calculation, and report generation functions are normal; deploy the system to the cloud server, use Nginx as the web server and Gunicorn as the application server.

[0025] In the above, the road damage, flatness, rutting, and lateral force coefficient are obtained as follows: The road damage coefficient uses advanced equipment installed on the inspection vehicle, using high-definition cameras and laser scanners to quickly collect road surface images and three-dimensional data while the vehicle is driving. Through image processing and analysis technology, it identifies and classifies road damage and calculates its related parameters.

[0026] The smoothness coefficient uses a laser sensor to measure the distance change between the road surface and the sensor, and the road surface smoothness index is obtained by computer processing the data. This method can quickly and accurately measure the smoothness of the road surface, and can also obtain three-dimensional information of the road surface at the same time.

[0027] The rutting coefficient is measured by a laser rutting meter installed on a testing vehicle. The laser scans the cross section of the road surface while the vehicle is driving, quickly and accurately measures parameters such as the depth, width and shape of the rutting, and automatically records and processes the data.

[0028] The lateral force coefficient is measured by using a special friction coefficient measuring vehicle, which measures the lateral force coefficient by using the friction between the test wheel and the road surface when driving at a certain speed. The test wheel usually uses special tires and maintains a certain vertical load and slip rate during the test.

[0029] Example 3 Further described in conjunction with Examples 1-2, the specific steps of step S3 are: S31. Collect the PQI, MQI and sub-item index values ​​in the highway technical condition assessment report of step S2, and integrate them with the collected historical test data, traffic volume data and climate data; use the method based on local outlier factor to detect outliers, and for the sample points in the data set , through the formula Calculate the local reachability density using the formula Calculate the local outlier factor when Process outliers when they are greater than the set threshold; construct derived features, including: moving average of traffic volume , Comprehensive index of climate factors ,in Determined by principal component analysis; S32, using composite kernel function , for the training data set , calculate the covariance matrix , whose elements , considering the observation noise Get the covariance matrix with noise ; S33. Define the log-likelihood function for Gaussian process regression , using the Bayesian optimization algorithm to find the optimal hyperparameters ; S34. Generate forecast input data based on the time span of the next 1-5 years and the trend of historical data changes , calculate the covariance matrix between the prediction input and the training data and the covariance matrix of the prediction input itself , through the formula Calculate the predicted mean using the formula Calculate forecast covariance; S35. Based on the prediction covariance matrix The confidence interval of the predicted value is calculated, and the predicted mean and confidence interval are plotted over time to form a pavement performance prediction curve for the next 1-5 years.

[0030] The specific steps are: Collect the PQI, MQI and sub-index values ​​in the highway technical condition assessment report in step S2, historical inspection data (pavement performance related indicators over the years), traffic volume data (traffic volume at different times, vehicle type ratio, etc.), and climate data (temperature, precipitation, sunshine duration, etc.). Integrate these data into a multidimensional dataset according to time series and road section information.

[0031] The outliers are detected using a method based on the local outlier factor (LOF). , calculate its local reachable density and local outlier factor .

[0032] Local reachability density formula:

[0033] in yes of The nearest neighbor sample set, , It is a sample To its The distance to the nearest neighbor, It is a sample and The Euclidean distance between .

[0034] Local outlier factor formula:

[0035] when When it is greater than the set threshold (such as 1.5), it is considered It is an outlier and can be processed by interpolation or deletion. The purpose of this method is to remove noise points in the data to avoid adverse effects on model training.

[0036] In addition to the original data features, some derived features are constructed. Calculate the moving average of traffic volume ,in It is Traffic volume at the time, is the window size of the moving average. It is also possible to construct a comprehensive index of climate factors, such as ,in are weight coefficients determined by principal component analysis (PCA). These derived features help the model capture the underlying regularities of the data.

[0037] The composite kernel function is used to combine the radial basis function (RBF) and the rational quadratic kernel function (RQ). The formula of the composite kernel function is: ; in and are the signal variances of the RBF and RQ kernel functions, and are their length scales, is the shape parameter of the rational quadratic kernel function. The composite kernel function can better adapt to different types of data features and improve the flexibility of the model.

[0038] For the training dataset , the covariance matrix Elements At the same time, considering the observation noise , and obtain the noisy covariance matrix ,in yes The covariance matrix describes the correlation between data points and is the core of Gaussian process regression.

[0039] The log-likelihood function for Gaussian process regression is: ; in is the observation vector of the training data, is the model’s hyperparameter vector, yes The likelihood function measures the probability of the observed data occurring under given hyperparameters.

[0040] Use Bayesian optimization algorithm to find the optimal hyperparameters Bayesian optimization estimates the posterior distribution of the objective function (here, the log-likelihood function) by building a proxy model (such as a Gaussian process model), and then selects the next hyperparameter point to be evaluated based on the acquisition function (such as the expected improvement EI). This process is repeated until the optimal hyperparameter combination is found. Bayesian optimization can find better hyperparameters with fewer evaluations, improving optimization efficiency.

[0041] Generate forecast input data based on the time span of 1-5 years in the future, combining the trend of historical traffic volume and climate data Traffic volume can be predicted using time series prediction methods, and the predicted value can be used as input.

[0042] Compute the covariance matrix between the prediction input and the training data ,element , and the covariance matrix of the predicted input itself .

[0043] Prediction mean formula:

[0044] Prediction covariance formula:

[0045] The prediction mean gives the predicted value of the pavement performance, and the prediction covariance reflects the uncertainty of the prediction.

[0046] Based on the prediction covariance matrix , calculate the confidence interval for the predicted value. The 95% confidence interval is ,in yes This helps to assess the reliability of the forecast results.

[0047] The predicted mean and confidence interval are plotted over time to form a pavement performance prediction curve for the next 1-5 years. This curve intuitively shows the changing trend and uncertainty range of pavement performance, providing a scientific basis for highway maintenance decisions.

[0048] Example 4 Further explanation in conjunction with Examples 1-3: The specific steps of step S4 are: S41, collect the pavement performance prediction curve data, maintenance standards and capital budget data of step S3; use the method based on the interquartile range to process the outliers, and use the formula Determine the outlier judgment interval , linear interpolation method is used to correct abnormal values; min-max normalization formula is used to normalize the road performance prediction curve data; S42. Construct a network structure model including pavement performance, traffic impact, and economic cost element sets, analyze the interdependence between elements; construct a judgment matrix through expert scoring method ; Solve Get the eigenvectors and combine them into an unweighted supermatrix ; Determine the element set association weight matrix , calculate the weighted supermatrix ;right Perform power calculation Get the limit supermatrix , determine maintenance priorities; S43. Define the maintenance requirement status set and the observation set ; Initialize the state transition probability matrix , observation probability matrix and the initial state probability vector ; Using the Baum-Welch algorithm, the forward probability recursive formula And the backward probability recursion formula Train the model and update the parameters ; Using the Viterbi algorithm, through the recursive formula Find the most likely sequence of states; S44. Combine the priorities determined by ANP and the trends predicted by the hidden Markov model, generate a maintenance demand list including road section information, maintenance time, maintenance demand status, and estimated required funds according to the road section number and time sequence, and calculate the maintenance demand list according to the capital budget formula. Adjust and screen maintenance items.

[0049] The specific implementation is as follows: Extract the pavement performance prediction curve data of each road section in the next 1-5 years from step S3, covering key indicators such as PQI and MQI. At the same time, collect detailed maintenance standards, maintenance measures and quality requirements corresponding to different PQI and MQI intervals; obtain capital budget data to clarify the total funds available for maintenance and the proportion of funds allocated in each year ( ).

[0050] Check the pavement performance prediction curve data, identify and handle outliers. For outliers, use the method based on the interquartile range (IQR). Assume that the lower quartile of the data is , the upper quartile is , then IQR = The outlier judgment interval is , values ​​beyond this interval are considered abnormal values ​​and can be corrected by linear interpolation. The pavement performance prediction curve data is normalized.

[0051] The factors that affect maintenance priority are divided into multiple element sets. Including pavement performance element set , Traffic Impact Elements Set , Economic Cost Element Set .

[0052] Analyze dependencies within and between sets of elements. For example, pavement performance affects traffic flow, which in turn affects maintenance costs. Build a network structure to visualize these relationships.

[0053] For the associated element pairs in the network, a judgment matrix is ​​constructed using the expert scoring method. Elements in and elements , judgment matrix ,in Representation elements Relative to element The importance of is determined by using a 1-9 scale. , calculate its eigenvector , by solving Get, where is a matrix The eigenvalue of . The eigenvectors corresponding to each judgment matrix are combined into an unweighted supermatrix The unweighted supermatrix reflects the relative importance of the elements, but does not take into account the association weights between element sets. Determine the association weight matrix between element sets , which is also obtained through expert scoring and eigenvector calculation. Weighted supermatrix The weighted supermatrix comprehensively considers the relationship weights within and between element sets.

[0054] Weighted Supermatrix Perform power operation, when When it is big enough, Converges to the limiting supermatrix The sum of the elements in each column of the extreme supermatrix is ​​1, and its element value represents the relative importance of each road section under the influence of all factors, that is, the maintenance priority. The power operation formula is .

[0055] Define the maintenance requirement status set according to the maintenance standards , which respectively mean “no maintenance required”, “light maintenance”, “moderate maintenance” and “heavy maintenance”.

[0056] Select observable variables that are closely related to maintenance needs as the observation set , such as normalized PQI, traffic flow, and temperature (considering the impact of climate on pavement performance).

[0057] State transition probability matrix : , indicating that at time In state At the moment Transfer to state The initial value can be estimated based on the statistical results of historical maintenance data. : , indicating that at time In state When it is observed The probability of . It is also initialized based on historical data.

[0058] Initial state probability vector : , indicating that at the initial moment it is in state The probability of is determined based on the initial road performance prediction data. and the backward probability ,in is the model parameter.

[0059] Forward probability recursion formula: , used to calculate at time In state And observed before The probability of observations. Backward probability recursion formula: , used to calculate at time In state And the probability of observing subsequent observations. Through continuous iteration and updating , and , so that the observation sequence The likelihood probability Maximum, to achieve optimization of model parameters.

[0060] Using the trained hidden Markov model and observation data such as road performance prediction curves, the most likely state sequence is found through the Viterbi algorithm. The Viterbi algorithm is based on the idea of ​​dynamic programming and defines Indicates at time In state And observed before The maximum probability of observations. The recursive formula is , and record the backtrace pointer , and finally trace back to obtain the most likely state sequence, that is, the changing trend of maintenance demand.

[0061] Combining the maintenance priority determined by ANP and the maintenance demand change trend predicted by the hidden Markov model, a maintenance demand list is generated according to the road section number and time sequence. The list includes: Section number, section length, starting and ending point locations, etc.

[0062] Determine the specific time point when maintenance is required for each road section based on the predicted trend of changes in maintenance demand.

[0063] Clarify the maintenance requirement status of each road section at the corresponding time point, including light maintenance and moderate maintenance.

[0064] Estimate the funds required for each maintenance of each road section based on maintenance standards and maintenance demand status At the same time, combined with the capital budget data, check whether the total maintenance funds for each year exceed the capital allocation for that year. If it exceeds, the maintenance items will be adjusted and screened according to the maintenance priority.

[0065] Example 5 Further described in conjunction with Examples 1-4, the specific steps of step S5 are: S51, extracting the road section maintenance demand information and estimated required funds from the maintenance demand list of step S4 and maintenance priorities , clarify the total amount of funds budget and the annual funding allocation ratio ; Define the objective function ,in ( ), ; Set constraints and priority penalty ; S52, Cuckoo search algorithm steps: random generation The length is The binary vector of is used as the initial solution; set the discovery probability , maximum number of iterations , step size scaling factor Parameters; using Lévy flight formula Generation step ,pass Generate new solutions ;calculate The objective function value of And check the constraints and modify the solution if they are not satisfied; compare and ,like Use replace ; with probability Randomly select nests to update and generate new solutions and evaluate replacements; When the maximum number of iterations is reached The algorithm is terminated when ; the solution with the largest objective function value is selected as the optimal solution ;according to Generate a funding allocation plan, determine the maintenance sections and allocate funds; clarify the maintenance plan; calculate the implementation cost and expected effect data .

[0066] The specific implementation method is: Extract the maintenance demand information of each road section from the maintenance demand list in step S4, including the required maintenance type (including light maintenance, medium maintenance, heavy maintenance), the estimated funds required ( represents the road section number) and the maintenance priority determined by ANP . At the same time, the total amount of funds budget should be clarified and the proportion of funds allocated in each year ( ).

[0067] Defining the maintenance benefit function ,in It is a road section The expected benefit improvement after maintenance can be calculated based on the improvement of indicators such as PQI and MQI in the pavement performance prediction curve, for example: ( and is the weight coefficient, determined by expert scoring or historical data statistics). is the decision variable, , Indicates the road section Carry out maintenance, Indicates that no maintenance is performed.

[0068] Defining the maintenance cost function Taking all factors into consideration, the objective function It can be expressed as ,in and It is the weight coefficient for weighing maintenance benefits and costs, and can be adjusted according to actual conditions. , ensuring that the funds allocated do not exceed the total budget.

[0069] For sections with higher maintenance priority, priority should be given to ensuring that they have higher Possibility of taking values. Priority penalty items can be introduced , try to minimize the penalty term during the optimization process.

[0070] Random Generation initial solutions (cuckoo nests), each of which is a length The binary vector ( ),in Indicates Solution Section Whether maintenance is carried out.

[0071] Set the algorithm's parameters, including the probability of discovery (used to control the probability of the cuckoo's nest being discovered and replaced), maximum number of iterations , step size scaling factor For every cuckoo's nest , using Lévy flight to generate new solutions . Lévy flight stride length By formula Calculate, where and is a random variable that follows a normal distribution, The value is usually 1.5. The formula for generating the new solution is ,in is a random vector whose elements are , which is used to randomly change some decision variables in the solution.

[0072] New calculation solution The objective function value of , and check whether it satisfies the constraints. If the total funding limit or maintenance priority constraint is not met, the solution is modified, such as randomly reducing some until the constraints are met.

[0073] Relatively new solution and original solution The objective function value of , then use replace With probability Randomly select some nests to update. For the selected nests , randomly generate a new solution , and similarly evaluate its objective function value and the situation of satisfying the constraints. If it is better than the original solution, it will be replaced.

[0074] When the maximum number of iterations is reached When , the algorithm terminates.

[0075] Select the solution with the largest objective function value from all cuckoo nests as the optimal solution .

[0076] According to the optimal solution , determine which sections of road will be maintained and the funds allocated to each section. Section , allocate its estimated funds required .

[0077] Clarify the type of maintenance to be adopted for each road section to be maintained and determine it based on the maintenance needs list.

[0078] Calculate the total cost of implementing the maintenance plan .

[0079] According to the benefit calculation method in the objective function, calculate the expected benefit improvement value after implementing the maintenance plan .

[0080] Example 6 Further explained in conjunction with Examples 1-5, the specific steps of step S6 are: inputting data such as the maintenance plan, implementation cost, and expected effect into the analysis system, and the analysis system uses the grey ideal solution to evaluate the improvement of the pavement performance before and after maintenance; The analysis system then uses the super-efficiency-based DEA model to evaluate the economic and social benefits of maintenance and generate a maintenance benefit analysis report.

[0081] In the preferred solution, S61, collect maintenance plan, implementation cost, expected effect data and enter them into the analysis system; Standardization is performed and positive indicators are used , negative indicators use Formula, eliminating the dimension effect; S62. Determine the standardized data matrix The ideal solution ( ) and negative ideal solution ( ); Calculate the grey correlation coefficient with the ideal solution ( ) and the grey correlation coefficient with the negative ideal solution ( ); Calculate the grey relational degree and ; Calculate the road performance improvement index ; S63. Use the super-efficiency DEA model to evaluate economic and social benefits: Take each maintenance plan as a decision-making unit, determine the input implementation cost and output economic and social benefit indicators; establish a super-efficiency DEA model with the objective function as follows: , the constraints are , , , and solve for the super-efficiency value ; S64. Report includes cover page and table of contents, project overview, and pavement performance improvement assessment , economic and social benefit evaluation , comprehensive evaluation and suggestions and appendix; The specific implementation method is: Collect detailed data on the maintenance plan, implementation cost, and expected effect obtained in step S5. The expected effect covers various pavement performance indicators before and after maintenance, such as pavement damage rate, flatness, anti-skid performance, etc. The implementation cost includes labor, materials, equipment rental and other expenses; social benefit data may include the degree of traffic congestion relief, surrounding residents' satisfaction, etc. Enter these data into the analysis system.

[0082] Since different indicators have different dimensions and value ranges, linear transformation is used for standardization to make the indicators comparable. Suppose the original data matrix is ,in is the number of maintenance options, is the number of indicators.

[0083] For positive indicators (the larger the value, the better, such as road surface smoothness improvement rate), the standardized formula is: ; For negative indicators (the smaller the value, the better, such as implementation cost), the standardized formula is: ; Eliminate the impact of different indicator dimensions and make comparisons among indicators on a unified scale, thus providing a basis for subsequent evaluation.

[0084] Suppose the standardized data matrix is . Ideal solution is a vector composed of the optimal values ​​of each indicator, that is, ,in ; Negative ideal solution is a vector of the worst values ​​of each indicator, that is, ,in . Serves as a reference standard for evaluating the pavement performance improvement of various maintenance plans.

[0085] Calculate the grey correlation coefficient between each maintenance scheme and the ideal solution and negative ideal solution. Maintenance plan The grey correlation coefficient of the ideal solution The calculation formula is: ; in , is the resolution coefficient, usually taken as 0.5.

[0086] Grey correlation coefficient with negative ideal solution The calculation formula is: ; in The above algorithm is used to measure the closeness of the index value of each maintenance plan to the ideal solution and the negative ideal solution.

[0087] The grey correlation coefficient of each maintenance scheme is weighted and summed to obtain the grey correlation degree with the ideal solution. Grey correlation degree of sum and negative ideal solution . Let the indicator weight vector be ,and ,but: ; ; Taking into account the influence of various indicators, the overall correlation degree of each maintenance scheme with the ideal solution and the negative ideal solution is obtained.

[0088] Pavement performance improvement index The calculation formula is:

[0089] Function: Quantify the improvement of pavement performance before and after each maintenance plan. The closer it is to 1, the greater the improvement in road performance.

[0090] Each maintenance plan is considered as a decision-making unit (DMU). The input indicators are various costs related to the implementation cost, and the output indicators are economic benefits (such as the economic benefits brought by the improvement of road traffic capacity after maintenance) and social benefits (such as the quantitative value of traffic congestion relief, the value of residents' satisfaction improvement, etc.).

[0091] For DMU, the objective function of the super-efficiency DEA model is: ; The constraints are: ; ; ; in For the DMU Input indicator values, For the DMU Output indicator value, and are the weights of output and input indicators respectively, is a non-Archimedean infinitesimal, usually taken as .

[0092] The model is solved by linear programming method and the Super efficiency value of a DMU .

[0093] The relative efficiency of each maintenance plan in terms of economic and social benefits is evaluated, and a super-efficiency value greater than 1 indicates that the plan is better in terms of input-output.

[0094] Cover and table of contents: include the report name, compiling unit, date and other information, and list the contents of each part of the report.

[0095] Project Overview: Introduce the background, objectives and main maintenance plans of the maintenance project.

[0096] Pavement performance improvement assessment: Detailed list of pavement performance improvement indexes for each maintenance plan , rank the various options, and analyze the key factors affecting the improvement of pavement performance.

[0097] Economic and social benefit evaluation: presenting the super efficiency value of each maintenance plan , analyze the advantages and disadvantages of each plan in terms of input and output, and point out the main ways to improve economic and social benefits.

[0098] Comprehensive evaluation and suggestions: Taking into account the improvement of pavement performance and the evaluation results of economic and social benefits, the maintenance plans are ranked comprehensively, the optimal maintenance plan is recommended, and improvement measures are proposed for the problems existing in each plan.

[0099] Example 7 Further described in conjunction with Examples 1-6, the highway technical condition assessment report, pavement performance prediction curve, maintenance demand list and funding allocation plan of steps S2 - S5 are integrated; the geographical location information of the road section is converted into geographical coordinates suitable for WebGL rendering; Use the WebGL library to create a 3D scene and load the geographic basemap; create a 3D model of the road segment based on the geographical location of the road segment, using the formula Mark the location of the disease with a specific geometric shape; in is the base size, is the standardized disease severity data, is the scaling factor, which is used to control the range of disease mark size; According to the maintenance needs list and funding allocation plan, maintenance measures are represented by different colored lines or textures, and the formula Use animation to show maintenance progress; in and are the starting color and the ending color respectively. progress is the progress of the maintenance plan; Bind the integrated data with Echarts to create bar charts, line charts, and pie charts; add interactive events and query input boxes to the charts to implement interactive queries; provide summation and average statistical analysis functions, and display the statistical results in new charts or tables; set the data update cycle to implement dynamic data updates; Using the formula Calculate comprehensive benefit indicators; in , and is the weight coefficient, which is determined by expert scoring or historical data statistics. is the improvement value of pavement performance PQI, cost is the maintenance cost, It is the quantitative value of social benefits; Provide decision-makers with advice on maintenance plan adjustment, fund allocation optimization, and other decision-making based on the indicators.

[0100] The specific implementation method is: Collect the highway technical condition assessment report in step S2, the pavement performance prediction curve in step S3, the maintenance demand list in step S4, and the funding allocation plan in step S5. These data are associated and integrated according to key information such as section number and geographical location to build a unified data set.

[0101] The pavement performance indicators (such as PQI, MQI), disease severity and other data are standardized so that their value range is between [0, 1] to facilitate subsequent visualization.

[0102] For the geographical location information of the road segment, convert it into geographical coordinates (such as longitude and latitude) suitable for WebGL rendering. If the original data is relative location information, it can be converted through the relevant algorithms of the Geographic Information System (GIS).

[0103] Use the WebGL related library (Three.js) to create a 3D scene. Define the basic parameters of the scene such as size, background color, lighting, etc.

[0104] Load the geographic basemap, you can use open source map data (OpenStreetMap), and convert it into a 3D terrain model as the basis for the entire road condition display.

[0105] According to the geographical location information of the road segment, a 3D model of the road segment is created in the 3D scene. Cylinders or cuboids can be used to represent different types of road segments, and their length, width, and height can be set according to the size of the actual road segment.

[0106] The damage location is marked with a specific geometric shape on the road segment model. The size and color of the damage can be set according to the severity of the damage (through the standardized damage degree data). For example, the more serious the disease, the larger the marked sphere and the brighter the color.

[0107] Disease mark size calculation formula: ; in is the base size, is the standardized disease severity data, is a scaling factor that controls the range of variation in the size of the disease marker. The purpose of this formula is to visually display the severity of the disease.

[0108] Visualize the maintenance plan for each road segment based on the maintenance needs list and funding allocation plan. Use different colored lines or textures to represent different types of maintenance measures (such as light maintenance, medium maintenance, heavy maintenance).

[0109] The time progress of the maintenance plan can be displayed through animation effects. Use color gradients or dynamic line growth to indicate the progress of the maintenance plan.

[0110] Maintenance progress color gradient formula: ; in and are the starting color and the ending color, respectively, and progress is the completion progress of the maintenance plan (the value range is [0, 1]). This formula is used to dynamically display the implementation status of the maintenance plan.

[0111] Bind the integrated data with Echarts to create various charts according to different analysis needs, such as bar charts (used to show the comparison of pavement performance indicators of different road sections), line charts (used to show pavement performance prediction curves), pie charts (used to show the proportion of fund allocation), etc.

[0112] Add interactive events to the chart, such as mouse hover, click, etc. When the user hovers over a chart element, the detailed data information corresponding to the element is displayed, such as the section number, specific pavement performance index value, maintenance plan details, etc.

[0113] Implement a query input box, where users can enter keywords (such as road section number, maintenance type, etc.) to filter and query data. The filtered data is updated in real time and displayed in charts.

[0114] Provides a variety of statistical analysis functions, such as sum, average, maximum, minimum, etc. Users can select different indicators and analysis dimensions (such as by road section, by time, etc.) for statistical analysis.

[0115] Statistical analysis results are presented in new charts or tables. For example, the total capital investment for different maintenance types can be presented in a bar chart.

[0116] Dynamic update mechanism: Set the data update cycle and regularly obtain the latest data from the data source. When new data is updated, the display content in the Echarts chart and WebGL 3D scene will be automatically refreshed to ensure the real-time nature of the data.

[0117] Based on the results of visualization and statistical analysis, some decision-related indicators are calculated. For example, the comprehensive benefit index of each road section is calculated, taking into account factors such as road performance improvement, maintenance cost, and social benefits.

[0118] Comprehensive benefit index calculation formula: ; in , and is the weight coefficient, which is determined by expert scoring or historical data statistics. is the improvement value of pavement performance PQI, cost is the maintenance cost, It is the quantitative value of social benefits. This formula is used to comprehensively evaluate the maintenance benefits of each road section and provide a basis for decision-making.

[0119] According to the calculated decision indicators, the system provides suggestions for decision makers on maintenance plan adjustment, fund allocation optimization, etc. For example, for road sections with higher comprehensive benefit indicators, the maintenance fund investment can be appropriately increased; for road sections with lower comprehensive benefit indicators, the rationality of the maintenance plan should be re-evaluated.

[0120] Example 8 Further illustrate with reference to Example 1, Figure 1-7 The structure shown includes a data collection and preprocessing module, a road automation detection system, a pavement performance prediction module, a maintenance decision module, a maintenance benefit evaluation module, and a visualization and analysis module; The data collection and preprocessing module collects multi-source data of basic data, auxiliary basic data and maintenance decision data from the public network and highway management intranet, and uses density-based spatial clustering algorithm for cleaning and integration, and adopts local linear embedding algorithm for data dimension reduction processing; The road automation detection system is built with the Python-based Django framework, which uses the rough set theory algorithm to determine the detection indicator weights and calculate the PQI and MQI values ​​for data collection, indicator calculation, report generation and display, system testing and deployment functions; The pavement performance prediction module is used to integrate relevant data, process outliers and construct derived features. It uses a composite kernel function and Gaussian process regression model combined with a Bayesian optimization algorithm to predict pavement performance and generate a pavement performance prediction curve for the next 1 to 5 years. The maintenance decision module is used to determine maintenance priorities, predict the trend of maintenance demand changes to generate a maintenance demand list, use the cuckoo search algorithm to generate a funding allocation plan, and determine the maintenance plan, implementation cost, and expected effect data; The maintenance benefit evaluation module uses the grey ideal solution and the super-efficiency DEA model to evaluate the improvement of pavement performance before and after maintenance, economic benefits and social benefits, and generates a maintenance benefit analysis report; The visualization and analysis module uses WebGL-based 3D visualization technology to display road condition distribution, disease location, and maintenance plan information. It uses Echarts-based interactive query and statistical analysis functions to support dynamic data updates and calculate comprehensive benefit indicators to assist decision-making.

[0121] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A scientific decision-making method for intelligent maintenance of highway pavement, characterized by: The method includes: S1. Collect multi-source data of basic data, auxiliary basic data and maintenance decision data on the public network and highway management intranet; pre-process the multi-source data; S2. Build an automated road detection system, input the road damage, flatness, rutting, and lateral force coefficient of the road section that needs maintenance into the automated road detection system, and the automated road detection system generates a highway technical condition assessment report based on the PQI and MQI values; S3, based on the highway technical condition assessment report in step S2, combined with historical test data, traffic volume data, and climate data, a Gaussian process regression model is used to predict road performance, and a road performance prediction curve for the next 1-5 years is generated; S4, according to the pavement performance prediction curve in step S3, a prediction curve, maintenance standards, and capital budget data are obtained, and the prediction curve, maintenance standards, and capital budget data are used to determine the maintenance priorities of different road sections by using ANP; Combined with the priorities determined by ANP, the hidden Markov model is applied to predict the changing trend of maintenance needs and generate a maintenance needs list; S5. Based on the maintenance demand list in step S4, a cuckoo search algorithm is used to generate a funding allocation plan to obtain maintenance plans, implementation costs, and expected effect data; S6. Input the maintenance plan, implementation cost, and expected effect data into the analysis system to evaluate the improvement of pavement performance before and after maintenance, the economic and social benefits of maintenance, and generate a maintenance benefit analysis report; S7, using WebGL-based 3D visualization technology to display road condition distribution, disease location, and maintenance plan information for the highway technical condition assessment report, pavement performance prediction curve, maintenance demand list, and funding allocation plan in S2-S5; Then use the interactive query and statistical analysis functions based on Echarts.

2. According to claim 1, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: Using density-based spatial clustering algorithm to clean and integrate the multi-source data to remove noise data; The local linear embedding algorithm is used to reduce the dimension of data; Among them, basic data include: map GIS data, road network information data, highway pavement and subgrade data, and pavement technical condition assessment data; Auxiliary basic data include: traffic flow data, meteorological data, geological environment data, and administrative management data; Maintenance decision data: maintenance material category data, maintenance equipment data, maintenance method data, and maintenance project unit price data.

3. According to claim 1, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: Use Python-based Django framework to build an automated road detection system; Use the jango - admin startproject command to create a Django project, and then execute the python manage.pystartapp command to create a specific application; Configure database connection information in the project configuration file, and select MySQL or SQLite database; define the database model for road detection data and indicator calculation, including road damage, flatness, rutting, lateral force coefficient detection data, and PQI, MQI indicator data, and execute python manage.py makemigrations and pythonmanage.py migrate commands to create the database table corresponding to the database model; Write Python scripts to collect data on road damage, flatness, rutting, and lateral force coefficients for sections that need maintenance; use Django's REST framework to develop an API interface to receive the collected data and store it in a database; Write Python code in Django application to implement rough set theory algorithm and build information system with collected data ,in is a collection of objects, ,in is the set of conditional attributes, i.e., various detection indicators. is the decision attribute set, i.e., the overall performance of the highway. is a set of attribute values, is the information function; Attribute reduction is performed by calculating the difference matrix, the formula is: （ ）； Calculating attribute importance ,in , ; Finally, according to Determine the weight of each indicator; The collected raw data of PCI, RQI, RDI, and PBI that constitute the PQI are standardized again; Among them, PCI is the pavement damage index, RQI is the pavement ride quality index, RDI is the pavement rutting depth index, and PBI is the pavement skid resistance index; According to Calculate the PQI value; For MQI, the original data of its constituent indicators PQI, BCI, TCI, and SCI are subjected to the same standardization process, and the rough set theory is used again to determine the weight of each indicator. Calculate the MQI value; Among them, PQI is the pavement performance index, MQI is the highway technical condition index; Based on the calculated PQI and MQI values, combined with the pre-set evaluation grade standards, a highway technical condition evaluation report including basic information of the road section, indicator calculation results, evaluation grade, analysis and maintenance suggestions is generated in the Django application; Use Matplotlib and Echarts libraries to visualize the assessment results; use the Django template system to develop a Web interface to view data and assessment report information; Use the Django testing framework to test each functional module of the system to ensure that data collection, calculation, and report generation functions are normal; Deploy the system to a cloud server, using Nginx as the web server and Gunicorn as the application server.

4. According to claim 1, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: The specific steps of step S3 are: S31. Collect the PQI, MQI and sub-item index values ​​in the highway technical condition assessment report of step S2, and integrate them with the collected historical test data, traffic volume data, and climate data; use the method based on local outlier factors to detect outliers. , through the formula Calculate the local reachability density using the formula Calculate the local outlier factor when Process outliers when they are greater than the set threshold; in yes of The nearest neighbor sample set, , It is a sample To its The distance to the nearest neighbor, It is a sample and The Euclidean distance between S32, using composite kernel function , for the training data set , calculate the covariance matrix , whose elements , considering the observation noise Get the covariance matrix with noise ; S33. Define the log-likelihood function for Gaussian process regression , using the Bayesian optimization algorithm to find the optimal hyperparameters ; S34. Generate forecast input data based on the time span of the next 1-5 years and the trend of historical data changes , calculate the covariance matrix between the prediction input and the training data and the covariance matrix of the prediction input itself , through the formula Calculate the predicted mean using the formula Calculate forecast covariance; S35. Based on the prediction covariance matrix The confidence interval of the predicted value is calculated, and the predicted mean and confidence interval are plotted over time to form a pavement performance prediction curve for the next 1-5 years.

5. According to claim 4, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: The specific steps of step S4 are: S41, collect the pavement performance prediction curve data, maintenance standards and capital budget data of step S3; use the method based on the interquartile range to process the outliers, and use the formula Determine the outlier judgment interval , linear interpolation method is used to correct outliers; Use the min-max normalization formula to normalize the pavement performance prediction curve data; S42. Construct a network structure model including pavement performance, traffic impact, and economic cost element sets, analyze the interdependence between elements; construct a judgment matrix through expert scoring method ; Solve Get the eigenvectors and combine them into an unweighted supermatrix ; Determine the element set association weight matrix , calculate the weighted supermatrix ;right Perform power calculation Get the limit supermatrix , determine maintenance priorities; S43. Define the maintenance requirement status set and the observation set ; Initialize the state transition probability matrix , observation probability matrix and the initial state probability vector ; Using the Baum-Welch algorithm, the forward probability recursive formula And the backward probability recursion formula Train the model and update the parameters ; Using the Viterbi algorithm, through the recursive formula Find the most likely sequence of states; S44. Combine the priorities determined by ANP and the trends predicted by the hidden Markov model, generate a maintenance demand list including road section information, maintenance time, maintenance demand status, and estimated required funds according to the road section number and time sequence, and calculate the maintenance demand list according to the capital budget formula. Adjust and screen maintenance items.

6. According to claim 5, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: The specific steps of step S5 are: S51, extracting the road section maintenance demand information and estimated required funds from the maintenance demand list of step S4 and maintenance priorities , clarify the total amount of funds budget and the annual funding allocation ratio ; Define the objective function ,in: The maintenance benefit function is defined as ; The maintenance cost function is defined as: ; Setting Constraints and priority penalty ; S52, in the cuckoo search algorithm step, a solution with the largest objective function value is selected as the optimal solution; Generate a funding allocation plan based on the optimal solution, determine the maintenance section and allocated funds, clarify the maintenance plan; and calculate the total implementation cost and expected effect data .

7. According to claim 6, a scientific decision-making method for intelligent maintenance of highway pavement, characterized in that: step S6 The maintenance plan, implementation cost, and expected effect data are input into the analysis system, which uses the grey ideal solution to evaluate the improvement of pavement performance before and after maintenance; The analysis system then uses the super-efficiency-based DEA model to evaluate the economic and social benefits of maintenance and generate a maintenance benefit analysis report.

8. According to claim 7, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: The specific steps of step S6 are: S61, collecting the maintenance plan, implementation cost, and expected effect data in steps S51 and S52 and entering them into the analysis system; standardizing the original data matrix to eliminate the dimensional effect; S62, determining a standardized data matrix and calculating a road performance improvement index; S63. Use the super-efficiency DEA model to evaluate economic and social benefits: Each maintenance plan is used as a decision-making unit to determine the input implementation cost and output economic and social benefit indicators; a super-efficiency DEA model is established, and the objective function is ; The constraints are: , , , and solve for the super-efficiency value ; in For the DMU Input indicator values, For the DMU Output indicator value, and are the weights of output and input indicators respectively, is a non-Archimedean infinitesimal; S64. Report includes cover page and table of contents, project overview, and pavement performance improvement assessment , economic and social benefit evaluation , comprehensive evaluation and suggestions, and appendix.

9. According to claim 8, a scientific decision-making method for intelligent maintenance of highway pavement is characterized by: Integrate the highway technical condition assessment report, pavement performance prediction curve, maintenance demand list and funding allocation plan in steps S2-S5; convert the geographical location information of the road section into geographical coordinates suitable for WebGL rendering; Use the WebGL library to create a 3D scene and load the geographic basemap; create a 3D model of the road segment based on the geographic location of the road segment, using the formula Mark the location of the disease with a specific geometric shape; in is the base size, is the standardized disease severity data, is the scaling factor, which is used to control the range of disease mark size; According to the maintenance needs list and funding allocation plan, maintenance measures are represented by different colored lines or textures, and the formula Use animation to show maintenance progress; in and are the starting color and the ending color respectively. progress is the progress of the maintenance plan; Bind the integrated data with Echarts to create bar charts, line charts, and pie charts; add interactive events and query input boxes to the charts to implement interactive queries; provide summation and average statistical analysis functions, and display the statistical results in new charts or tables; set the data update cycle to implement dynamic data updates; Using the formula Calculate comprehensive benefit indicators; in , and is the weight coefficient, which is determined by expert scoring or historical data statistics. is the improvement value of pavement performance PQI, cost is the maintenance cost, It is the quantitative value of social benefits; Provide decision-makers with suggestions on maintenance plan adjustment and fund allocation optimization based on PQI and MQI indicators.

10. A scientific decision-making method for intelligent maintenance of a highway pavement according to any one of claims 1 to 8, characterized in that: The decision-making system of the decision-making method includes a data collection and preprocessing module, a road automation detection system, a pavement performance prediction module, a maintenance decision module, a maintenance benefit evaluation module, and a visualization and analysis module; The data collection and preprocessing module collects multi-source data of basic data, auxiliary basic data and maintenance decision data from the public network and highway management intranet, and uses density-based spatial clustering algorithm for cleaning and integration, and adopts local linear embedding algorithm for data dimension reduction processing; The road automation detection system is built with the Python-based Django framework, which uses the rough set theory algorithm to determine the detection indicator weights and calculate the PQI and MQI values ​​for data collection, indicator calculation, report generation and display, system testing and deployment functions; The pavement performance prediction module is used to integrate relevant data, process outliers and construct derived features. It uses a composite kernel function and Gaussian process regression model combined with a Bayesian optimization algorithm to predict pavement performance and generate a pavement performance prediction curve for the next 1 to 5 years. The maintenance decision module is used to determine maintenance priorities, predict the trend of maintenance demand changes to generate a maintenance demand list, use the cuckoo search algorithm to generate a funding allocation plan, and determine the maintenance plan, implementation cost, and expected effect data; The maintenance benefit evaluation module uses the grey ideal solution and the super-efficiency DEA model to evaluate the improvement of pavement performance before and after maintenance, economic benefits and social benefits, and generates a maintenance benefit analysis report; The visualization and analysis module uses WebGL-based 3D visualization technology to display road condition distribution, disease location, and maintenance plan information. It uses Echarts-based interactive query and statistical analysis functions to support dynamic data updates and calculate comprehensive benefit indicators to assist decision-making.

Images