Asphalt pavement green maintenance decision-making method and system based on long-term performance monitoring

Abstract

The invention relates to the technical field of data processing, and discloses an asphalt pavement green maintenance decision-making method and system based on long-term performance monitoring, which are used for solving the problems of low decision-making efficiency and insufficient resource utilization caused by the fact that global optimum is difficult to solve by multi-objective optimization under budget constraint in a traditional method. The method comprises the following steps: firstly, acquiring and sorting pavement performance data through a data acquisition and storage module, establishing a performance attenuation model and a green index system, and forming a multi-objective function framework; setting a budget hard constraint and a target weight, and performing initial solution space division and iterative adjustment, including boundary expansion and cross validation, on a function; and finally, outputting a decision scheme and applying the decision scheme to pavement management to realize balance optimization of cost, performance and environmental influence. The decision-making efficiency is improved, and sustainable maintenance is supported.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a decision-making method and system for green maintenance of asphalt pavement based on long-term performance monitoring. Background Technology

[0002] As a core component of highway transportation infrastructure, the long-term performance monitoring and green maintenance decisions of asphalt pavement directly affect the safety, durability, and environmental sustainability of roads. With increasing traffic loads and intensified climate change, traditional pavement maintenance methods are no longer sufficient to meet the requirements of high efficiency and low carbon emissions. In existing technologies, performance degradation models are typically established by collecting data on pavement structure, load, and environment, and maintenance decisions are made in conjunction with multi-objective optimization algorithms to achieve a balance between minimizing costs, maximizing performance, and minimizing environmental impact. However, under budget constraints, multi-objective optimization often faces the technical challenge of finding the global optimum, leading to low decision-making efficiency and resource waste.

[0003] For example, Chinese patent CN113030450A discloses a method and system for monitoring and evaluating the full-cycle performance of asphalt pavement. This system optimizes the selection of maintenance schemes through real-time data acquisition and performance degradation curve plotting. This method emphasizes monitoring the pavement's life cycle, but its decision-making process relies on the analytic hierarchy process (AHP) and network-level optimization. When budgets are limited, it needs to consider multiple objective functions such as cost, performance, and sustainability simultaneously. However, because the optimization problem is NP-hard, this method struggles to efficiently solve for the global optimum and is prone to getting trapped in local optima. This results in maintenance strategies in practical applications failing to fully consider green indicators, such as minimizing carbon emissions, thus increasing long-term environmental burden and additional economic costs. (Patent CN114...) Chinese patent 118800A discloses a method for evaluating the condition of asphalt pavement and selecting maintenance materials based on the law of service performance degradation. This method achieves dynamic adjustment of maintenance decisions by establishing a deterioration model and a green material database. This patent highlights the green attributes of preventive maintenance, but its multi-objective optimization framework also has the bottleneck of finding the global optimum under budget constraints. Specifically, when the objective function includes economic cost, material durability and environmental impact, the computational complexity of the algorithm increases exponentially, and it is impossible to obtain the global optimum solution in a reasonable time. This makes the decision-making process rely on experience adjustment, which reduces the reliability and universality of the system. Especially in areas with limited resources, it is easy to cause delays in maintenance or excessive intervention, amplifying the risk of pavement distress.

[0004] In summary, while existing technologies have made progress in long-term performance monitoring and green maintenance decision-making for asphalt pavements, they generally suffer from the difficulty of finding the global optimum when performing multi-objective optimization under budget constraints. This problem stems from the non-convexity and high dimensionality of the optimization problem, leading to excessive consumption of computational resources and insufficient exploration of the solution space, ultimately affecting the scientific nature and sustainability of the decision-making. Therefore, a new method is needed to overcome this technical challenge and improve the efficiency and greenness of maintenance decisions. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a decision-making method and system for green maintenance of asphalt pavements based on long-term performance monitoring. This method solves the technical problems of low decision-making efficiency and insufficient resource utilization caused by the difficulty in finding the global optimum in multi-objective optimization under budget constraints in traditional methods.

[0007] (II) Technical Solution

[0008] To achieve the goals of improving the efficiency and environmental friendliness of maintenance decisions mentioned in the background section, this invention provides the following technical solution:

[0009] Decision-making methods for green maintenance of asphalt pavements based on long-term performance monitoring include:

[0010] S1, collect long-term performance monitoring data of asphalt pavement, classify and organize the data and store it in the database;

[0011] S2. Based on the stored data, a pavement performance degradation model and a green maintenance index system are established. By integrating performance indicators, environmental factors and cost parameters, a multi-objective function framework is formed.

[0012] S3, based on the established framework, sets budget constraints, incorporates the budget ceiling as a hard constraint into the multi-objective function, and defines the specific objective weights for maximizing performance, minimizing environmental impact, and minimizing cost.

[0013] S4. Using the set constraints and weights, the initial solution space of the multi-objective function is partitioned.

[0014] S5, within the partitioned solution space, performs iterative adjustments, and performs boundary expansion and cross-validation for each feasible solution;

[0015] S6, based on the solution set after iteration, outputs the final maintenance decision scheme and applies the selected scheme to actual road management.

[0016] In a preferred embodiment, long-term performance monitoring data of asphalt pavement is collected, the data is classified, organized, and stored in a database, including:

[0017] A sensor network was deployed on a selected asphalt pavement monitoring section to collect various types of data.

[0018] Data is transmitted to the edge gateway via a wireless module for synchronization and then uploaded to the cloud.

[0019] The data is subject-based, derived metrics are calculated, and data cleaning is performed, including outlier removal, noise interpolation, and missing data labeling.

[0020] The cleaned data is written to a distributed database in a structured format. The main table is linked to multiple sub-tables through foreign keys, and indexes are created.

[0021] In a preferred embodiment, a pavement performance degradation model and a green maintenance index system are established based on stored data. By integrating performance indicators, environmental factors, and cost parameters, a multi-objective function framework is formed, including:

[0022] Extract structural performance, service intensity, and external impact data corresponding to pavement segment numbers and time series from the database;

[0023] Establish a pavement performance degradation model to simulate the strength variation with time and load;

[0024] Establish a green maintenance indicator system to quantify carbon emissions and material recycling rates;

[0025] Integrating performance indicators, environmental factors, and cost parameters, a vector-based multi-objective function framework is formed.

[0026] In a preferred embodiment, based on the established framework, budget constraints are set, incorporating the budget ceiling as a hard constraint into the multi-objective function. Simultaneously, specific objective weights for maximizing performance, minimizing environmental impact, and minimizing cost are defined. Specifically, the implementation is as follows:

[0027] Extract the budget ceiling value from the database and embed it into a multi-objective function framework;

[0028] Add a new cost constraint vector and set a linear summation of the total cost of candidate solutions;

[0029] Schemes exceeding the budget limit are eliminated; performance and green maintenance indicators are calculated only for schemes that meet the conditions.

[0030] Weights are determined and adjusted based on expert scoring to reflect priority.

[0031] Scores are summarized and normalized according to the target categories, and weight sets are formed after reconciling the differences;

[0032] The weights are multiplied by the corresponding target vectors and then summed to form a weighted composite target vector.

[0033] In a preferred embodiment, the initial solution space of the multi-objective function is partitioned using set constraints and weights, including:

[0034] Load the budget hard constraints and target weight vector from the configuration file;

[0035] The core parameter set is discretized into gridded intervals to construct a high-dimensional solution space;

[0036] Filter grid points layer by layer and eliminate schemes that exceed the limits;

[0037] Calculate the normalized weighted composite score and retain high-potential points in order;

[0038] Perform cross-dimensional consistency verification to further converge the solution set;

[0039] Set up a feedback adjustment mechanism to fine-tune the grid boundaries and re-filter;

[0040] The refined solution set is stored in a structured format in the database.

[0041] In a preferred embodiment, within the partitioned solution space, iterative adjustments are performed, including boundary expansion and cross-validation for each feasible solution, including:

[0042] Load a dynamic solution space structure from the refined feasible solution set, and select representative schemes hierarchically based on the comprehensive score;

[0043] Perform a boundary expansion operation around the center point to generate an expanded variant;

[0044] Construct a pairing matrix to perform cross-validation, calculate compatibility metrics, and eliminate low-compliance variants;

[0045] Fine-tune the parameters of the retained variants, adjust the relevant parameters in conjunction, and calculate the composite score;

[0046] Perform a global comparison and merge on the adjusted variants, sort by score and merge similar combinations;

[0047] The optimized solution set is stored in a structured format in the database.

[0048] In a preferred embodiment, based on the iteratively obtained solution set, a final maintenance decision scheme is output, including:

[0049] Load the parameter combinations and score list from the optimized solution set file;

[0050] Perform high-level filtering on the list, calculate a score threshold, and retain combinations that are above the threshold;

[0051] Examine the diversity of parameter distributions and supplement alternatives from combinations with lower scores;

[0052] The final maintenance plan is selected by prioritizing the options based on performance, environment, and cost.

[0053] Implementation information is extracted from parameters to estimate construction windows and resource quantification, and a timetable is generated.

[0054] In a preferred embodiment, the selected scheme is applied to actual road management, including:

[0055] The detailed data package is output to the user interface, displaying a layered overview of the solution, schedule, resource panel and effect comparison. It supports users to simulate fine-tuning parameters and refresh in real time, and records confirmation operations and feedback logs.

[0056] The confirmed plan is converted into a structured format and uploaded to the central server of the road management system to update the road segment status fields;

[0057] The decomposition plan is converted into operation instructions, which send task notifications to the construction terminal, generate allocation instructions to the inventory subsystem, and collect maintenance data from the monitoring subsystem and compare it with expectations.

[0058] Generate an execution confirmation file, summarizing the scheme details, import logs, and execution results, and store them back in the database in a structured format and establish a relationship.

[0059] On the other hand, the present invention provides a decision-making system for green maintenance of asphalt pavement based on long-term performance monitoring, comprising:

[0060] Data acquisition and storage module: responsible for deploying sensor networks to collect road surface deformation, traffic load and environmental data, classifying and organizing the data, cleaning outliers, and storing the data in a structured format in the database;

[0061] Model and framework building module: Based on the stored data, extract key datasets, establish a performance degradation model and a green maintenance index system, and integrate performance, environmental and cost parameters to form a multi-objective function framework;

[0062] Constraint and Weight Setting Module: Introduces hard budget constraints on top of the functional framework and defines the weights of performance, environmental, and cost objectives;

[0063] Solution space partitioning module: The function is divided into gridded intervals using constraints and weights. Invalid intervals are eliminated by filtering layer by layer, and the set of feasible solutions is converged.

[0064] Iterative adjustment module: Selects representative solutions within the feasible solution space, performs boundary expansion and cross-validation, and optimizes parameter combinations to cover more paths;

[0065] Decision Output and Application Module: Based on the iterative solution set, the highest-scoring solution is selected, its details are extracted and output to the interface, and imported into the management system to guide the execution of the solution through API linkage with on-site equipment.

[0066] Compared with existing technologies, this invention provides a decision-making method and system for green maintenance of asphalt pavement based on long-term performance monitoring, which has the following beneficial effects:

[0067] 1. This invention, through long-term continuous collection of multi-source monitoring data such as structural response, traffic load, and environmental effects on typical road sections, establishes a performance degradation model coupled with traffic and climate factors, and constructs a green maintenance index system with carbon emissions, material recycling rate, and ecological restoration as its core. Performance, environmental, and cost objectives are unified into a single multi-objective function framework, with a hard constraint of a budget ceiling introduced within this framework. Target weights obtained through expert scoring quantitatively characterize the importance of different objectives. Combined with grid-based solution space partitioning, pre-budget filtering, and hierarchical screening, along with iterative adjustment mechanisms such as boundary expansion, cross-validation, and parameter fine-tuning, the search range of the high-dimensional non-convex solution space is effectively narrowed while ensuring the total cost does not exceed the predetermined budget. This improves the efficiency of finding globally optimal solutions and avoids the optimization process getting stuck in local optima and relying on experience-based decisions. Thus, under limited resource conditions, a coordinated balance is achieved between pavement performance, maintenance costs, and green indicators. This significantly improves the technical problems in existing technologies where multi-objective optimization under budget constraints struggles to find the global optimum, leading to low decision-making efficiency and insufficient resource utilization.

[0068] 2. This invention, by deploying an integrated sensor network on representative road sections, continuously acquires full-process monitoring data on pavement structure response, traffic load, and environmental effects. This data is then refined in a database according to road section, time, and indicators. These long-term sequences are input into a performance degradation model and a green maintenance indicator system, enabling quantitative prediction and graded assessment of pavement remaining life, deterioration rate, and green level. This transforms the maintenance triggering conditions from a crude model relying on manual inspections and experience-based judgment to a refined and forward-looking decision-making process based on monitoring data and model output. Simultaneously, the implementation results of the maintenance plan are used to write back the model parameters through monitoring data, forming a closed-loop iteration of decision-making and execution. This improves the accuracy of deterioration trend identification and the rationality of maintenance timing selection, reduces over-maintenance and delayed maintenance, extends pavement service life, and reduces disturbance to traffic operations. Attached Figure Description

[0069] Figure 1 This is a flowchart of the green maintenance decision-making method for asphalt pavement based on long-term performance monitoring, as per the present invention.

[0070] Figure 2 This is a schematic diagram of the green maintenance decision system for asphalt pavement based on long-term performance monitoring, as per the present invention. Detailed Implementation

[0071] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] Example 1: Figure 1 A decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring is presented, including:

[0073] S1, collect long-term performance monitoring data of asphalt pavement, classify and organize the data and store it in the database;

[0074] S2. Based on the stored data, a pavement performance degradation model and a green maintenance index system are established. By integrating performance indicators, environmental factors and cost parameters, a multi-objective function framework is formed.

[0075] S3, based on the established framework, sets budget constraints, incorporates the budget ceiling as a hard constraint into the multi-objective function, and defines the specific objective weights for maximizing performance, minimizing environmental impact, and minimizing cost.

[0076] S4. Using the set constraints and weights, the initial solution space of the multi-objective function is partitioned.

[0077] S5, within the partitioned solution space, performs iterative adjustments, and performs boundary expansion and cross-validation for each feasible solution;

[0078] S6, based on the solution set after iteration, outputs the final maintenance decision scheme and applies the selected scheme to actual road management.

[0079] S1. Collect long-term performance monitoring data of asphalt pavement, classify and organize the data, and store it in the database. The specific implementation is as follows:

[0080] Representative asphalt pavement monitoring sections were selected from the target highway. An integrated sensor network was deployed in each monitoring section, including embedded and surface sensors. The embedded sensors included fiber Bragg grating strain sensors, platinum resistance temperature sensors, and capacitive humidity sensors. The strain sensors were placed at the interface between the bottom of the asphalt pavement surface layer and the top of the base layer, with a group installed every 50m along the longitudinal direction of the pavement. Each group contained sensor units in the transverse, longitudinal, and oblique directions to acquire the tensile, compressive, and shear deformation of the pavement under vehicle loads. The temperature and humidity sensors were vertically embedded in the middle of the surface layer, the middle of the base layer, and the top of the subbase layer, respectively, at depths of 5cm, 20cm, and 40cm. They were used to monitor the temperature field distribution and moisture content in different structural layers, providing basic data for analyzing the long-term changes in material modulus and pavement strength caused by temperature and humidity coupling.

[0081] The surface sensors include a laser scanner and a vehicle load counter. The laser scanner is fixedly installed on the shoulder guardrails on both sides of the monitoring section, at a height of approximately 2.5 meters, with a field of view covering the entire width of the road surface. It collects point cloud data of the road surface elevation using a high-frequency pulse mode. Subsequently, by performing surface fitting on the point cloud data and combining it with the elevation change threshold, the surface crack boundaries and rut shape are extracted, and indicators such as road surface smoothness index, crack width, and rut depth are calculated. The vehicle load counter uses a piezoelectric axle load sensor, with the sensing element buried approximately 5 mm below the lane markings. Two rows of sensors are arranged along the driving direction in each lane. The amplitude and time interval of the sensing signal are used to identify vehicle passing events, distinguish vehicle axle type, axle load, and vehicle speed, and record a precise timestamp for each vehicle passing, thus forming a continuous traffic load record.

[0082] Each sensor is connected to the roadside edge computing gateway via a low-power wide-area network wireless communication module. The communication method can adopt narrowband IoT protocol or long-range wireless communication protocol to meet the connectivity needs of remote road sections. Each sensor node is equipped with a solar panel and lithium battery power supply combination. The solar panel area is, for example, 0.2 square meters, and the tilt angle is set to 35 degrees according to the latitude of the monitoring area to ensure high light energy utilization efficiency throughout the year. The data acquisition module adopts a dynamic frequency strategy: when no vehicle load event is detected, it operates in environmental monitoring mode, collecting background data such as temperature, humidity and static strain every 5 minutes; when the piezoelectric signal amplitude output by the vehicle load counter exceeds the preset trigger threshold, it is determined that a vehicle passage event has occurred, and the system immediately switches to high-frequency mode, synchronously collecting strain, axle load and vehicle speed data at an interval of about 0.1 seconds, thereby recording the transient response of the entire process of vehicle load action while ensuring controllable energy consumption.

[0083] The edge computing gateway is deployed in a roadside cabinet to manage the cache and time synchronization of the uploaded raw monitoring data. The gateway connects to the timing module of the BeiDou or Global Positioning System and uses the timing signal to correct the timestamps of the data reported by each sensor node, so that the data of different nodes are aligned on a unified time axis and the time error is controlled within the millisecond range. This ensures that the data from multiple sources can be accurately matched for the same vehicle event or the same environmental process. When the network is normal, the gateway sends the cached data to the cloud data center in batches according to the preset upload cycle. When a network anomaly or communication link interruption is detected, the gateway continuously caches the data locally for a certain period of time and triggers the retransmission process after the link is restored to avoid irrecoverable data loss.

[0084] In the cloud data center, the received raw data is first classified by topic, and corresponding derived indices are calculated. Data related to pavement structure deformation and surface distress geometry are categorized into the structural performance group, including transverse strain, longitudinal strain, shear strain, and indices such as crack width, crack length, and rut depth obtained from point cloud processing. Data related to single vehicle load and traffic flow characteristics are categorized into the service intensity group, including actual axle load, vehicle speed, vehicle type, and equivalent axle load calculated based on standard axle load. The cumulative equivalent axle load is calculated as follows: based on the various vehicle axle types and axle loads identified by the vehicle load counter, each axle load is compared with the design standard axle load. The axle load ratio is obtained by division. The axle load ratio is then converted to a power of 1 using a preset power rule, such as the fourth power rule. Finally, it is weighted and summed based on the number of times the corresponding vehicle passes through to obtain the cumulative equivalent axle load within the monitoring time interval. Data related to environmental and internal climate conditions are classified into the external influence group. The daily temperature difference is obtained by calculating the difference between the highest and lowest temperatures of the same sensor within 24 hours. The number of freeze-thaw cycles is obtained by counting the number of times the temperature passes back and forth between 0°C and 0°C. The moisture saturation is obtained by converting the volumetric water content measured by the humidity sensor with the porosity of the structural layer. This is used to reflect the impact of environmental changes on the road surface aging process.

[0085] After classification, a unified data cleaning and quality control process is performed on all types of data. First, outlier thresholds are set based on physically reasonable ranges. For example, records with absolute strain values ​​exceeding 5000 microstrain, temperatures below -40℃ or above 80℃, or humidity less than 0% or greater than 100% are considered inconsistent with actual working conditions, directly marked as outliers, and removed from the analysis dataset. Then, a continuity verification method based on a sliding time window is used. The window average value of the indicator is calculated within a time window matching the sampling frequency, using monitoring segments and indicator types as units. In environmental monitoring mode, the window length is, for example, 5 minutes; in high-frequency monitoring mode, the window length is set to several... Aggregate data across sampling periods; when the deviation of the observed values ​​at three consecutive time points from the average value of the corresponding window exceeds 20%, the data at these three time points are identified as noise points, and linear interpolation is performed using the two most recent valid data points to reduce the interference of short-term sudden noise on long-term trend analysis; for timestamp integrity checks, the data sequences of each sensor are traversed along a unified time axis, and when there is a time interval of more than 1 minute without data reporting in a certain index sequence of the same monitoring segment, the time interval is marked as a missing segment, and the missing flag is recorded in the database through a dedicated status field for easy identification and processing in subsequent modeling and statistical analysis;

[0086] The cleaned data is written to a structured cloud-based distributed database. The database contains a main table and multiple sub-tables. The main table uses the monitoring segment number and data collection timestamp as a composite primary key, recording basic information such as road mileage markers, location coordinates, lane numbers, and sensor node numbers. The structural performance sub-table stores fields for lateral strain, longitudinal strain, shear strain, and surface deformation such as crack width, crack length, and rut depth. The usage strength sub-table stores fields for single-axle load, cumulative equivalent axle loads, traffic flow, and vehicle speed statistics. The external influence sub-table stores fields for temperature, humidity, daily temperature range, freeze-thaw cycle count, and moisture saturation. Each sub-table is linked to the main table via foreign keys. Indexes are created on the monitoring segment number, data collection timestamp, and data type fields, supporting fast queries by time interval, road segment range, and indicator category. For example, all data from the section of a highway from K100+000 to K100+500 can be partitioned by day, making it easier to archive and statistically analyze data monthly or yearly. The database is deployed with master and standby nodes and configured with a real-time synchronization mechanism, while also incorporating a regular off-site backup strategy to improve data storage security and system fault recovery capabilities.

[0087] Through the aforementioned sensor deployment, data acquisition, time synchronization, classification and organization, derived index calculation, and cleaning and storage process, the constructed database covers typical asphalt pavement structures in the spatial dimension, continuously records the entire service process of the pavement in the temporal dimension, and comprehensively represents structural response, traffic load, and environmental effects in the index dimension, providing a complete and reliable data foundation for the subsequent establishment of pavement performance degradation models and green maintenance index systems.

[0088] S2, based on the stored data, establish a pavement performance degradation model and a green maintenance index system. By integrating performance indicators, environmental factors, and cost parameters, a multi-objective function framework is formed, specifically implemented as follows:

[0089] Long-term performance monitoring data is extracted from the cloud database constructed earlier. First, using pavement segment number and time series as query conditions, structural performance group data is retrieved to obtain parameters such as transverse strain, longitudinal strain, shear strain, and surface crack propagation rate. These parameters originate from the monitoring results of embedded sensors and laser scanners and are aggregated according to monthly or quarterly time scales to form deformation sequences reflecting the long-term evolution trend of each monitoring segment. Simultaneously, usage intensity group data is extracted, including parameters such as single-axle load, cumulative equivalent axle load, vehicle speed distribution, and peak-hour traffic flow. By filtering peak-hour records, the usage of different pavement segments under high-load conditions is characterized. Intensity characteristics; further extraction of external impact group data, where daily temperature difference, freeze-thaw cycle number and moisture saturation are calculated based on data recorded by temperature and humidity sensors, and annual average rainfall is obtained through data interface with the regional meteorological monitoring system. The annual average is calculated based on the daily rainfall records of the meteorological station corresponding to the monitoring section and written into the environmental impact data table; during extraction, the data are associated by road segment number and time interval, so that deformation data, traffic load data and environmental impact data are one-to-one in time and space, forming a key dataset with road segment and observation period as rows and various parameters as columns, providing a unified input for subsequent modeling;

[0090] A pavement performance degradation model is established based on key datasets. To introduce strength indicators that reflect the structural bearing capacity, pavement strength reference data is first obtained. For example, falling weight deflectometers are set up at typical monitoring sections, and field deflection tests are carried out at different service stages. The equivalent resilient modulus of each monitoring section is obtained through structural back-calculation. Alternatively, laboratory mechanical tests are conducted on pavement core samples to calculate the compressive strength and elastic modulus of each structural layer. The strength indicators are written into the strength indicator table according to the pavement section number and observation time, and associated with the deformation data and traffic load data in the monitoring database. It is preferred to match the monitoring data within a preset time window before and after the strength test date so that the strength indicators correspond to long-term monitoring records of the same service stage.

[0091] Based on the alignment of strength indicators with long-term monitoring data, the cumulative equivalent axle number is used as the main independent variable, and the equivalent resilient modulus or other strength indicators are used as the dependent variable. Distinguishing between different pavement types such as highways and urban roads, and grouping high-traffic and low-traffic sections, regression fitting or piecewise function fitting is performed on the aggregated monthly or quarterly strength indicators and the cumulative equivalent axle number sequence to obtain a basic model describing the gradual decline in strength with accumulated load. To reflect the influence of environmental factors on the attenuation process, freeze-thaw cycles, moisture saturation, and annual average rainfall are introduced into the model as environmental moderating variables. By setting coefficients related to environmental variables in the parameters of the basic attenuation model, or constructing a multiple regression model with cumulative equivalent axle number and environmental variables as multiple independent variables, a larger attenuation rate parameter can be fitted under high humidity, high rainfall, or high freeze-thaw frequency conditions. In areas where low temperature has a significant effect, the weight coefficient related to brittle failure is increased, thus forming a multi-factor pavement performance attenuation model that comprehensively considers traffic load and environmental effects. The deterioration rate and expected remaining life are output on a per-pavement-segment basis to determine the maintenance urgency of different monitoring segments.

[0092] While establishing the attenuation model, a green maintenance indicator system is constructed. Based on the process type, material ratio, and construction procedures of different maintenance schemes, carbon emission indicators are defined. By statistically analyzing the material consumption and machinery operation time in each scheme, the corresponding carbon emission factors per unit material and per unit procedure are obtained from tables. These emission factors are pre-established based on current carbon emission accounting standards or industry statistical data to create parameter tables, and then the total carbon emissions of each scheme are calculated. A material recycling rate indicator is also defined. Based on the ratio of recycled waste materials to total material usage in each scheme, the closed-loop efficiency from road milling to recycling is quantified. For example, a recycling rate of 50% is used as a benchmark value, and values ​​higher than this benchmark value are considered as follows. A portion is included in the positive contribution; an ecological restoration indicator is defined, which is calculated by conducting on-site surveys or remote sensing analysis of the vegetation coverage and soil erosion status around the monitoring section before and after maintenance, and assessing the effectiveness of soil erosion control. The results are then normalized into an ecological restoration score; the above three indicators of carbon emissions, material recycling rate and ecological restoration are integrated into a comprehensive green maintenance indicator by weighted summation. The weight coefficients are preset by the road maintenance management department in combination with environmental protection policy objectives, resource utilization requirements and management experience, or determined by expert scoring methods, and stored and retrieved in the form of a parameter table, so that the green weight configuration can be flexibly adjusted in different regions and at different stages;

[0093] Based on the pavement performance degradation model and the green maintenance index system, a multi-objective function framework is formed. First, the deterioration rate output by the degradation model is linked with the pavement performance index. Among them, the pavement smoothness index is calculated from the longitudinal profile elevation data obtained by laser scanning mentioned above, and the international smoothness index is converted according to the smoothness evaluation standard. This index and its rate of change are used as core performance variables reflecting driving comfort and structural service level. In the multi-objective optimization, it is required that the degree of deviation from the preset limit be minimized. Second, the green maintenance comprehensive index is included as part of the environmental objectives, and carbon emissions and resource recycling levels are comprehensively considered. The annual average rainfall and temperature range are not used as separate optimization objectives, but as external parameters describing regional climate characteristics. They are used to adjust the weights of water-related and low-temperature-related factors in the green index, so that water-resistant schemes are given priority in hot and humid areas, and freeze-thaw-resistant schemes are given priority in cold areas.

[0094] Furthermore, the maintenance cost, comprising material procurement costs, machinery usage costs, and construction labor costs, is used as the economic objective. By breaking down and summarizing each cost item according to the work process, a total cost expression is constructed. The performance objective vector, the green objective vector, and the cost objective vector are organized in the same optimization framework. Each objective vector contains corresponding indicators and weight coefficients, forming a set of multi-objective functions that can be used for solving. The optimization module can use a weighted summation method to transform the multi-objective into a single objective for solution, or it can use a multi-objective optimization algorithm based on the Pareto front to search for a set of non-dominated solutions. Under given budget constraints and weight settings, it provides the optimal or suboptimal combination of performance, greenness, and cost for different maintenance schemes, providing a structured decision-making basis for formulating specific maintenance schemes and allocating funding budgets in subsequent steps.

[0095] S3, based on the established framework, sets budget constraints, incorporating the budget ceiling as a hard constraint into the multi-objective function, and defines specific objective weights for maximizing performance, minimizing environmental impact, and minimizing cost. The specific implementation is as follows:

[0096] Starting from the vector-based multi-objective function framework established earlier, budget constraints are introduced to enhance the practical feasibility of decision-making and resource control. First, the source and method of retrieving budget data are determined. The upper limit of the budget is extracted from the project budget table pre-stored in the cost management module of the cloud database. The budget table is stored in a standardized spreadsheet format, recording the total annual budget, material costs, labor costs, and other subcategories, as well as the corresponding road section number. The system queries the corresponding records according to the current decision-making cycle through an interface. For example, for a highway maintenance project, the annual budget upper limit is read as 5 million yuan. This value is an example value; the actual budget upper limit is determined based on the financial allocation items in the project planning document. During the retrieval, the version number and update time of the budget table are simultaneously verified. If the update time is earlier than a preset threshold for the current decision date, a prompt mechanism is triggered to require updating the budget data, ensuring the accuracy and timeliness of the budget input.

[0097] The obtained budget ceiling is embedded as a hard constraint into the multi-objective function framework. The embedding process transforms the budget ceiling into a boundary condition of the cost dimension. A new cost constraint vector is added to the original framework. The cost constraint vector includes parameters such as material procurement costs, construction equipment usage costs, and labor input costs. Each cost item corresponds to the material cost, equipment cost, and labor cost fields in the budget table, ensuring the consistency of the total cost calculation with the budget source data in terms of subject. According to the parameter configuration of each candidate scheme, the system expresses the total cost as a linear sum of the above cost items and sets a constraint that the total cost does not exceed the budget ceiling. This constraint is attached to the joint expression of the performance objective vector, environmental objective vector, and cost objective vector, thus forming a set of multi-objective functions with budget constraints. During the solution process, the system first filters out schemes whose total cost exceeds the budget ceiling based on the budget constraint. Only the schemes that meet the budget conditions are calculated for performance indicators and green maintenance indicators, ensuring that the candidate schemes meet the funding constraints at the generation stage and avoiding infeasible solutions that need to be eliminated later. This allows the budget ceiling to play a role as a priority constraint condition in the entire framework.

[0098] While embedding hard budget constraints, the specific weights of each objective are further defined to guide balanced optimization under constraints. For the performance maximization objective, in one preferred configuration, its weight can be set to 0.4. This objective covers indicators reflecting structural safety and service level, such as the International Roughness Index (IRI) and strength ratio. A higher weight reflects the priority given to road safety and service life. For the environmental impact minimization objective, the weight can be set to 0.3. This objective includes carbon emissions and ecological restoration scores, emphasizing minimizing negative environmental impacts while meeting safety and economic constraints. The cost minimization objective can also be set to 0.3, covering cost items such as material costs and labor costs, allowing for some optimization space for cost control outside of hard budget constraints. The above weight values ​​can be adjusted according to different project types and management strategies, and are not limited to this set of values. When the project is more focused on green development, the weight of environmental objectives can be appropriately increased, and when funds are tight, the weight of cost objectives can be appropriately increased to adapt to the actual needs under different decision-making scenarios.

[0099] The weights are determined through an expert scoring method. Specifically, multiple experts are selected from fields such as transportation engineering, environmental assessment, and financial management. Each expert scores the importance of performance, environmental, and cost objectives within a range of 0 to 1. The system summarizes and normalizes the scoring results according to the objective categories to obtain an initial weight distribution. Subsequently, experts are organized to discuss and coordinate scores with significant differences. If necessary, adjustments are made again to form a set of objective weights that have been collectively confirmed. This set is stored in a cloud database in the form of a parameter table for use by the multi-objective function framework in different decision-making cycles. This ensures that the weight configuration reflects both technical security and environmental policy requirements, as well as the actual situation of financial constraints.

[0100] The adjusted weights are directly integrated into the multi-objective function calculation. The system multiplies the performance objective vector, green objective vector, and cost objective vector by their respective weight coefficients, and sums the results to form a weighted comprehensive objective vector. This comprehensive objective is then optimized within the feasible region defined by the budget hard constraint. To evaluate the impact of weight settings on the optimization results, the system can conduct sensitivity tests. For example, it can re-solve the multi-objective optimization problem when the budget ceiling is reduced by a preset percentage, and compare the changes in performance and green indicators under different configurations. If it is found that the score of the comprehensive green maintenance indicator decreases by more than the preset allowable percentage under the configuration with higher performance weights, experts will be reorganized for review, and the weights of each objective will be appropriately adjusted to restore a reasonable balance between performance and environmental objectives, thereby reflecting the dynamic grasp of the relationship between safety, environmental protection, and cost in actual decision-making.

[0101] By introducing the aforementioned hard budget constraints and setting target weights, the cost boundaries and optimization directions of feasible solutions are clearly defined within the multi-objective function framework. On the one hand, the budget ceiling defines the feasible region where the total cost does not exceed the ceiling, ensuring that candidate solutions are financially feasible. On the other hand, performance, environmental, and cost objectives are coordinated in the weighted comprehensive objective. This can be achieved by using a weighted summation method or a multi-objective optimization algorithm based on the Pareto front to compare maintenance solutions that meet the budget constraints, obtaining a set of optimal or suboptimal solutions that compromise between performance, environmental friendliness, and cost. The output of this step includes a multi-objective function configuration file with embedded budget constraints and target weights. The budget ceiling, constraint expressions, and weight parameters are stored back to the cloud database in a structured format, providing a traceable input basis for subsequent steps in solution space partitioning and optimization.

[0102] S4, using the set constraints and weights, performs initial solution space partitioning for the multi-objective function, specifically as follows:

[0103] The system loads the budget hard constraints and objective weight vectors from the constraint weighted function framework file generated earlier as initial input to initiate the solution space partitioning of the multi-objective function. The system parses structured configuration files, such as JSON files, through an interface to extract the hard boundary expression corresponding to the budget ceiling value, such as 5 million yuan, as well as weight vector components such as performance weight 0.4, environmental weight 0.3, and cost weight 0.3. These elements are then injected into the overall structure of the multi-objective function, ensuring that the decision domain is simultaneously constrained by both budget constraints and weight allocation, providing clear boundary conditions and optimization priority information for subsequent solution space partitioning.

[0104] After loading constraints and weights, the parameter ranges of possible solutions are constructed using a gridded interval, discretizing continuous design variables into a finite number of interval points to form a high-dimensional discrete solution space. During construction, the core parameter set is first identified, including the type of maintenance material, construction period, resource allocation ratio, and intervention intensity level. These parameters correspond one-to-one with the specific components in the aforementioned performance, environmental, and cost objectives. Maintenance material types are divided into two main categories: recycled asphalt and conventional asphalt. Recycled asphalt is further subdivided into basic recycled grade and high-performance recycled grade, while conventional asphalt is subdivided into standard grade and modified grade, resulting in four grades in the material dimension. Construction periods are divided into short-term and long-term periods, with short-term periods for example, 1 to 3 months, and long-term periods for example, 4 to 12 months, with each period based on a weekly unit. The system can be further subdivided into approximately eight construction cycle levels; resource allocation ratios are divided into three sub-dimensions: labor ratio, equipment ratio, and material ratio. Each sub-dimension is discretized from 0% to 100% in 10% increments, resulting in 11 ratio levels; intervention intensity levels are divided into three levels: light repair (e.g., surface sealing), moderate repair (e.g., partial milling), and heavy repair (e.g., full-layer renovation); by combining the four levels of the material dimension, the approximately eight levels of the construction cycle dimension, the eleven levels of the resource ratio dimension, and the three levels of the intervention intensity dimension, approximately 1056 initial grid points can be obtained. The actual number of grid points can be adjusted based on computational resources and project complexity to achieve a balance between spatial coverage and computational overhead.

[0105] After gridding is completed, a layer-by-layer screening stage begins, evaluating the feasibility and merits of each solution corresponding to each grid point. The first layer of screening uses the hard budget constraint as the primary filtering condition. For each grid point, the total cost of its corresponding solution is calculated. The total cost is a linear sum of material costs, labor costs, equipment usage costs, and necessary management and miscellaneous costs. The unit prices of materials, labor, and equipment shifts used in the total cost calculation are read from the preset unit price table in the cost management module of the cloud database and are consistent with the budget table data mentioned above in terms of subject settings. The calculated total cost is compared with the budget limit. When the total cost exceeds the budget limit, the grid point is marked as infeasible and removed. For example, if a combination solution using high-performance recycled asphalt with a long construction period and high equipment investment has an estimated total cost of 5.5 million yuan, which exceeds the budget limit of 5 million yuan, then the grid point is directly removed from the solution space. Through point-by-point traversal and filtering, a first-layer feasible solution set that satisfies the hard budget constraint is formed.

[0106] Based on the first-level feasible solution set, a second-level screening is performed, focusing on the calculation of a weighted comprehensive score. The system calculates a composite score for each remaining grid point based on the performance target value, environmental target value, and cost target value defined in S2 and S3. To avoid imbalances in the weighting results due to differences in the dimensions and value ranges of different indicators, the performance target value, environmental target value, and cost target value are first normalized according to preset maximum, minimum, or quantile values ​​before participating in the weighted calculation, mapping each indicator to a unified range of 0 to 1. After normalization, the performance target value, such as the predicted IRI improvement rate, is multiplied by a performance weight of 0.4, and the environmental target value, such as the carbon emission reduction rate, is multiplied by an environmental weight of 0.3. Multiply the cost target value, such as the reciprocal of the cost per unit area, by the cost weight of 0.3, and add the three results to obtain the comprehensive score for each grid point. Sort all grid points in descending order of comprehensive score, and prioritize retaining solutions with higher scores. The retention ratio can be set to the top percentage, such as the top 60%. This ratio is a preset parameter of the system and can be adjusted according to the project scale and diversity requirements. At the same time, a preset threshold is set. When the comprehensive score of some grid points is lower than the threshold, they are removed. The threshold can be selected as the current overall average score, or it can be determined as an appropriate quantile based on the score distribution characteristics, thereby further narrowing the candidate range while ensuring the diversity of the solution space.

[0107] The third layer of screening performs cross-dimensional consistency checks and refined convergence on the high-potential schemes retained from the second layer. At this layer, the system verifies the matching relationships between various parameters. For example, it checks the reasonable combination of material type and construction period. If the combination of recycled asphalt and an extremely short construction period results in a low carbon emission reduction rate and insufficient ecological restoration effect, the priority of such combinations can be reduced or they can be eliminated. Regarding the relationship between performance and cost, it checks whether improving performance indicators requires excessively high unit area costs. When performance weights are high, cost growth is still required to be controlled within a preset allowable range. Regarding the relationship between environment and budget, it requires that the comprehensive environmental score under hard budget constraints not be lower than a preset benchmark value, such as 0.5, to avoid schemes that meet cost constraints but have significantly insufficient green performance. Through the above multi-dimensional verification and convergence, the initial approximately 1056 grid points can be gradually compressed into a smaller number of high-potential areas, for example, shrinking to approximately 200 representative grid points. These grid points are concentrated in areas with high comprehensive scores and all constraints are met.

[0108] To prevent the overly aggressive partitioning process from mistakenly eliminating potentially excellent solutions, a feedback adjustment mechanism is implemented after each layer of screening. When the rejection rate after a certain layer of screening exceeds a preset rate, such as 30%, the system can trigger a backtracking operation to fine-tune the grid boundaries of some parameter dimensions. For example, the short-term construction cycle range can be expanded from 1 to 3 months to 1 to 4 months, or the resource ratio step size can be appropriately relaxed to cover boundary areas that may have been missed previously. After feedback adjustment, the screening is re-executed to ensure that the solution space exploration remains balanced, avoiding both excessive shrinkage and leaving a large number of obviously infeasible solutions in subsequent calculations. By combining grid partitioning, pre-budget filtering, weighted score sorting, and cross-dimensional verification, this step forms an initial feasible solution set with a clear structure and appropriate size under multi-objective constraints.

[0109] The entire partitioning and screening process ultimately outputs a refined set of feasible solutions to support subsequent fine-grained optimization. The file is stored in a structured text format, such as comma-separated value format, recording the parameter combinations, corresponding performance scores, environmental scores, cost indicators, and comprehensive scores for each retained grid point, along with feasible domain boundary information. After the file is generated, it is uploaded to the cloud database via a database interface, linked with the aforementioned constraint weighting function framework file, and directly imported as the initial solution space for iterative optimization. At the same time, the file records metadata such as screening levels, elimination ratios at each level, and key parameter adjustment records, providing a basis for subsequent auditing, solution reproduction, and parameter readjustment. Thus, under the premise of close connection with budget constraints and target weight settings, it lays a traceable starting point for the global optimization of subsequent maintenance schemes.

[0110] S5, within the partitioned solution space, performs iterative adjustments, expanding the boundaries and cross-validating each feasible solution. Specifically, this is implemented as follows:

[0111] The refined feasible solution set file output from the previous section is directly imported into the current processing module. This file, for example, contains the parameter coordinates, weighted preliminary scores, and boundary descriptions of approximately 200 high-potential grid points. The system parses the structured text file, such as comma-separated value format, and loads each grid point into a dynamic solution space structure. This structure is organized in the form of a multi-dimensional array, with each array element corresponding to a complete parameter set of a scheme, including material type grid, construction cycle level, resource allocation ratio, and intervention intensity level. During the loading process, the integrity of the file structure and the number of records is checked. When the number of valid grid points is lower than a preset threshold, such as less than 150, a backtracking notification is triggered to readjust the grid boundaries or screening parameters of the previous stage, thereby providing a stable and reliable starting basis for iterative adjustments.

[0112] After obtaining the refined solution space, a representative set of schemes is first selected as the starting point for iteration. The selection logic is as follows: the grid points are stratified and grouped according to the weighted comprehensive preliminary scores mentioned above. For example, the proportions of high-level, medium-level, and low-level strata can be preset to 60%, 30%, and 10%, respectively. This proportion can be adjusted according to project needs. Within each stratum, schemes that can cover a larger range of parameter variations are prioritized. For example, in the high-level stratum, a scheme using recycled asphalt and configured with a short construction cycle is selected as a performance-oriented representative, while in the same stratum, a scheme using traditional asphalt and configured with a longer construction cycle is selected as a cost-oriented representative. This ensures that the representative schemes have sufficient differences in dimensions such as material type, construction cycle, resource allocation, and intervention intensity. The number of representative schemes can be controlled at a certain proportion of the total number of points, such as about 20%. When the total number of grid points is 200, about 40 representative schemes can be selected. The actual proportion can be appropriately adjusted according to the computational resource conditions and diversity requirements.

[0113] After determining representative schemes, a boundary expansion operation is performed on each representative scheme. The parameters of each scheme are considered as the center point, and the expansion is gradually extended to neighboring grids around the center within an allowable range. During expansion, the expansion radius is set as a multiple of the original grid step size; for example, it can be preset to 1.5 times the original step size. The specific value is configured based on parameter sensitivity and computational capability. For material proportion parameters, such as 50% recycled asphalt in the initial scheme, it can be expanded to a neighborhood range of 45% to 55% without exceeding the previously established upper and lower limits. For construction cycle parameters, such as a short-term construction cycle of 1 to 3 months, it can be expanded to 1 to 4 months. The duration should be no less than one month and no more than the long-term upper limit set above. The expansion direction can be adjusted according to the optimization objective preference: in performance-oriented expansion, the intervention intensity should be fine-tuned to a higher level to improve the smoothness improvement rate; in environment-oriented expansion, the material consumption should be reduced or the proportion of recycled materials should be increased to increase the carbon emission reduction; in cost-oriented expansion, resource allocation should be compressed to get as close as possible to the budget lower limit. Each representative scheme can generate several expansion variants, such as 5 to 10. All variants must meet the budget hard constraints and relevant boundary conditions. The expansion results are linked to the original scheme in a tree structure to facilitate the tracking of the source and evolution process of each variant.

[0114] After boundary expansion, cross-validation is performed on the expanded variants. Pairing relationships are constructed, and each expanded variant is compared with other schemes in the representative set to form a pairing matrix. For each pair of schemes, parameter compatibility indicators are calculated, including whether the combination of material type and construction cycle is reasonable, whether the resource allocation meets the construction feasibility, and whether the intervention intensity is adapted to the pavement performance degradation characteristics. The above compatibility indicators can be converted into percentage scores according to preset rules for comparison with preset compatibility thresholds. Compatibility assessment can be carried out in batches. The first batch focuses on checking the balance between performance objectives and environmental objectives, combining a performance weight of 0.4 and an environmental weight of 0.3 to determine whether the expanded scheme maintains a certain level of greenness while improving performance. The second batch focuses on checking the relationship between cost and budget to ensure that the total cost of the expanded variants does not exceed the budget limit. The system can set a compatibility threshold. For example, when the compatibility pairing ratio of a certain expanded variant with other schemes is lower than a preset ratio, such as 70%, the variant is marked as low compatibility and removed or downweighted, thereby retaining a subset of expanded schemes with reasonable structure and compatible parameters.

[0115] Based on cross-validation, the parameters of the retained extended variants are fine-tuned. The parameter adjustment logic performs local optimization on the imbalances between performance, environment, and cost exposed in the paired comparison. For example, when a variant has a high performance score but a low carbon emission reduction rate, the proportion of recycled materials in the material mix can be increased by a preset step size, such as 5%, without reducing the intensity of intervention, and the construction period can be appropriately extended to take into account construction quality. When the cost of a variant is close to the budget limit and the performance improvement is limited, the proportion of equipment input can be reduced, the proportion of labor input can be appropriately increased, and the cost and performance indicators can be re-estimated. Parameter adjustment can be performed in stages. The first stage adjusts only a single parameter, and the second stage adjusts two related parameters in a coordinated manner, such as the construction period and the resource allocation ratio. After each adjustment, the composite score is recalculated using the target system defined above, and the number of iterations for a single scheme is limited. For example, each scheme is preset to a maximum of 5 adjustments to prevent excessive modification and excessive computational overhead. At the same time, the parameter changes and score changes in each adjustment are recorded to form a complete adjustment trajectory.

[0116] After local adjustments are completed, the adjusted variants are re-integrated into the dynamic solution space for global comparison and merging. The system sorts all representative solutions and their extended variants according to the updated composite scores, from high to low, and compares the parameter differences of adjacent solutions one by one. When the relative difference between two solutions in the main parameters is lower than a preset threshold, such as less than 2%, they are considered highly similar combinations and can be merged into a single representative solution, retaining the one with the higher score, and recording the merging information at the same time. Global comparison and merging can be performed in multiple rounds, for example, the preset is 3 rounds. The first round optimizes local combinations within the same parameter cluster, the second round merges solutions with both high performance scores and high environmental scores between different parameter clusters, and the third round focuses on examining variants in the boundary region to avoid missing potential excellent solutions in the edge region. Through multiple rounds of comparison and merging, the solution space can gradually converge from the extended set starting with the representative solutions, for example, from several hundred points formed by the initial approximately 40 representative solutions and their extended variants to a set of optimized solutions of approximately 150. The above number is an example value and can be adjusted according to the actual project.

[0117] Through an iterative adjustment process combining boundary expansion, cross-validation, parameter fine-tuning, and global comparison and merging, coarse-to-fine adaptive optimization was achieved within the initial feasible solution space generated earlier. This allowed for the gradual selection of more balanced solution combinations across performance, environment, and cost, while maintaining the hard budget constraints and target weights. The entire iterative adjustment process ultimately outputs an optimized solution set file, stored in a structured text format, such as comma-separated value format. This file records the parameter combinations, updated performance scores, environmental scores, cost indicators, and overall scores for each retained solution, along with an adjustment trajectory summary and solution space coverage indicators, such as the proportion of different material types and construction cycle combinations in the final solution set, used to evaluate solution diversity. This file is uploaded to a cloud database via a database interface and linked with the aforementioned constraint weighting function framework file and refined feasible solution set file.

[0118] S6, based on the iteratively obtained solution set, outputs the final maintenance decision scheme and applies the selected scheme to actual road management. The specific implementation is as follows:

[0119] Import the optimized solution set file generated earlier into the decision module. This file contains several parameter combinations, corresponding optimization scores, and adjustment trajectory logs. The optimization scores are weighted comprehensive scores calculated based on the target weights set in S3. Parse the structured text file and load each parameter combination into a list sorted in descending order of comprehensive score. Each combination element includes material type, construction period, resource allocation ratio, intervention intensity level, and derived indicators such as predicted performance improvement value, environmental score, and cost indicators. During the loading process, the integrity of the trajectory logs and parameter fields is checked. When it is detected that the number of adjustment rounds for a certain combination is less than the preset number of rounds, the combination is marked as an object that needs to be reviewed. The preset number of rounds can be adjusted according to project requirements.

[0120] After obtaining the optimized solution set, a high-level filtering is performed on the sorted list to form candidate subsets. The filtering logic sets a dynamic threshold based on the statistical characteristics of the scores. First, the average and standard deviation of the comprehensive scores of all combinations are calculated. Then, the screening threshold is set to a certain proportion of the average plus the standard deviation. Only combinations with comprehensive scores higher than this threshold are retained as candidate subsets, and the size of the candidate subset is controlled to a certain proportion of the total number of loads, such as about 30%. To avoid the candidate subsets being too concentrated in terms of material type or construction strategy, the parameter distribution in the candidate subsets is checked. When it is found that the proportion of a certain material type in the candidate subset exceeds the preset proportion, alternative solutions are added from combinations with slightly lower comprehensive scores but different material types, so that the candidate subsets maintain diversity in the main parameter dimensions.

[0121] A detailed comparison is conducted based on the candidate subsets to select the final maintenance scheme. The comparison logic adopts a multi-dimensional and hierarchical sorting approach. First, the candidate subsets are sorted according to the performance dimension, and several combinations with high performance scores are selected to form the first candidate pool, for example, the top 10 combinations. Then, within the first candidate pool, they are re-sorted according to the environmental dimension, prioritizing the retention of combinations with high environmental scores and small differences, for example, retaining the top 5 schemes with relatively balanced performance and environmental impact. Finally, within this range, they are sorted according to the cost dimension, comparing cost indicators with budget utilization, and selecting one or a few schemes with the highest comprehensive score that meet cost constraints as the final decision scheme. For example, a combination scheme that includes recycled asphalt sealing and preventive crack repair can be selected. Specific parameters can be automatically matched by the system according to the project scale, budget, and construction conditions, and manually confirmed by management personnel.

[0122] After selecting the final scheme, detailed implementation information is extracted from the corresponding parameters and an implementation plan is generated. Based on the pavement performance degradation model and green maintenance index system constructed in S2, combined with the monitoring data of the current pavement section and environmental data such as temperature and rainfall, a suitable construction window is calculated, and the construction start date is set as the start date of the next suitable maintenance period. The construction duration matches the construction cycle in the scheme, and the entire process is subdivided into preparation, construction, and acceptance stages. The required resources are quantified according to the resource allocation ratio. The system automatically calculates the quantity of labor, equipment, and materials based on the construction area, layer thickness, and process ratio, and verifies that the total resource cost meets the budget hard constraint set in S3. The expected results are obtained by simulating the performance degradation model in S2 by substituting the selected scheme parameters, and obtaining the number of years the pavement's remaining life is extended, the degree of smoothness improvement, and the improvement of ecological restoration indicators.

[0123] The aforementioned implementation information is organized into a structured detail data package, recording scheme parameters, schedules, resource requirements, and expected results in key-value pairs, and output to the user interface for management review. The interface centrally displays a scheme overview, phase schedule, resource list, and comparative charts of performance and environmental indicators before and after construction. The interface allows management to simulate and fine-tune some parameters within permissible limits and instantly refresh expected results to assist in decision-making and evaluation. Unless management explicitly submits an update in the confirmation interface, simulation and fine-tuning will not automatically modify the selected scheme and imported execution plan. User confirmation operations and feedback in the interface are logged as a reference for subsequent weight settings and iterative strategy adjustments.

[0124] After management confirms the plan, the selected plan is imported into the road management system and driven to actual execution. The detailed data package is converted into Extensible Markup Language (XML) format, including nodes such as road segment identification, plan parameters, time schedule, resource plan, and expected results. It is then uploaded to the central server of the road management system via an interface, updating the status fields of relevant road segments, such as changing the status from monitoring to maintenance plan pending or maintenance in progress. The current plan is also linked and archived with historical plan records. After the plan is imported, it is linked with the field execution subsystem through the application programming interface (API), decomposing the decision plan into executable operation instructions. The scheduling system then uses the construction time and location parameters in the plan to execute these instructions. The system sends task notifications to the construction team's terminals. The inventory management subsystem checks the inventory of materials and equipment based on resource requirements. When the available inventory of a certain key material is lower than the required quantity in the plan, a purchase or allocation instruction is generated. The monitoring subsystem collects data through a sensor network during and after the maintenance process. It compares the results of the international flatness index and structural response obtained after construction with the expected values ​​to generate a performance deviation report. The construction equipment regularly reports its work status and progress information through the Internet of Things module. The system compares this information with the plan's timetable and target progress. When the actual progress is significantly behind the plan, it triggers a reminder or warning and provides suggestions for adjusting resource allocation or construction pace.

[0125] To enhance the traceability of the implementation and decision-making process, an implementation confirmation document is generated, summarizing the details of the plan, import logs, on-site linkage instructions, and key monitoring results during the implementation process. This document is then stored back in a structured format in the cloud database and linked with the aforementioned monitoring data, model parameters, and optimization results. When subsequent long-term monitoring data indicates that the changes in pavement performance deviate from expectations, the system can analyze the causes based on the implementation confirmation document and historical records. If necessary, the model correction and plan re-optimization process can be re-executed, enabling the maintenance strategy to be continuously iterated and updated within the same data and model framework.

[0126] The solution in this embodiment first deploys an integrated sensor network, including embedded strain, temperature, and humidity sensors, as well as a surface laser scanner and a vehicle load counter. This network wirelessly collects road surface deformation, traffic load, and environmental data in real time, and stores the data in a structured format in a cloud database after classification and cleaning. Next, key datasets are extracted from the database to establish a road surface performance degradation model and construct a green maintenance index system. Performance indicators, environmental factors, and cost parameters are integrated to form a vector-based multi-objective function framework. Based on this, the budget ceiling is read and embedded as a hard constraint into the multi-objective function. Simultaneously, the weights for maximizing performance, minimizing environmental impact, and minimizing cost are determined using an expert scoring method. Finally, the multi-objective function is initially partitioned into a gridded solution space, and intervals that do not meet the constraints are selected and eliminated hierarchically. High-potential areas are gradually identified based on weighted scores. Within the convergence space, representative solutions are selected, and several variants are generated by expanding their boundaries. Through cross-validation and parameter adjustment, these variants are repeatedly compared and merged to optimize parameter combinations and cover more potential superior paths. Finally, the highest-scoring maintenance solution is selected from the optimized solution set, such as a combination of recycled material sealing and preventive crack repair. Based on this, the implementation schedule, resource requirements, and expected results are derived and output to the interactive interface for management personnel to review and confirm. After confirmation, the solution is imported into the pavement management system. Through the application programming interface, the system links on-site equipment to schedule construction teams, monitor construction progress, and transmit execution data, forming a closed-loop control mechanism integrating monitoring, decision-making, and implementation. This ensures that maintenance decisions are efficient and feasible, results are traceable, and continuous iterative optimization can be achieved in long-term operation.

[0127] Example 2: Figure 2 This invention presents a decision-making system for green maintenance of asphalt pavement based on long-term performance monitoring, comprising:

[0128] Data acquisition and storage module: responsible for deploying sensor networks to collect road surface deformation, traffic load and environmental data, classifying and organizing the data, cleaning outliers, and storing the data in a structured format in the database;

[0129] Model and framework building module: Based on the stored data, extract key datasets, establish a performance degradation model and a green maintenance index system, and integrate performance, environmental and cost parameters to form a multi-objective function framework;

[0130] Constraint and Weight Setting Module: Introduces hard budget constraints on top of the functional framework and defines the weights of performance, environmental, and cost objectives;

[0131] Solution space partitioning module: The function is divided into gridded intervals using constraints and weights. Invalid intervals are eliminated by filtering layer by layer, and the set of feasible solutions is converged.

[0132] Iterative adjustment module: Selects representative solutions within the feasible solution space, performs boundary expansion and cross-validation, and optimizes parameter combinations to cover more paths;

[0133] Decision Output and Application Module: Based on the iterative solution set, the highest-scoring solution is selected, its details are extracted and output to the interface, and imported into the management system to guide the execution of the solution through API linkage with on-site equipment.

[0134] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0135] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0136] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0137] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0138] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0139] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0140] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0141] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0142] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring, characterized in that, include: S1, collect long-term performance monitoring data of asphalt pavement, classify and organize the data and store it in the database; S2. Based on the stored data, a pavement performance degradation model and a green maintenance index system are established. By integrating performance indicators, environmental factors and cost parameters, a multi-objective function framework is formed. S3, based on the established framework, sets budget constraints, incorporates the budget ceiling as a hard constraint into the multi-objective function, and defines the specific objective weights for maximizing performance, minimizing environmental impact, and minimizing cost. S4. Using the set constraints and weights, the initial solution space of the multi-objective function is partitioned. S5, within the partitioned solution space, performs iterative adjustments, and performs boundary expansion and cross-validation for each feasible solution; S6, based on the solution set after iteration, outputs the final maintenance decision scheme and applies the selected scheme to actual road management.

2. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, Collect long-term performance monitoring data for asphalt pavement, classify and organize the data, and store it in a database, including: A sensor network was deployed on a selected asphalt pavement monitoring section to collect various types of data. Data is transmitted to the edge gateway via a wireless module for synchronization and then uploaded to the cloud. The data is subject-based, derived metrics are calculated, and data cleaning is performed, including outlier removal, noise interpolation, and missing data labeling. The cleaned data is written to a distributed database in a structured format. The main table is linked to multiple sub-tables through foreign keys, and indexes are created.

3. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, Based on the stored data, a pavement performance degradation model and a green maintenance index system are established. By integrating performance indicators, environmental factors, and cost parameters, a multi-objective function framework is formed, including: Extract structural performance, service intensity, and external impact data corresponding to pavement segment numbers and time series from the database; Establish a pavement performance degradation model to simulate the strength variation with time and load; Establish a green maintenance indicator system to quantify carbon emissions and material recycling rates; Integrating performance indicators, environmental factors, and cost parameters, a vector-based multi-objective function framework is formed.

4. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, Based on the established framework, budget constraints are set, with the budget ceiling as a hard constraint integrated into the multi-objective function. Specific objective weights for maximizing performance, minimizing environmental impact, and minimizing cost are defined, and the implementation is as follows: Extract the budget ceiling value from the database and embed it into a multi-objective function framework; Add a new cost constraint vector and set a linear summation of the total cost of candidate solutions; Schemes exceeding the budget limit are eliminated; performance and green maintenance indicators are calculated only for schemes that meet the conditions. Weights are determined and adjusted based on expert scoring to reflect priority. Scores are summarized and normalized according to the target categories, and weight sets are formed after reconciling the differences; The weights are multiplied by the corresponding target vectors and then summed to form a weighted composite target vector.

5. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, Using the defined constraints and weights, the initial solution space of the multi-objective function is partitioned, including: Load the budget hard constraints and target weight vector from the configuration file; The core parameter set is discretized into gridded intervals to construct a high-dimensional solution space; Filter grid points layer by layer and eliminate schemes that exceed the limits; Calculate the normalized weighted composite score and retain high-potential points in order; Perform cross-dimensional consistency verification to further converge the solution set; Set up a feedback adjustment mechanism to fine-tune the grid boundaries and re-filter; The refined solution set is stored in a structured format in the database.

6. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, Within the partitioned solution space, iterative adjustments are performed, and boundary expansion and cross-validation are conducted for each feasible solution, including: Load a dynamic solution space structure from the refined feasible solution set, and select representative schemes hierarchically based on the comprehensive score; Perform a boundary expansion operation around the center point to generate an expanded variant; Construct a pairing matrix to perform cross-validation, calculate compatibility metrics, and eliminate low-compliance variants; Fine-tune the parameters of the retained variants, adjust the relevant parameters in conjunction, and calculate the composite score; Perform a global comparison and merge on the adjusted variants, sort by score and merge similar combinations; The optimized solution set is stored in a structured format in the database.

7. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, Based on the iteratively obtained solution set, the final maintenance decision scheme is output, including: Load the parameter combinations and score list from the optimized solution set file; Perform high-level filtering on the list, calculate a score threshold, and retain combinations that are above the threshold; Examine the diversity of parameter distributions and supplement alternatives from combinations with lower scores; The final maintenance plan is selected by prioritizing the options based on performance, environment, and cost. Implementation information is extracted from parameters to estimate construction windows and resource quantification, and a timetable is generated.

8. The decision-making method for green maintenance of asphalt pavement based on long-term performance monitoring according to claim 1, characterized in that, The selected solutions will be applied to actual road management, including: The detailed data package is output to the user interface, displaying a layered overview of the solution, schedule, resource panel and effect comparison. It supports users to simulate fine-tuning parameters and refresh in real time, and records confirmation operations and feedback logs. The confirmed plan is converted into a structured format and uploaded to the central server of the road management system to update the road segment status fields; The decomposition plan is converted into operation instructions, which send task notifications to the construction terminal, generate allocation instructions to the inventory subsystem, and collect maintenance data from the monitoring subsystem and compare it with expectations. Generate an execution confirmation file, summarizing the scheme details, import logs, and execution results, and store them back in the database in a structured format and establish a relationship.

9. A green maintenance decision-making system for asphalt pavement based on long-term performance monitoring, used to implement the green maintenance decision-making method for asphalt pavement based on long-term performance monitoring as described in any one of claims 1-8, characterized in that, include: Data acquisition and storage module: responsible for deploying sensor networks to collect road surface deformation, traffic load and environmental data, classifying and organizing the data, cleaning outliers, and storing the data in a structured format in the database; Model and framework building module: Based on the stored data, extract key datasets, establish a performance degradation model and a green maintenance index system, and integrate performance, environmental and cost parameters to form a multi-objective function framework; Constraint and Weight Setting Module: Introduces hard budget constraints on top of the functional framework and defines the weights of performance, environmental, and cost objectives; Solution space partitioning module: The function is divided into gridded intervals using constraints and weights. Invalid intervals are eliminated by filtering layer by layer, and the set of feasible solutions is converged. Iterative adjustment module: Selects representative solutions within the feasible solution space, performs boundary expansion and cross-validation, and optimizes parameter combinations to cover more paths; Decision Output and Application Module: Based on the iterative solution set, the highest-scoring solution is selected, its details are extracted and output to the interface, and imported into the management system to guide the execution of the solution through API linkage with on-site equipment.

Images