Rigid pavement texture engraving parameter intelligent optimization design method and system

Abstract

The invention discloses a rigid pavement texture carving parameter intelligent optimization design method and system, and relates to the field of pavement texture design, and the method comprises the steps: collecting three types of core data to construct a multi-dimensional pavement working condition database, carrying out the data processing and working condition risk quantification, carrying out the hierarchical training through a CatBoost algorithm, solidifying a texture performance mapping model, extracting a performance weight, and carrying out the calculation of the performance weight. And then defining parameter engineering constraints, generating a Pareto optimal frontier through an NSGA-III algorithm, screening an optimal scheme in combination with an engineering preference weight, and finally outputting a standardized design scheme and a quantitative association report. According to the method, a multi-dimensional association database is constructed by collecting three types of core data, quantitative modeling is carried out, refined design of rigid pavement texture parameters is achieved, multi-performance collaborative optimization is achieved by focusing on anti-sliding, noise reduction and drainage core requirements, the limitation of single target optimization is broken, design efficiency and global optimization capacity are improved by relying on an efficient algorithm, and the method is suitable for large-scale popularization and application. Industry specifications are met, and project landing feasibility is ensured.

Description

Technical Field

[0001] This invention relates to the field of road surface texture design, specifically to a method and system for intelligent optimization design of rigid road surface texture carving parameters. Background Technology

[0002] The macroscopic texture (such as roughening and grooving) of rigid pavement (such as cement concrete pavement) is a key factor that directly determines its anti-skid performance, drainage capacity, driving noise and ride comfort, and is of great importance to driving safety and environmental protection. However, the current design of its engraving parameters mainly relies on the experience and specifications of engineers. This method lacks quantitative basis, is difficult to adapt to complex working conditions, and the design process is often conservative, making it difficult to fully explore the performance potential of the texture.

[0003] Existing optimization methods often only target a single performance objective (such as skid resistance or noise reduction), ignoring the inherent contradictions and synergistic requirements among multiple performance indicators. This makes it difficult to meet the comprehensive requirements of modern roads for safety, environmental protection, and comfort. At the same time, due to the large number of parameter combinations and the complexity of performance calculations, traditional methods and simple optimization algorithms suffer from low design efficiency and insufficient global optimization capabilities, making it difficult to quickly obtain the optimal solution for multi-objective balance. This seriously restricts the large-scale and refined design application of high-quality pavements. Summary of the Invention

[0004] To address the aforementioned technical problems, this paper provides an intelligent optimization design method and system for rigid pavement texture carving parameters. This technical solution solves the problem that some existing optimization methods mentioned in the background technology often only target a single performance objective (such as anti-skid or noise reduction), ignoring the inherent contradictions and synergistic requirements between multiple performance indicators. This makes it difficult to meet the comprehensive requirements of modern roads for safety, environmental protection, and comfort. At the same time, due to the large number of parameter combinations and the complexity of performance calculations, traditional methods and simple optimization algorithms have the defects of low design efficiency and insufficient global optimization ability, making it difficult to quickly obtain the optimal solution for multi-objective balance, which seriously restricts the large-scale refined design and application of high-quality pavements.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method and system for intelligent optimization design of rigid pavement texture engraving parameters, including: Collect historical engineering data, indoor and outdoor test data, and construction site inspection data to construct a multi-dimensional pavement condition database containing texture parameters and performance indicators; Obtain non-numerical pavement condition data and numerical pavement condition data from the multidimensional pavement condition database, merge low-frequency non-numerical pavement condition data, standardize numerical pavement condition data, and divide the multidimensional pavement condition database into training data and validation data based on stratified sampling of pavement conditions. Based on the multidimensional pavement condition database and historical engineering data, the priority of pavement conditions is classified, and a risk coefficient benchmark range is set to correspond to the priority classification of pavement conditions, thereby quantifying the risk level of pavement conditions. A comprehensive coefficient benchmark is obtained based on the risk level and risk coefficient range of the road surface conditions, and the balance and consistency with engineering logic are verified. Based on the features of a multi-dimensional road condition database, the control parameters of the CatBoost algorithm are set, the training data and validation data are fed into the algorithm and the categorical features are automatically identified, and the sample weights are bound to the multi-dimensional data. The training data is used to conduct hierarchical training of the model, and the validation data is simultaneously connected to verify the model in real time and limit the training step size. After completion, it is labeled as a texture performance mapping model. The model is subjected to hierarchical cross-validation under different operating conditions. An engineering optimization threshold is set. If the threshold is not met, the algorithm parameters are iteratively adjusted. Once the threshold is met, the model is solidified, and the texture parameters are extracted to map the performance weights corresponding to the texture parameters and performance indicators in the model. Define the engineering constraints of texture parameters, select a multi-objective evolutionary algorithm, conduct a global parallel search, generate the Pareto optimal front, set engineering preference weights, screen the texture parameter scheme with the best overall performance, generate a standardized design scheme, and simultaneously output a quantitative correlation report of texture parameters and performance indicators.

[0006] Preferably, the step of collecting historical engineering data, indoor and outdoor test data, and construction site inspection data to construct a multi-dimensional pavement condition database containing texture parameters and performance indicators specifically includes: Extract texture parameters and performance indicators from the design phase, construction records and long-term performance tracking data from historical engineering data, test data and experimental conditions of specimens with different texture parameters from indoor and outdoor experimental data, actual texture parameter data and actual pavement performance data from construction site inspection data, and collectively refer to them as multidimensional acquisition data. Filter the multidimensional collected data, where the texture parameter data should include at least texture depth, texture width, texture spacing, texture distribution form and texture distribution angle, and the performance index data should include at least skid resistance coefficient, noise reduction performance and drainage performance. Then, associate the multidimensional collected data with road surface type, climate conditions and traffic load level to identify working condition characteristics. Based on the selected texture parameters, performance indicators, and working condition characteristics, a structured integration is performed to establish data correspondence and construct a multi-dimensional road condition database.

[0007] Preferably, the step of acquiring non-numerical and numerical pavement condition data from the multi-dimensional pavement condition database, merging low-frequency non-numerical pavement condition data, standardizing the numerical pavement condition data, and dividing the multi-dimensional pavement condition database into training and validation data based on stratified sampling according to the pavement condition, specifically includes: Non-numerical pavement condition data are extracted from the multi-dimensional working condition database. The non-numerical pavement condition data is cleaned, and the sample proportion of the cleaned non-numerical pavement condition data categories is calculated and filtered. Non-numerical pavement condition data with a sample proportion lower than the preset standard are integrated as low-frequency categories, a standardized non-numerical working condition data category system is established, and the non-numerical pavement condition data is standardized and coded. Numerical pavement condition data are extracted from the multidimensional working condition database and grouped into working condition combinations. Missing samples in the numerical pavement condition data are supplemented and removed. The processed numerical pavement condition data is standardized and then the processed numerical pavement condition data and the non-numerical pavement condition data encoded in the standardized non-numerical working condition data category system are referred to as the full data. Based on the combination of road surface type, climate conditions, and traffic load level, the full data is divided into training data and validation data according to a preset ratio, and the training data and validation data are verified.

[0008] Preferably, the step of classifying the priority of pavement conditions based on a multi-dimensional pavement condition database and historical engineering data, setting a risk coefficient benchmark range to correspond to the pavement condition priority classification, and quantifying the risk level of pavement conditions specifically includes: Based on the condition feature data in the multidimensional pavement condition database and the defect correlation data in historical engineering data, the pavement conditions of rigid pavement are classified into priority levels. The condition feature data includes three categories: pavement type, climate conditions, and traffic load level. For road conditions with different priorities, corresponding risk coefficient benchmark intervals are set so that the numerical range of the risk coefficient benchmark intervals is positively correlated with the priority level of the road conditions. By breaking down the risk impact dimensions of road surface conditions and combining the correlation analysis results of defect data in historical engineering data, the risk impact weight of each dimension is obtained. The risk level of different conditions is then quantitatively calculated to obtain the dimensional risk coefficient of a single condition.

[0009] Preferably, the step of obtaining a comprehensive coefficient benchmark based on the risk level and risk coefficient range of the road surface conditions, and verifying the balance and consistency with engineering logic, specifically includes: Match the risk coefficient of a single working condition with the benchmark range of the risk coefficient of the corresponding priority, and use the product of the risk coefficients of each dimension and the superposition of the lower limit of the benchmark range of the risk coefficient of the corresponding priority as the comprehensive coefficient benchmark. The obtained comprehensive coefficient benchmark is verified bidirectionally to check the balance of the comprehensive coefficient benchmark between different priority working conditions and to check the consistency between the comprehensive coefficient benchmark and the road engineering risk logic.

[0010] Preferably, the step of setting CatBoost algorithm control parameters based on multi-dimensional road condition database features, feeding training and validation data into the algorithm and automatically identifying categorical features, and binding sample weights with multi-dimensional data specifically includes: Based on the sample size, condition complexity, and feature dimensions of the multidimensional pavement condition database, the core control parameters of the CatBoost algorithm are set to form an algorithm parameter system adapted to the learning of the mapping relationship between rigid pavement texture parameters and performance indicators. Based on the standardized non-numerical working condition data category system, the non-numerical working condition data in the sample is standardized and encoded. The encoded sample dataset containing categorical features, numerical texture parameters and performance indicators is fed into the CatBoost algorithm to identify the correlation between categorical features and performance indicators and quantify the importance of categorical features. The comprehensive coefficient benchmark of the working condition is transformed into sample weights, which are then associated and bound with the feature data and performance index labels in the multi-dimensional pavement working condition database to form weighted training data units.

[0011] Preferably, the step of using training data to perform hierarchical training of the model, synchronously accessing validation data for real-time verification and limiting the training step size, and labeling the completed model as a texture performance mapping model, specifically includes: Using the three-dimensional combination of road surface type, climate conditions, and traffic load level as the basis for working condition stratification, the training data is divided into several working condition strata, and stratified training is carried out in accordance with the hierarchical order. The common correlation rules between texture parameters and performance indicators in each stratum are integrated, and the nonlinear mapping logic of texture parameters-performance indicators under different working conditions is integrated across strata. Synchronously access validation data, set a performance check to be performed every 10 rounds of CatBoost decision tree iterations, and limit the model training step size to avoid model overfitting; Once the model training meets the preset verification criteria, it is labeled as a texture performance mapping model.

[0012] Preferably, the step of performing stratified cross-validation on the model under different operating conditions, setting an engineering optimization threshold, iteratively adjusting algorithm parameters if the threshold is not met, solidifying the model after the threshold is met, and extracting the performance weights corresponding to the texture parameters and performance indicators in the texture parameter mapping model, specifically includes: Using road surface type, climate conditions, and traffic load level as the basis for working condition stratification, the trained texture performance mapping model is subjected to working condition stratified cross-validation. The model is trained and validated in multiple rounds of data subset rotation to evaluate its adaptability and generalization ability to different working conditions. An engineering optimization threshold is set for model performance. The threshold includes the overall fitting accuracy index, the core working condition prediction accuracy index, and the performance trend consistency index. If the model performance does not meet the engineering optimization threshold, the core control parameters of the CatBoost algorithm are iteratively adjusted until the texture performance mapping model performance meets the standard. Once the texture performance mapping model meets the preset optimization threshold, it is engineered and solidified to form a reusable texture performance mapping model. Extract the performance weights of each texture parameter and performance index in the texture performance mapping model, and clarify the priority of the influence of texture parameters on each performance index.

[0013] Preferably, the process of defining the engineering constraints of texture parameters, selecting a multi-objective evolutionary algorithm, conducting a global parallel search, generating a Pareto optimal front, setting engineering preference weights, screening the texture parameter scheme with the best overall performance, generating a standardized design scheme, and simultaneously outputting a quantitative correlation report of texture parameters and performance indicators specifically includes: Based on the design specifications for cement concrete pavement of highways, engineering compliance constraints are set for the texturing parameters of rigid pavement, and the upper and lower limits of the values ​​of each texturing parameter are clarified. Based on the requirement of coordinated optimization of multiple performance indicators of rigid pavement, a suitable multi-objective evolutionary algorithm is selected, and a multi-objective optimization solution framework is built. Within the defined engineering constraints of texture parameters, the NSGA-Ⅲ algorithm was selected as the multi-objective evolutionary algorithm to carry out a global parallel search and generate the Pareto optimal frontier of the performance index trade-off relationship. Based on the core requirements of specific engineering scenarios, engineering preference weights are set, and texture parameter schemes with the best overall performance are selected from the Pareto optimal frontier based on these preference weights. The selected optimal texture parameter scheme is transformed into a standardized engineering design scheme, and a quantitative correlation report of texture parameters and performance indicators is output simultaneously.

[0014] Preferably, the intelligent optimization design system for rigid pavement texture engraving parameters for implementing the intelligent optimization design method for rigid pavement texture engraving parameters includes: The data management module is used for data collection and processing. It is responsible for the collection, screening, and structured storage of historical engineering data, indoor and outdoor experimental data, and construction site testing data, as well as the low-frequency category merging, data cleaning, standardization processing, and condition stratified sampling of non-numerical pavement condition data to divide training and validation data. The model prediction module is used to carry the solidified texture performance mapping model and realize the prediction of performance indicators of texture parameters and working conditions. Specifically, it includes receiving the input texture parameters and corresponding working conditions data, calling the texture performance mapping model to output the predicted values ​​of core performance indicators such as anti-slip coefficient, noise reduction performance and drainage efficiency, and extracting the performance weights corresponding to each performance indicator in the model. The multi-objective optimization solution module is used to carry out multi-objective collaborative optimization of texture parameters of rigid pavement. Specifically, it includes defining the engineering constraint range of texture parameters in combination with road engineering industry standards, building an optimization solution framework by loading the NSGA-Ⅲ multi-objective evolutionary algorithm, carrying out global parallel search based on the performance evaluation results of the model prediction module, and generating Pareto optimal frontier that represents the performance trade-off relationship. The interactive decision-making module is used to realize the interactive setting of engineering preference weights and the selection of optimal solutions. Specifically, it provides designers with a weight adjustment interface, supports setting preference weights for anti-skid, noise reduction, and drainage performance based on road segment functions, climate conditions, traffic load levels, and engineering scenario requirements, and selects the texture parameter scheme with the best overall performance from the Pareto optimal frontier based on preference weights. The parameter output module is used to generate standardized design results and complete engineering output. Specifically, it transforms the selected optimal solution into a standardized design solution that includes texture geometry parameters, engraving process adaptation suggestions, and working condition adaptability descriptions. It also generates a quantitative correlation report of texture parameters and performance indicators, and generates engineering-usable files in various formats. The visualization and interaction module is used to realize the visualization of data and results throughout the entire process. Specifically, it includes multi-dimensional retrieval visualization of multi-dimensional road condition database, visualization of texture parameter-performance index correlation curve of texture performance mapping model, visualization of performance trade-off relationship graph of Pareto optimal frontier, and visualization of performance comparison between optimal solution and traditional solution, to help designers make design decisions intuitively.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: In this invention, a multi-dimensional pavement condition database with three-dimensional correlation between working conditions, parameters, and performance is constructed by collecting three types of core data: historical engineering data, indoor and outdoor experimental data, and construction site data. This breaks through the reliance of traditional design on engineers' experience and achieves quantitative and refined design. By using the CatBoost algorithm to construct a high-precision texture performance mapping model and the NSGA-Ⅲ algorithm for global parallel search, a Pareto optimal frontier is generated and the optimal solution is selected by combining engineering preference weights. This solves the limitations of single performance optimization, achieves multi-objective collaborative balance, and transforms the optimal solution into a standardized design result containing construction technology and quality acceptance requirements. A quantitative correlation report is output simultaneously to ensure the feasibility of project implementation. Attached Figure Description

[0016] Figure 1This is a flowchart of the present invention; Figure 2 Flowchart for establishing the multidimensional road condition database in this invention; Figure 3 A flowchart for establishing the texture performance mapping model in this invention; Figure 4 This is a flowchart of the system processing in this invention. Detailed Implementation

[0017] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0018] Reference Figure 1 As shown, the intelligent optimization design method and system for rigid pavement texture carving parameters collects historical engineering data, indoor and outdoor experimental data, and construction site test data to construct a multi-dimensional pavement condition database containing texture parameters and performance indicators. Obtain non-numerical pavement condition data from the multidimensional pavement condition database, merge low-frequency non-numerical pavement condition data, process the multidimensional pavement condition database, and divide the multidimensional pavement condition database into training data and validation data based on stratified sampling of pavement conditions. Based on the multidimensional pavement condition database and historical engineering data, the priority of pavement conditions is classified, and a risk coefficient benchmark range is set to correspond to the priority classification of pavement conditions, thereby quantifying the risk level of pavement conditions. A comprehensive coefficient benchmark is obtained based on the risk level and risk coefficient range of the road surface conditions, and the balance and consistency with engineering logic are verified. Based on the features of a multi-dimensional road condition database, the control parameters of the CatBoost algorithm are set, the training data and validation data are fed into the algorithm and the categorical features are automatically identified, and the sample weights are bound to the multi-dimensional data. The training data is used to conduct hierarchical training of the model, and the validation data is simultaneously connected to verify the model in real time and limit the training step size. After completion, it is labeled as a texture performance mapping model. The model is subjected to hierarchical cross-validation under different operating conditions. An engineering optimization threshold is set. If the threshold is not met, the algorithm parameters are iteratively adjusted. Once the threshold is met, the model is solidified, and the texture parameters are extracted to map the performance weights corresponding to the texture parameters and performance indicators in the model. Define the engineering constraints of texture parameters, select a multi-objective evolutionary algorithm, conduct a global parallel search, generate the Pareto optimal front, set engineering preference weights, screen the texture parameter scheme with the best overall performance, generate a standardized design scheme, and simultaneously output a quantitative correlation report of texture parameters and performance indicators.

[0019] Reference Figure 2As shown, specifically, the collection of historical engineering data, indoor and outdoor test data, and construction site inspection data to construct a multi-dimensional pavement condition database containing texture parameters and performance indicators specifically includes: Extract texture parameters and performance indicators from the design phase, construction records and long-term performance tracking data from historical engineering data, test data and experimental conditions of specimens with different texture parameters from indoor and outdoor experimental data, actual texture parameter data and actual pavement performance data from construction site inspection data, and collectively refer to them as multidimensional acquisition data. The design texture parameters and performance data refer to the initial design parameters and corresponding expected performance indicators of the rigid pavement texture. The construction record refers to the construction process parameters, construction environment and construction process data of the texture carving. The long-term performance tracking data are the texture performance decay data and disease correlation data during the service life of the pavement. The indoor and outdoor experimental data are mainly the core performance data of anti-skid, noise reduction and drainage obtained by the equipment test of different texture geometric shape specimens made in the laboratory, as well as the experimental environment and load conditions during the test. The actual texture parameter data are mainly the three-dimensional morphological parameters of the texture measured after completion. The actual pavement performance data are mainly the pavement performance sampling data in the early stage of completion. Based on the three main data, a closed-loop verification of design-construction-performance is constructed, and multiple data sources are realized to avoid single data. All data are uniformly named, standardized and collected, and labeled as multi-dimensional collected data. Filter the multidimensional collected data, where the texture parameter data should include at least texture depth, texture width, texture spacing, texture distribution form and texture distribution angle, and the performance index data should include at least skid resistance coefficient, noise reduction performance and drainage performance. Then, associate the multidimensional collected data with road surface type, climate conditions and traffic load level to identify working condition characteristics. The multidimensional collected data was standardized and screened to remove invalid samples with inconsistent data formats, missing test conditions, or abnormal performance data. The screening criteria for texture parameters were texture depth of 0.5mm-1.5mm, texture width of 2mm-5mm, and texture spacing of 15mm-30mm. Samples with texture distribution angles of 0 degrees to 90 degrees and angles with the driving direction of 45 degrees to 60 degrees were also classified separately. The performance indicators were screened for skid resistance coefficient of 60dB-75dB (in the marked road section environment), noise reduction performance of 60dB-75dB (at a test speed of 50 km / h), and drainage performance of 30s-120s (with a water accumulation thickness of 10mm). A classification standard for working condition characteristics was established based on road surface type, climate conditions, and traffic load level. Based on the classification standard for working condition characteristics and typical working condition association cases in historical project data, the screened texture parameters, performance indicator data, and the three types of working condition characteristics were associated to form a three-dimensional association dataset of working condition-parameter-performance. Based on the selected texture parameters, performance indicators and working condition characteristics, the data are structurally integrated to establish data correspondence and construct a multi-dimensional road working condition database. The selected texture parameters, performance indicators, and working condition characteristic data are standardized and formatted. Numerical pavement working condition data are converted to floating-point format, and categorical data are converted to standardized encoding format. Each data point is assigned a unique sample number, following the format: working condition type code - data source code - data batch code - sequence number. A hierarchical data correspondence is established based on the logic of working condition characteristics as the classification dimension, texture parameters as the input dimension, and performance indicators as the output dimension. The hierarchical data correspondence is logically verified, and abnormal correlation units are removed to ensure the accuracy and uniqueness of the data correspondence. A multidimensional pavement working condition database is built based on a distributed relational database architecture, and the processed data is stored in the basic layer, the correlation layer, and the application layer, respectively.

[0020] Reference Figure 3 As shown, specifically, the process of acquiring non-numerical and numerical pavement condition data from the multi-dimensional pavement condition database, merging low-frequency non-numerical pavement condition data, standardizing the numerical pavement condition data, and dividing the multi-dimensional pavement condition database into training and validation data based on stratified sampling according to the pavement condition includes: Non-numerical pavement condition data are extracted from the multi-dimensional working condition database. The non-numerical pavement condition data is cleaned, and the sample proportion of the cleaned non-numerical pavement condition data categories is calculated and filtered. Non-numerical pavement condition data with a sample proportion lower than the preset standard are integrated as low-frequency categories, a standardized non-numerical working condition data category system is established, and the non-numerical pavement condition data is standardized and coded. Non-numerical pavement condition data, including pavement type subclasses, climate condition subclasses, traffic load scenario subclasses, and texture distribution form subclasses, were extracted from a multi-dimensional pavement condition database. Samples with incorrect labeling and ambiguous classification were removed. Missing fields for climate conditions were filled in according to pavement condition attributes, and missing traffic load levels were filled in according to road segment registration levels. The proportion of each subclass in the multi-dimensional pavement condition database was calculated, and invalid samples with incorrect labeling and ambiguous classification were filtered out. Non-numerical pavement condition data with a proportion less than 5% were extracted and classified as low-frequency non-numerical pavement condition data. Low-frequency non-numerical pavement condition data were integrated according to the principles of similar functional characteristics, common engineering needs, and similar performance impacts. A three-level hierarchical standard was constructed based on pavement type. A standardized non-numerical working condition data category system was established, comprising a four-level coding rule: category identifier - first-level major category code - second-level sub-category code - third-level original type code (example: road type - ordinary cement concrete road surface is coded as LM-01-00-00, climate conditions - humid zone - semi-humid zone is coded as QH-01-02-01). Subsequently, the logical consistency of the standardized non-numerical working condition data category system was rationally verified based on industry standard compliance and the homology of working condition characteristics. Based on engineering safety baseline constraints and engineering feasibility requirements, the standard compliance rate was set at 98%. Based on the balance of major category sample proportions and the complementarity of sub-category data, the sample proportion of each major category in the total non-numerical data was calculated, and the result was determined by subtracting the largest proportion from the smallest proportion. The proportion of core categories is set at ≥8% based on the principles of adapting to the needs of stratified sampling and cross-validation for suitable working conditions and avoiding model bias caused by data imbalance. This is based on four dimensions: minimum effectiveness for engineering requirements, critical values ​​for data statistics, safety net requirements for risk management, and adaptability to quantitative practice. For example, if durability (freeze-thaw resistance) in high-altitude and cold regions is a core category, and its proportion is only 5%, even with a standardized score of 0.9, its contribution to the final comprehensive weight is only 0.045, far lower than skid resistance (0.5 × 0.85 = 0.425) and drainage (0.3 × 0.8 = 0.24). This would lead to design schemes neglecting freeze-thaw protection and significantly shortening the service life of the road surface. However, when the proportion is ≥8%, the contribution is ≥0.0. 72 (8% × 0.9), which can effectively force the solution to take into account the durability requirements), the maximum proportion - minimum proportion ≤ 30% is qualified (used to balance the priority of the core dimension and the synergy of multiple dimensions, to avoid the risk of engineering imbalance and model performance defects caused by excessive concentration of weight. For example, in the cold region, the skid resistance coefficient accounts for 60%, the durability accounts for 25%, and the difference is 35% (over 30%). At this time, the design solution will overemphasize the depth of the skid resistance texture, resulting in insufficient road surface freeze-thaw resistance, and the service life in the cold region will be reduced from 15 years to 8 years; if the durability accounts for 30% (the difference is 30%, which is just qualified), then the weight gradient of skid resistance and durability is reasonable, and the solution can simultaneously meet the requirements of skid resistance safety and freeze-thaw protection), to check the rationality of data balance; Numerical pavement condition data are extracted from the multidimensional working condition database and grouped into working condition combinations. Missing samples in the numerical pavement condition data are supplemented and removed. The processed numerical pavement condition data is standardized and then the processed numerical pavement condition data and the non-numerical pavement condition data encoded in the standardized non-numerical working condition data category system are referred to as the full data. Numerical pavement condition data (including numerical texture parameters and numerical performance indicators, where numerical texture parameters include texture depth, texture width, texture spacing, and texture distribution angle, and numerical performance indicators include skid resistance coefficient, noise reduction decibel value, and water receding time) are extracted from the multidimensional working condition database. The data is then grouped according to working condition combinations of pavement type, climate conditions, and traffic load level (e.g., ordinary cement concrete pavement - humid area - heavy load, steel fiber cement concrete pavement - high-altitude freezing area - medium load). Missing numerical texture parameters are filled with the mean of the same working condition combination. If the proportion of missing performance indicators is ≤3%, the median is used for filling; if it is >3%, the sample is directly removed. The processed numerical pavement condition data is standardized using the linear normalization method. Finally, the processed numerical pavement condition data and the coded non-numerical pavement condition data in the standardized non-numerical working condition data category system are referred to as the full data. Based on the combination of road surface type, climate conditions, and traffic load level, the full data is divided into training data and validation data according to a preset ratio, and the training data and validation data are verified. Based on the combination of working conditions according to road type, climate conditions, and traffic load level (typical working condition combinations include ordinary cement concrete road surface - wet area - heavy load), the full data is divided into training data and validation data according to a preset ratio of 7:3. The sample validity rate in the training data and validation data is calculated separately. If the sample validity rate reaches 98%, the integrity requirement is met. The sample ratio of working condition combination in the training data and validation data is calculated and compared with the original ratio in the full data. When the deviation between the original ratio and the sample ratio is less than 5%, the balanced distribution requirement is met.

[0021] Specifically, the process of classifying pavement conditions by priority based on a multi-dimensional pavement condition database and historical engineering data, setting a risk coefficient benchmark range to correspond to the pavement condition priority classification, and quantifying the risk level of pavement conditions includes: Based on the condition feature data in the multidimensional pavement condition database and the defect correlation data in historical engineering data, the pavement conditions of rigid pavement are classified into priority levels. The condition feature data includes three categories: pavement type, climate conditions, and traffic load level. Based on the pavement condition characteristic data (including three core data categories: pavement type, climate conditions, and traffic load level) and historical engineering defect correlation data, defect incidence rate (the percentage of projects with the target defect under the condition combination), defect severity (classified into four levels: minor, moderate, severe, and extremely severe according to the "Highway Technical Condition Assessment Standard"), defect repair cycle, and defect impact coefficient on traffic safety (calculated based on the proportion of accident statistics, ranging from 0 to 1, with higher values ​​indicating more severe impact) from the multidimensional pavement condition database, an expert evaluation method was adopted. A three-dimensional weighted evaluation method based on safety, cost, and durability was used to prioritize the pavement conditions of rigid pavements. The system is tiered and prioritized based on the logic of engineering risk gradient control, industry assessment practices, cost reduction and efficiency improvement, and the patterns of historical engineering defects. A comprehensive score of ≥80 is designated as the first priority level (80 is considered the excellent score based on the highway technical condition assessment standards, highway engineering risk assessment specifications, and historical engineering defect data statistics, avoiding omissions of high-risk scenarios and insufficient high-risk sample data). A score of 60-79 is designated as the second priority level, and a score <60 is designated as the third priority level (60 is considered the passing score based on the highway technical condition assessment standards, highway engineering risk assessment specifications, and historical engineering defect data statistics, avoiding over-design that leads to cost waste). For road conditions with different priorities, corresponding risk coefficient benchmark intervals are set so that the numerical range of the risk coefficient benchmark intervals is positively correlated with the priority level of the road conditions. For road surface conditions classified with different priorities, based on the core principles of matching risk levels with engineering management needs, non-overlapping hierarchical intervals, and traceable quantitative logic, corresponding risk coefficient benchmark intervals are set. The risk coefficient value range is uniformly [0, 1]. The larger the value, the higher the safety risk, defect risk, and engineering management difficulty of the road surface condition. Moreover, the value range of the risk coefficient benchmark interval is positively correlated with the priority level of the road surface condition: for each level of priority reduction, the lower and upper limits of the benchmark interval shift downwards simultaneously. At the same time, a reasonable interval span is maintained to adapt to the risk differences of different conditions within the same level, ensuring clear distinction between levels and accurate quantification within levels. The risk coefficient benchmark interval is set based on historical engineering risk data, industry quantitative practices, and technical process integration. Among them, the risk coefficient benchmark interval for the first priority road surface condition is set as [0.8, 1.0] (based on historical engineering risk data statistics, industry standard quantitative settings, and consistent with highway engineering risk assessment). The estimation standard uses a risk coefficient greater than or equal to 0.8 as the threshold for major risk judgment. The interval span of 0.2 covers the differences between different high-risk working conditions and avoids insufficient quantification accuracy due to an excessively large span. The benchmark interval for the risk coefficient of the second-level priority road working condition is [0.5, 0.79] (covering common risk scenarios, ensuring clear quantification boundaries between medium-risk and low-risk working conditions, and not overlapping with the benchmark interval for the risk coefficient of the first-level priority road working condition, preventing the risk coefficients of high and medium-risk working conditions from crossing, and ensuring that the risk differences between the two types of working conditions can be distinguished during subsequent comprehensive coefficient benchmark verification). The benchmark interval for the risk coefficient of the third-level priority (low-risk) road working condition is [0.2, 0.49] (adapting to low-risk scenarios; below 0.2, the parameter-performance correlation of such scenarios may be ignored during model training, affecting generalization ability, and not overlapping with the benchmark interval for the risk coefficient of the second-level priority road working condition, ensuring that there is no ambiguity in the judgment of medium and low-risk working conditions). By breaking down the risk impact dimensions of road surface conditions and combining the correlation analysis results of defect data in historical engineering data, the risk impact weight of each dimension is obtained. The risk level of different conditions is quantitatively calculated to obtain the dimensional risk coefficient of a single condition. The risk impact dimensions of road surface conditions are broken down, adhering to the principles of covering core risk sources, conforming to engineering realities, and being quantifiable with data support. Five key risk impact dimensions are identified. Combining historical engineering defect data, the risk level of each condition in different dimensions is first scored on a percentage basis. Then, the preliminary data weights of each dimension are calculated through statistical correlation analysis of historical defect data. An expert team is organized to calibrate these preliminary weights based on engineering experience and regional specificities, resulting in a comprehensive risk impact weight with a total of 1 (wherein, safety risk is set at 0.4 based on the primary bottom line of engineering risk control, defect risk is set at 0.3 based on engineering economics and service life, cost risk is set at 0.15 based on design optimization costs, construction process costs, and subsequent repair costs, durability risk is set at 0.1 based on long-term engineering benefits, and construction difficulty risk is set at 0.05 based on the impact on overall risk). Finally, the scores of each dimension are weighted and summed with the corresponding comprehensive risk impact weights to complete the risk quantification calculation, ultimately obtaining the dimensional risk coefficient of a single condition (value range [0,1], with larger values ​​representing higher risks).

[0022] Specifically, the process of obtaining a comprehensive coefficient benchmark based on the risk level and risk coefficient range of road surface conditions, and verifying the balance and consistency with engineering logic, includes: Match the risk coefficient of a single working condition with the benchmark range of the risk coefficient of the corresponding priority, and use the product of the risk coefficients of each dimension and the superposition of the lower limit of the benchmark range of the risk coefficient of the corresponding priority as the comprehensive coefficient benchmark. The risk coefficient of a single working condition is matched with the benchmark range of the risk coefficient of the corresponding priority. With the core principles of precise range matching, risk synergistic quantification, and result engineering adaptation, a unique benchmark range of risk coefficients is first locked based on the working condition priority. Then, the synergistic superposition effect of multiple risk dimensions is quantified by multiplying the risk coefficients of each dimension. Finally, it is superimposed with the lower limit of the corresponding range to obtain the comprehensive coefficient benchmark (the value range is consistent with the benchmark range of the risk coefficient of the corresponding priority to ensure that the risk level does not cross levels). The obtained comprehensive coefficient benchmark is verified in two directions to verify the balance of the comprehensive coefficient benchmark between different priority working conditions and to verify the consistency between the comprehensive coefficient benchmark and the road engineering risk logic. One-way ANOVA was used to validate the composite coefficient benchmark variables. Based on general statistical standards and historical engineering data, a significance level of 0.05 was set. A value less than 0.05 was considered significant, and the validation was considered successful. This ensured the distinguishability of risk quantification between different priority working conditions. The Spearman rank correlation coefficient between the composite coefficient benchmark and key indicators in the historical engineering defect database was calculated to ensure that the coefficient benchmark was consistent with the risk patterns of pavement engineering. Both validations had quantitative thresholds and judgment criteria. If the criteria were not met, the dimensional risk coefficient or the benchmark interval boundary was adjusted in reverse until the validation was successful.

[0023] Specifically, the step of setting CatBoost algorithm control parameters based on multi-dimensional road condition database features, feeding training and validation data into the algorithm and automatically identifying categorical features, and binding sample weights with multi-dimensional data includes: Based on the sample size, condition complexity, and feature dimensions of the multidimensional pavement condition database, the core control parameters of the CatBoost algorithm are set to form an algorithm parameter system adapted to the learning of the mapping relationship between rigid pavement texture parameters and performance indicators. Based on the sample size, condition complexity, and feature dimensions of the multidimensional pavement condition database, the core control parameters of the CatBoost algorithm are selected. The value range is set through data feature-parameter adaptation logic. The algorithm is trained in combination with training data to form an algorithm parameter system that adapts to the nonlinear mapping relationship between rigid pavement texture parameters and performance indicators. Based on the standardized non-numerical working condition data category system, the non-numerical working condition data in the sample is standardized and encoded. The encoded sample dataset containing categorical features, numerical texture parameters and performance indicators is fed into the CatBoost algorithm to identify the correlation between categorical features and performance indicators and quantify the importance of categorical features. Based on the standardized non-numerical working condition data category system (a four-level structure: category identifier - first-level major category code - second-level sub-category code - third-level original type code), the non-numerical working condition data (road surface type, climate conditions, traffic load scenarios, and texture distribution patterns) in the sample are standardized and coded. The coding strictly follows the preset rules (category identifier: road surface type LM, climate conditions QH, traffic load scenarios JT, texture distribution pattern WL, each level of coding is a 2-digit number). The coded categorical features are integrated with the processed numerical texture parameters (texture depth, width, spacing, distribution angle) and numerical performance indicators (skid resistance coefficient, noise reduction decibel value, water receding time) into a complete sample dataset. After being structured according to the categorical feature column - numerical texture parameter column - performance indicator column, it is fed into the CatBoost algorithm. The algorithm automatically identifies the categorical features. This study explores the quantification patterns of hierarchical associations and numerical features, uncovers the nonlinear mapping relationship between different category combinations and performance indicators, and employs a dual-indicator approach of split gain and permutation importance to quantify the importance of categorical features. The CatBoost algorithm's native index calculates the contribution of decision tree splits to reducing prediction loss. It iterates through all possible split points for each categorical feature, calculates the difference in loss function before and after splitting, and this difference represents the split gain of that feature at that split node. The total split gain of a single feature is the sum of its gains across all decision tree split nodes. Finally, the standardized split gain is calculated as the proportion of a single-class feature in the sum of gains. Keeping other features unchanged, the values ​​of a certain categorical feature in the validation set are randomly permuted, and the model's prediction performance is recalculated; the magnitude of the performance decrease is the permutation importance of that feature. This clarifies the priority of each categorical feature's impact on performance indicators. The comprehensive coefficient benchmark of the working condition is transformed into sample weights, which are then associated and bound with the feature data and performance index labels in the multi-dimensional road working condition database to form weighted training data units. The comprehensive coefficient benchmark for the working condition is transformed into sample weights based on linear normalization and risk adaptation principles (based on the normalization of the comprehensive benchmark coefficient). The formula for mapping to the target weight interval is: in: Based on the weights, This is the upper limit of the weight. As the lower limit of the weight, =1.0, Subsequently, the basic weights of the first-priority working conditions were modified according to the risk adaptation principle. The weight of the first-priority working conditions was increased by 0.05 for the first-level core safety scenario. Then, the target weights were calculated. All weights must fall within the range of 0.5-2.0. If they exceed this range, they are directly truncated to (0.5, 2.0) to ensure that high-risk working condition samples obtain higher learning weights. Then, they are associated and bound with the feature data (encoded non-numerical working condition features + standardized numerical texture parameters) and performance index labels (anti-skid coefficient, noise reduction decibel value, water receding time) in the multi-dimensional road working condition database to form a well-structured and clearly weighted training data unit.

[0024] Specifically, the process of using training data to perform hierarchical model training, synchronously accessing validation data for real-time verification and limiting the training step size, and then labeling the completed model as a texture performance mapping model, specifically includes: Using the three-dimensional combination of road surface type, climate conditions, and traffic load level as the basis for working condition stratification, the training data is divided into several working condition strata, and stratified training is carried out in accordance with the hierarchical order. The common correlation rules between texture parameters and performance indicators in each stratum are integrated, and the nonlinear mapping logic of texture parameters-performance indicators under different working conditions is integrated across strata. Using the three-dimensional combination of road surface type, climate conditions, and traffic load level as the core stratification basis, 45 basic working condition combinations were first decomposed according to road surface type (3 categories) × climate conditions (5 categories) × traffic load level (3 categories). Then, based on the working condition priority level (Level 1 high risk, Level 2 medium risk, Level 3 low risk), they were classified and merged to form 6 working condition layers (2 high risk, 2 medium risk, and 2 low risk). CatBoost model stratification training was carried out in the hierarchical order of high risk layer - medium risk layer - low risk layer (based on working condition priority). The model is divided into core and layer dimensions, and adopts a sequential training strategy of high-priority training followed by low-priority iterative adaptation. The first-level high-risk working condition is taken as the core of engineering safety to ensure its prediction accuracy. Then, the first-level model is used as the basis for fine-tuning to adapt to the second and third-level working conditions. A three-layer training method is adopted (with five-fold cross-validation within each layer). The common correlation between texture parameters and performance indicators within each layer is explored. Then, the nonlinear mapping logic under different working conditions is integrated through the cross-layer feature fusion mechanism of the CatBoost algorithm to ensure that the model can adapt to specific working condition requirements and has global generalization ability. Synchronously access validation data, set a performance check to be performed every 10 rounds of CatBoost decision tree iterations, and limit the model training step size to avoid model overfitting; Based on the core principles of hierarchical correspondence and identical sample distribution, validation data is extracted from a multi-dimensional road condition database according to the hierarchical proportion of the condition. This data is then input into the CatBoost algorithm in parallel with the training data. The model performance is verified every 10 rounds of CatBoost decision tree iteration. The model's fit status is determined by comparing the multi-dimensional fitting indices of the training and validation data. At the same time, the model training step size is quantitatively limited from three dimensions: iteration step size, learning step size, and feature update step size. This forms a dual anti-overfitting mechanism of real-time verification and step size constraint, ensuring that the model fully learns the correlation between texture parameters and performance indices while avoiding overfitting caused by excessive iteration or excessive step size. Once the model training meets the preset verification criteria, it is labeled as a texture performance mapping model. After the model training meets the preset three-dimensional verification standards of overall performance, core working conditions and stability, it is labeled as a texture performance mapping model according to the process of unified naming rules, parameter solidification and version labeling. This ensures that the model is traceable and reusable, and clearly adapts to the quantitative mapping requirements of rigid pavement texture parameters and anti-skid, noise reduction and drainage performance.

[0025] Specifically, the process involves performing stratified cross-validation on the model under various operating conditions, setting engineering optimization thresholds, iteratively adjusting algorithm parameters if the thresholds are not met, solidifying the model once the thresholds are met, and extracting performance weights corresponding to texture parameters and performance indicators in the model. This includes: Using road surface type, climate conditions, and traffic load level as the basis for working condition stratification, the trained texture performance mapping model is subjected to working condition stratified cross-validation. The model is trained and validated in multiple rounds of data subset rotation to evaluate its adaptability and generalization ability to different working conditions. Using a three-dimensional combination of road surface type, climate conditions, and traffic load level as the basis for working condition stratification, a stratified five-fold cross-validation method is adopted to carry out multiple rounds of data subset rotation training and validation. The samples of each working condition stratum are independently divided into 5 subsets. In each round, one subset is selected as the validation subset and the other 5 subsets are selected as the training subset. The 5 rounds of rotation are completed in the order of high-medium-low risk level. By quantifying the model performance indicators of each stratum and each round, the adaptability and generalization ability of the model to different working conditions are comprehensively evaluated. An engineering optimization threshold is set for model performance. The threshold includes the overall fitting accuracy index, the core working condition prediction accuracy index, and the performance trend consistency index. If the model performance does not meet the engineering optimization threshold, the core control parameters of the CatBoost algorithm are iteratively adjusted until the texture performance mapping model performance meets the standard. To optimize model performance, quantitative, feasible, and engineering-relevant engineering optimization thresholds covering core dimensions are set. (Different optimization thresholds are set for different indicators. The overall fitting accuracy threshold is set based on the comprehensive coefficient of determination and the comprehensive root mean square error (RMSE), and is set to 0.82 based on historical engineering data. The comprehensive coefficient of determination, mean absolute error (MAE), and maximum absolute error (MaxAE) of the core performance indicators are set as the core working condition prediction accuracy indicators, and are set to 0.85 based on the design specifications for cement concrete pavement of highways. The performance trend consistency indicators are set based on common engineering knowledge, with performance trend consistency ≥95%, cross-layer trend consistency ≥90%, and extreme working condition trend compliance rate ≥98%.) The thresholds clearly include three categories: overall fitting accuracy indicators, core working condition prediction accuracy indicators, and performance trend consistency indicators. All indicators must meet the requirements simultaneously. If any indicator fails to meet the requirements, the core control parameters of the CatBoost algorithm are adjusted according to the logic of targeted adjustment, verification feedback, and iterative optimization until all indicators meet the thresholds. Once the texture performance mapping model meets the preset optimization threshold, it is engineered and solidified to form a reusable texture performance mapping model. After iterative optimization to meet all preset engineering thresholds, the texture performance mapping model is engineered and solidified through a four-step process: solidification preparation, core content solidification, encapsulation and delivery, and reuse verification. This results in a reusable texture performance mapping model that can be directly called across projects and scenarios, ensuring that the parameters remain unchanged, the accuracy is stable, and the logic is consistent when the model is called. Extract the performance weights of each texture parameter and performance index in the texture performance mapping model, and clarify the priority of the influence of texture parameters on each performance index. A combined approach of CatBoost feature importance quantification, SHAP value attribution, and partial correlation analysis was employed to extract performance weights corresponding to each texture parameter (texture depth, width, spacing, distribution angle) and performance indicators (skid resistance coefficient, noise reduction decibel value, and water receding time) from the engineered and solidified texture performance mapping model. A weighted fusion method was used to obtain a comprehensive performance weight (following the principle of prioritizing engineering requirements and supplementing with data support, the fusion was performed in three progressive levels: single-performance dimension indicator weight fusion, core evaluation indicator weight fusion, and working condition-level weight fusion). The fused weight scores were linearly normalized to eliminate dimensional differences. Based on the safety priority principle and highway cement concrete pavement design specifications, the weights for each single performance dimension were assigned, with skid resistance coefficient set at 0.5, drainage time at 0.3, and noise reduction decibel value at 0.2. The core evaluation indicators are weighted according to the following principle: core working condition accuracy is prioritized, overall fitting accuracy is secondary, and trend consistency is the last resort. Specifically, the core working condition prediction accuracy is assigned a weight of 0.45, the overall fitting accuracy is assigned a weight of 0.35, and the performance trend consistency is assigned a weight of 0.2. The working condition stratification weights are allocated based on an engineering classification logic that prioritizes high-risk working conditions, followed by medium-risk, and supplemented by low-risk conditions. Specifically, Level 1 is assigned a weight of 0.5, Level 2 a weight of 0.35, and Level 3 a weight of 0.15. Scores for each weight are calculated, and a weighted fusion weight is calculated based on the corresponding weights. The weights are then sorted from highest to lowest to clarify the priority of each texture parameter's impact on individual and comprehensive performance indicators. The rationality of the weights and priorities is verified by considering the engineering principles of rigid pavements, ensuring that the results align with both the model's learning patterns and the actual engineering logic.

[0026] Specifically, the process of defining the engineering constraints of texture parameters, selecting a multi-objective evolutionary algorithm, conducting a global parallel search, generating a Pareto optimal front, setting engineering preference weights, screening the texture parameter scheme with the best overall performance, generating a standardized design scheme, and simultaneously outputting a quantitative correlation report of texture parameters and performance indicators includes: Based on the design specifications for cement concrete pavement of highways, engineering compliance constraints are set for the texturing parameters of rigid pavement, and the upper and lower limits of the values ​​of each texturing parameter are clarified. Based on the design specifications for cement concrete pavement of highways, engineering compliance constraints are set for four core rigid pavement texture carving parameters, and the upper and lower limits of values ​​under different traffic levels and climate zones are clarified. At the same time, parameter coordination constraint rules are supplemented to ensure that the texture parameters not only meet the specifications but also adapt to comprehensive performance requirements. Based on the requirement of coordinated optimization of multiple performance indicators of rigid pavement, a suitable multi-objective evolutionary algorithm is selected, and a multi-objective optimization solution framework is built. Combining the requirements of multi-performance index synergistic optimization of rigid pavement anti-skid safety, noise reduction comfort, and efficient drainage, the Non-Dominated Sorting Genetic Algorithm III (NSGA-III) was selected as the core solution algorithm. With the texture performance mapping model as the prediction core, engineering compliance constraints as the boundary, and performance weights as the optimization guide, a full-link multi-objective optimization solution framework was built, which includes objective function construction, decision variable definition, constraint embedding, algorithm parameter configuration, and solution process design. This ensures that the output texture parameter scheme not only meets the synergistic optimization of multiple performance indicators, but also conforms to the actual engineering requirements. Within the defined engineering constraints of texture parameters, the NSGA-Ⅲ algorithm was selected as the multi-objective evolutionary algorithm to carry out a global parallel search and generate the Pareto optimal frontier of the performance index trade-off relationship. Based on the compliance constraints of texture parameter engineering, the NSGA-Ⅲ algorithm was selected as the multi-objective evolutionary algorithm. The Das-Dennis method was adopted for the reference point generation method. The search efficiency was improved by synchronously calculating the population fitness through multiple processes and performing crossover and mutation operations in parallel. At the same time, a constraint-guided search mechanism was introduced to ensure that the search process is always carried out within the compliance range. Finally, a Pareto optimal front covering the performance trade-offs of the entire scene was generated, providing a diverse set of optimal solutions for engineering scheme selection. Based on the core requirements of specific engineering scenarios, engineering preference weights are set, and texture parameter schemes with the best overall performance are selected from the Pareto optimal frontier based on these preference weights. Combining the core requirements of different engineering scenarios for rigid pavements, an approach of quantifying core requirements, adapting to specifications, and integrating performance weights is adopted to set engineering preference weights. (A 5-point Likert scale is used to quantify the core requirements of each scenario, which is then normalized to obtain the quantified weights. The core requirement weights are then adjusted upwards based on the requirements of current highway engineering specifications and the quantified weights. Subsequently, the requirement weights are mapped to three integrated levels: single performance dimension, three major evaluation indicators, and working condition stratification, to obtain the final engineering preference weights. Example: A medium-load urban expressway in a humid area, with core requirements of drainage and flood control, noise reduction, and environmental protection; a 5-point Likert scale is used to score the core requirements of the scenario, which is then normalized to convert them into initial weights, where drainage and flood control accounts for 5 points with an initial weight of 0.333.) The initial weights for noise reduction and environmental protection were 4 points (0.267); anti-skid safety was 3 points (0.200); construction scenario was 2 points (0.133); and durability was 1 point (0.067). The calculation method was the ratio of the core scenario requirement score to the total score. Subsequently, based on current highway engineering specifications, the initial weights were adjusted, strengthening the weights corresponding to mandatory requirements. Specifically: Drainage and flood control: According to the "Specifications for Design of Cement Concrete Pavement of Highways," the drainage time for expressways in humid areas should be ≤8s, a core mandatory requirement, and the initial weight was increased to 0.35; Noise reduction and environmental protection: According to the "Environmental Quality Standard for Noise," the noise level in residential areas on both sides of main traffic arteries should be ≤60dB, a core mandatory requirement, and the initial weight was increased to 0. 0.3; Skid Resistance Safety: According to the "Specifications for Design of Cement Concrete Pavement of Highway", the skid resistance coefficient of medium-load road sections is ≥60BPN, a basic safety requirement, and the initial weight remains at 0.2; Construction Scenario: According to the "Specifications for Construction Organization Design of Highway Engineering", the construction period of urban roads is ≤30 days, a non-core requirement, and the initial weight of non-core requirements is reduced to 0.10; Durability: According to the "Specifications for Design of Cement Concrete Pavement of Highway", the design life of medium-load pavement is ≥15 years, and the durability pressure under medium-load conditions is small, so the initial weight is reduced to 0.05; The adapted requirement weights are accurately mapped to the core level to form engineering preference weights that can be directly used for model training and scheme selection: the weight of the first level remains unchanged, the weight of the second level remains unchanged, and the weight of the third level remains unchanged. The three levels are weighted by three major evaluation indicators: the overall fitting accuracy indicator is set at 0.40 to address the diverse working conditions of urban expressways; the core working condition prediction accuracy indicator is set at 0.35 to address the core needs of drainage and noise reduction; and the performance trend consistency indicator is set at 0.25 to ensure that the texture parameters and performance trends conform to engineering logic. The third level is the working condition level weight, where: Level 1 high risk (humid + medium load) is set at 0.45, addressing the core working conditions of the scenario, with concentrated drainage and noise reduction needs and key safety management points, and has the highest weight; Level 2 medium risk (humid + light load) is set at 0.4, covering conventional working conditions such as road auxiliary roads and entrances / exits, with the highest proportion and the second highest weight; and Level 3 low risk (dry + light load) is set at 0.15. Low-risk conditions account for a low proportion within the scenario and have the lowest weight. The total weight is normalized to 1. Based on this preference weight, a weighted scoring model is constructed to comprehensively score all non-dominated solutions in the Pareto optimal frontier. The texture parameter scheme with the highest score is selected to ensure that the scheme not only meets the core requirements of the scenario but also takes into account the collaborative optimization of multiple performance indicators. The selected optimal texture parameter scheme is transformed into a standardized engineering design scheme, and a quantitative correlation report of texture parameters and performance indicators is output simultaneously. Combining the optimal texture parameter schemes for typical engineering scenarios, the parameter schemes are transformed into standardized engineering design schemes that include details of core parameters, construction process standards, and quality acceptance requirements. At the same time, based on the texture performance mapping model and Pareto optimal solution data, a quantitative correlation report of parameter-performance correlation model, sensitivity analysis, and verification results is output.

[0027] Reference Figure 4 As shown, specifically, the intelligent optimization design system for rigid road surface texture carving parameters, used to implement the intelligent optimization design method for rigid road surface texture carving parameters, specifically includes: The data management module is used for data collection and processing. It is responsible for the collection, screening, and structured storage of historical engineering data, indoor and outdoor experimental data, and construction site testing data, as well as the low-frequency category merging, data cleaning, standardization processing, and condition stratified sampling of non-numerical pavement condition data to divide training and validation data. The model prediction module is used to carry the solidified texture performance mapping model and realize the prediction of performance indicators of texture parameters and working conditions. Specifically, it includes receiving the input texture parameters and corresponding working conditions data, calling the texture performance mapping model to output the predicted values ​​of core performance indicators such as anti-slip coefficient, noise reduction performance and drainage efficiency, and extracting the performance weights corresponding to each performance indicator in the model. The multi-objective optimization solution module is used to carry out multi-objective collaborative optimization of texture parameters of rigid pavement. Specifically, it includes defining the engineering constraint range of texture parameters in combination with road engineering industry standards, building an optimization solution framework by loading the NSGA-Ⅲ multi-objective evolutionary algorithm, carrying out global parallel search based on the performance evaluation results of the model prediction module, and generating Pareto optimal frontier that represents the performance trade-off relationship. The interactive decision-making module is used to realize the interactive setting of engineering preference weights and the selection of optimal solutions. Specifically, it provides designers with a weight adjustment interface, supports setting preference weights for anti-skid, noise reduction, and drainage performance based on road segment functions, climate conditions, traffic load levels, and engineering scenario requirements, and selects the texture parameter scheme with the best overall performance from the Pareto optimal frontier based on preference weights. The parameter output module is used to generate standardized design results and complete engineering output. Specifically, it transforms the selected optimal solution into a standardized design solution that includes texture geometry parameters, engraving process adaptation suggestions, and working condition adaptability descriptions. It also generates a quantitative correlation report of texture parameters and performance indicators, and generates engineering-usable files in various formats. The visualization and interaction module is used to realize the visualization of data and results throughout the entire process. Specifically, it includes multi-dimensional retrieval visualization of multi-dimensional road condition database, visualization of texture parameter-performance index correlation curve of texture performance mapping model, visualization of performance trade-off relationship graph of Pareto optimal frontier, and visualization of performance comparison between optimal solution and traditional solution, to help designers make design decisions intuitively.

[0028] The advantages of this invention lie in constructing a multi-dimensional pavement condition database by collecting historical engineering data, indoor and outdoor experimental data, and on-site construction data. Non-numerical low-frequency data is merged and encoded, and numerical pavement condition data is standardized and then stratified for training and validation data. CatBoost algorithm parameters are set based on the features of the multi-dimensional database, sample weights are bound, and stratified training is conducted. A texture performance mapping model is formed by combining real-time validation with step size constraints. After stratified cross-validation and engineering optimization threshold verification, the model is solidified, and performance weights for texture parameters and performance indicators are extracted. The engineering constraints of texture parameters are defined in accordance with highway design specifications. The NSGA-Ⅲ multi-objective evolutionary algorithm is selected for global parallel search to generate the Pareto optimal front. Preference weights are set according to the core requirements of the engineering scenario to select the optimal comprehensive performance solution. Finally, a standardized engineering design scheme and a quantitative correlation report of texture parameters and performance indicators are output. This solves the problems of traditional methods relying on experience, single performance optimization, and low design efficiency, and is suitable for the multi-performance collaborative optimization needs of rigid pavements under complex working conditions.

[0029] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A rigid pavement texture carving parameter intelligent optimization design method, characterized in that, The method comprises the following steps: Collect historical engineering data, indoor and outdoor test data, and construction site detection data to build a multi-dimensional pavement working condition database containing texture parameters and performance indicators; Obtain non-numerical and numerical pavement working condition data from the multi-dimensional working condition database, merge the non-numerical pavement working condition data of the low-frequency category, standardize the numerical pavement working condition data, and divide the multi-dimensional pavement working condition database into training data and verification data according to stratified sampling of working conditions; Classify the priority of pavement working conditions according to the multi-dimensional pavement working condition database and historical engineering data, set the risk coefficient reference interval corresponding to the priority classification of pavement working conditions, and quantify the risk degree of pavement working conditions; Obtain the comprehensive coefficient reference according to the risk degree of pavement working conditions and the risk coefficient interval, and verify the balance and engineering logic consistency; Set the CatBoost algorithm control parameters based on the characteristics of the multi-dimensional pavement working condition database, input the training data and verification data into the algorithm, and automatically identify the category type features, bind the sample weight and multi-dimensional data; Use the training data to carry out stratified training of the model, simultaneously access the verification data for real-time verification and limit the training step length, and complete the labeling of the texture performance mapping model; Perform stratified cross-validation on the model, set the engineering optimization threshold, and iteratively adjust the algorithm parameters if the threshold is not met, and solidify the model after the threshold is met, and extract the performance weight corresponding to the texture parameters and performance indicators in the texture parameter mapping model; Define the texture parameter engineering constraint range, select a multi-objective evolutionary algorithm, perform global parallel search, generate a Pareto optimal frontier, set the engineering preference weight, select the texture parameter scheme with the optimal comprehensive performance, generate a standardized design scheme, and simultaneously output a texture parameter-performance indicator quantitative correlation report.

2. The rigid pavement texture carving parameter intelligent optimization design method according to claim 1, characterized in that, The method comprises the following steps: Extract the texture parameters and performance indicators in the design stage, construction records, and long-term performance tracking data from the historical engineering data, test data of different texture parameter specimens and experimental conditions from indoor and outdoor experimental data, and actual texture parameter data and pavement performance data from construction site detection data, and collectively refer to them as multi-dimensional collection data; Filter the multi-dimensional collection data, wherein the texture parameter data at least includes texture depth, texture width, texture spacing, texture distribution form, and texture distribution angle, and the performance indicator data at least includes skid resistance coefficient, noise reduction performance, and drainage performance, and associate the multi-dimensional collection data with pavement type, climate condition, and traffic load grade for working condition characteristics; According to the filtered texture parameters, performance indicators, and working condition characteristics, perform structured integration, establish data correspondence, and build a multi-dimensional pavement working condition database.

3. The rigid pavement texture carving parameter intelligent optimization design method according to claim 2, characterized in that, The method comprises the following steps: Obtain non-numerical and numerical pavement working condition data from the multi-dimensional working condition database, merge the non-numerical pavement working condition data of the low-frequency category, standardize the numerical pavement working condition data, and divide the multi-dimensional pavement working condition database into training data and verification data according to stratified sampling of working conditions; The non-numeric road surface working condition data in the multi-dimensional working condition database is extracted, the non-numeric road surface working condition data is cleaned, the sample proportion of the cleaned non-numeric road surface working condition data category is calculated and screened, the non-numeric road surface working condition data with a sample proportion lower than a preset standard is integrated as low-frequency category non-numeric road surface working condition data, a standardized non-numeric working condition data category system is established, and the non-numeric road surface working condition data is standardized and coded; The numeric road surface working condition data in the multi-dimensional working condition database is extracted, and the working condition combinations are grouped. The missing samples in the numeric road surface working condition data are supplemented and removed. The processed numeric road surface working condition data is standardized, and the processed numeric road surface working condition data and the coded non-numeric road surface working condition data in the standardized non-numeric working condition data category system are collectively referred to as full-quantity data. The full-quantity data is divided into training data and verification data according to a preset proportion based on the working condition combinations of road surface types, climate conditions, and traffic load grades, and the training data and the verification data are checked.

4. The rigid pavement texture carving parameter intelligent optimization design method according to claim 3, characterized in that, The priority of the road surface working condition is classified according to the multi-dimensional road surface working condition database and historical engineering data, a risk coefficient benchmark interval is set, which corresponds to the priority classification of the road surface working condition, the risk degree of the road surface working condition is quantified, and the method specifically includes: The road surface working condition of rigid pavement is divided into priority levels based on the working condition characteristic data in the multi-dimensional road surface working condition database and the disease correlation data in the historical engineering data, wherein the working condition characteristic data includes three types of data, i.e., road surface type, climate condition, and traffic load grade; For different priority road surface working conditions, corresponding risk coefficient benchmark intervals are set, so that the numerical range of the risk coefficient benchmark interval is in positive correlation with the priority level of the road surface working condition; The risk impact dimensions of the road surface working condition are disassembled, the risk impact weight of each dimension is obtained by combining the correlation degree analysis result of the disease data in the historical engineering data, the risk degree of different working conditions is quantitatively calculated, and the dimension risk coefficient of a single working condition is obtained.

5. The rigid pavement texture carving parameter intelligent optimization design method according to claim 4, characterized in that, The comprehensive coefficient benchmark is obtained according to the risk degree of the road surface working condition and the risk coefficient interval, and the balance and engineering logic consistency are checked, and the method specifically includes: The dimension risk coefficient of a single working condition is matched with the risk coefficient benchmark interval of the corresponding priority, and the product of the dimension risk coefficient and the lower limit value of the risk coefficient benchmark interval of the corresponding priority is taken as the comprehensive coefficient benchmark; The obtained comprehensive coefficient benchmark is checked in both directions, the balance of the comprehensive coefficient benchmark between different priority working conditions is checked, and the consistency of the comprehensive coefficient benchmark and the road surface engineering risk logic is checked.

6. The rigid pavement texture carving parameter intelligent optimization design method according to claim 5, characterized in that, The CatBoost algorithm control parameters are set based on the characteristics of the multi-dimensional road surface working condition database, the training data and the verification data are input into the algorithm, and the class type features are automatically identified, and the sample weight and the multi-dimensional data are bound, and the method specifically includes: Based on the sample size, working condition complexity, and feature dimension of the multi-dimensional road surface working condition database, the core control parameters of the CatBoost algorithm are set, and an algorithm parameter system suitable for learning the mapping relationship between the rigid pavement texture parameters and the performance indicators is formed. According to the standardized non-numerical working condition data category system, the non-numerical working condition data in the sample is standardized and coded, and the coded sample data set containing category type characteristics, numerical texture parameters and performance indicators is input into the CatBoost algorithm to identify the association rule between the category type characteristics and the performance indicators, and to quantify the importance of the category type characteristics; The comprehensive coefficient benchmark of the working condition is converted into sample weight, which is associated and bound with the feature data and performance indicator label in the multi-dimensional pavement working condition database to form a training data unit with weight.

7. The method of claim 6, wherein, The model hierarchical training is carried out using the training data, the verification data is accessed in real time to check and limit the training step length, and after completion, it is labeled as a texture performance mapping model, which specifically includes: Taking the three-dimensional working condition combination of pavement type, climate condition and traffic load level as the working condition hierarchical basis, the training data is divided into several working condition layers, and hierarchical training is carried out in order according to the hierarchical order, the common association rule of texture parameters and performance indicators in each level is fused, and the texture parameter-performance indicator nonlinear mapping logic under different working conditions is fused across levels; The verification data is accessed in real time, and performance verification is performed once every 10 rounds of CatBoost decision tree iteration, and the model training step length is limited to avoid overfitting of the model; After the model training meets the preset verification standard, it is labeled as a texture performance mapping model.

8. The method of claim 7, wherein, The working condition hierarchical cross-validation of the model is carried out, an engineering optimization threshold is set, if it does not meet the standard, the algorithm parameters are iteratively adjusted, if it meets the standard, the model is solidified, and the performance weight corresponding to the texture parameters and performance indicators in the texture parameter mapping model is extracted, which specifically includes: Taking the working condition combination of pavement type, climate condition and traffic load level as the working condition hierarchical basis, the working condition hierarchical cross-validation of the trained texture performance mapping model is carried out, and the adaptability and generalization ability of the model to different working conditions are evaluated according to the training and verification of multiple rounds of data subsets; An engineering optimization threshold is set for the model performance, which includes overall fitting accuracy indicators, core working condition prediction accuracy indicators and performance trend consistency indicators, if the model performance does not meet the engineering optimization threshold, the core control parameters of the CatBoost algorithm are iteratively adjusted until the texture performance mapping model performance meets the standard; After the texture performance mapping model performance meets the preset optimization threshold, it is solidified for engineering, forming a reusable texture performance mapping model; The performance weight corresponding to each texture parameter and performance indicator in the texture performance mapping model is extracted, and the priority of the influence of the texture parameters on each performance indicator is clarified.

9. The method of claim 8, wherein, The engineering constraint range of the texture parameters is defined, a multi-objective evolutionary algorithm is selected, global parallel search is carried out, a Pareto optimal frontier is generated, an engineering preference weight is set, the texture parameter scheme with the best comprehensive performance is selected, a standardized design scheme is generated, and a texture parameter-performance indicator quantitative association report is output simultaneously, which specifically includes: According to the rigid pavement texture carving parameters, the engineering compliance constraint range is set, and the upper and lower limits of the values of each texture parameter are clarified. According to the demand of rigid pavement multi-performance index collaborative optimization, the adaptive multi-objective evolutionary algorithm is selected to build the multi-objective optimization solution framework; Within the defined texture parameter engineering constraint range, NSGA-Ⅲ algorithm is selected as the multi-objective evolutionary algorithm to carry out global parallel search and generate the Pareto optimal front of performance index trade-off relationship; Combined with the core demand of specific engineering scene, the engineering preference weight is set, and the optimal texture parameter scheme with the best comprehensive performance is selected from the Pareto optimal front based on the preference weight; The selected optimal texture parameter scheme is converted into the engineering standardized design scheme, and the texture parameter-performance index quantitative correlation report is output synchronously.

10. A rigid pavement texture carving parameter intelligent optimization design system for implementing the rigid pavement texture carving parameter intelligent optimization design method of the above claims, characterized in that, It includes: Data management module, used for data collection and processing, responsible for the collection, screening, structured storage of historical engineering data, indoor and outdoor experimental data, construction site detection data, and the low-frequency category combination, data cleaning, standardization processing and working condition stratified sampling division of training data and verification data of non-numerical pavement working condition data; Model prediction module, used to carry the cured texture performance mapping model, realize the performance index prediction of texture parameters and working condition characteristics, including receiving input texture parameters and corresponding working condition data, calling texture performance mapping model to output anti-skid coefficient, noise reduction performance, drainage efficiency core performance index prediction value, and extracting performance weight of texture parameters and each performance index in the model; Multi-objective optimization solution module, used for multi-objective collaborative optimization of rigid pavement texture parameters, including defining texture parameter engineering constraint range combined with road engineering industry specification, loading NSGA-Ⅲ multi-objective evolutionary algorithm to build optimization solution framework, carrying out global parallel search based on performance evaluation results of model prediction module, and generating Pareto optimal front representing performance trade-off relationship; Interactive decision-making module, used to realize interactive setting of engineering preference weight and selection of optimal scheme, including providing weight adjustment interface for designers, supporting setting of preference weight of anti-skid, noise reduction and drainage performance according to road section function, climate condition, traffic load grade engineering scene demand, and selecting the optimal texture parameter scheme with the best comprehensive performance from the Pareto optimal front based on the preference weight; Parameter output module, used to generate standardized design results and complete engineering output, including converting the selected optimal scheme into the standardized design scheme containing texture geometric parameters, carving process adaptation suggestion and working condition adaptability description, synchronously generating texture parameter-performance index quantitative correlation report, and generating engineering available files in multiple formats; Visual interactive module, used to realize visual presentation of whole-process data and results, including multi-dimensional search visualization of multi-dimensional pavement working condition database, texture parameter-performance index correlation curve visualization of texture performance mapping model, performance trade-off relationship atlas visualization of Pareto optimal front, and performance comparison visualization of optimal scheme and traditional scheme, to assist designers to complete design decision intuitively.

Images