CN-121980190-A - Pavement performance prediction method for rural simple pavement

CN121980190ACN 121980190 ACN121980190 ACN 121980190ACN-121980190-A

Abstract

The invention discloses a pavement performance prediction method of a rural simple pavement, which comprises the steps of data standardization, construction of an original matrix based on disease evaluation values, standardization, formalization quantitative analysis, acquisition of disease, climate and traffic data, calculation of weights and pavement damage condition indexes through dynamic weight functions, multi-level summarization, construction of a rural highway usage performance evaluation system, PQI evaluation based on the sampled data, feature selection, system analysis of disease types, screening of main diseases through optimal subset regression, LASSO regression and random forests, calculation of index entropy and weights, quantification of disease entropy and standardization, and calculation of entropy weights based on information utility values. The invention solves the problems of lack of systematic research on rural simple paving materials, neglecting maintenance intervention influence, unscientific disease weight distribution and poor model generalization caused by grading dependence on manual experience in the prior art, so as to improve scientificity and accuracy of rural road maintenance.

Inventors

HAO JIANYUN
LUO SHAOJUN
XIAO YONGJIAN
LI GUOHUI
QI XIANGYU
WANG YI
YANG JING
YAN YUAN
SHAO GUOXIANG
HE YIPENG
LIU YINGBO
LI ZHENGMING

Assignees

云南省公路科学技术研究院

Dates

Publication Date: 20260505
Application Date: 20260208

Claims (10)

1. The pavement performance prediction method for the rural simple pavement is characterized by comprising the following steps of: Step S1, data standardization, namely constructing an original data matrix based on various disease evaluation values of a project to be evaluated, and carrying out standardization processing on elements in the matrix; Step S2, formalized quantitative analysis, namely constructing a road evaluation system based on a multi-level sampling structure, collecting disease data, climate environment factors and traffic data of detection points, calculating weights of various diseases through a dynamic weight function, further calculating pavement damage rate and pavement damage condition index, and carrying out multi-level summarization and statistical analysis; Step S3, constructing a rural highway pavement performance evaluation system, namely evaluating by adopting a pavement technical condition index PQI based on sampling detection data; S4, performing systematic analysis on various disease types of the rural highway simple pavement, and screening out main disease types affecting pavement condition indexes through comprehensive analysis of three methods of optimal subset regression, LASSO regression and random forest; And S5, calculating an index entropy value and a weight, namely receiving main disease types screened through feature selection, quantitatively analyzing entropy values of various disease indexes, performing standardization, and calculating entropy weights of various diseases based on the information utility value.
2. The method for predicting the pavement performance of a rural simple pavement according to claim 1, wherein the step S1 comprises: constructing an original data matrix Wherein Is the first Under the disease category The evaluation value of each item is calculated, As the total number of disease types, The construction form of the original data matrix R is as follows: The calculation formula for carrying out standardized processing on matrix elements is as follows: Wherein, the Respectively represent the indexes in the evaluation area Maximum and minimum values of (a).
3. The method for predicting pavement performance of rural simple pavement according to claim 1, wherein in the step S2, a multi-level sampling structure is constructed and a basic index is calculated by: Defining a set of all regions Wherein the method comprises the steps of Represent the first Defining a subset of sampled regions Wherein the method comprises the steps of The first selected for sampling Defining a county collection sampled in each region d Wherein the method comprises the steps of Represent the first Defining road set sampled and selected in every county Wherein the method comprises the steps of Represent the first Defining a set of disease types Wherein the method comprises the steps of Represent the first Defining detection point set of every road Wherein the method comprises the steps of Is the first First road A plurality of detection points with a detection point spacing of Rice; Calculating a detection point Is the first of (2) Area of disease-like disease Wherein, the method comprises the steps of, Represent the first The quasi-diseases are detected at the point of detection The length of the upper part, Representing average width, calculating detection points Total pavement surface area of (2) Wherein, the Is the total number of disease types; Collecting each detection point Climate environment factors and traffic data related to diseases form a multi-variable set Wherein, the Is the detection point Upper first The factors that influence the class of the factors, Representing the total number of influence factor categories; by a dynamic weighting function Calculating a detection point Location No Dynamic weighting of road surface damage When the variables are in a linear relationship with each other, Wherein, the Is the model parameter that needs to be fitted, When the variable is high-dimensional complex data, Wherein, the Representing a learnable nonlinear predictive model; is a parameter of the model and is a parameter of the model, Represent the first Coefficients of the individual feature variables; By an objective function For parameters And optimizing the process, wherein, the process comprises the steps of, Representing the global sample size for model training; A true value representing the road surface damage condition index PCI; Is the predicted value of the model The predictive model Expressed as: Wherein, the Is a parameter estimation value; To pair(s) And performing nonlinear feature mapping on the medium variable to obtain feature components.
4. The method for predicting the pavement performance of a rural simple pavement according to claim 3, wherein in the step S2, the pavement damage condition index of each level is obtained by the following summary calculation formula: Based on the multi-level sampling structure and disease data of each detection point, the following summary calculation is sequentially executed: Calculating the road surface breakage rate of the point Index of road surface damage condition : Wherein, the Is the total number of disease types; represent the first The cumulative area of road surface damage; Indicating the detection point Is a road surface investigation area; Is shown at the detection point Location No Dynamic weights for road surface damage; Representing model parameters, calculating road surface damage condition index of road r Wherein, the Representing the total number of detection points of the r-th road; Calculating road surface damage condition index of county c Road surface damage condition index of region d And provincial index Wherein, the Representing the total number of sampled roads in county c, Representing the total number of sampling counties in region d, Representing the total number of sampling regions in full province; based on the multi-level sampling structure and the pavement performance indexes of each level Determining the corresponding confidence interval; The confidence interval is for the road r: Wherein, the The PCI mean value of the road r is indicated, Representing road level standard deviation; is a confidence level The standard normal quantile below; the confidence interval is for county c: Wherein, the Representing the PCI mean value of county c, Representing county level standard deviation; the confidence interval is for region d: Wherein, the Represents the PCI mean value for region d, Standard deviation of regional level; the confidence interval is for the province level: Wherein, the A PCI mean value representing a province level, Representing the standard deviation of the full province.
5. The method for predicting pavement performance of rural simple pavement according to claim 1, wherein in the step S2, the weight model is optimized by the following dynamic update method: combining new annual detection data with historical data along with accumulation of data and environmental change, and dynamically updating weights by using full data Dynamic weighting function Simultaneously introducing time attenuation factors To reduce the effect of the over-time data, sample Is the post-decay weight of (2) The calculation is as follows: Wherein, the Representing a sample First, the The initial dynamic weight of road surface damage is similar, Time expressed as sample distance from the current year; Representing a sample Is a time decay weight of (2); Weighting the attenuated weight Performing global normalization processing to obtain a sample Final weights for model updates Where N represents the total number of samples and l represents the total number of disease types.
6. The method for predicting road surface performance of rural simple pavement according to claim 1, wherein in the step S2, the influence of the characteristic variables is evaluated by the following global sensitivity analysis method: the global sensitivity analysis adopts a Sobol method based on variance decomposition, and the method comprises the steps of representing a model as a function form Representing an input feature component; Function of Is decomposed into the sum of the variances resulting from the interaction of the individual parameters: Wherein, the Parameters (parameters) The resulting variance; Is a parameter Variance due to interactions; Showing the variance resulting from the co-action of the m parameters; Based on the variance decomposition, parameters are calculated First order sensitivity index of (a) Second order sensitivity index And an overall sensitivity index Wherein, the To remove parameters Variance due to the combined action of other parameters; And evaluating the influence degree of each variable on the road surface performance index according to the first-order sensitivity index, the second-order sensitivity index and the total sensitivity index.
7. The method for predicting the road surface performance of the rural simple pavement according to claim 1, wherein in the step S3, the PQI is obtained by weighting calculation of the road surface damage condition index PCI and the road surface running quality index RQI, and the calculation formula is as follows: Wherein, the Weights in PQI for PCI; weights for RQI in PQI; the PCI passing area road surface damage rate The calculation formula is as follows: Wherein, the Is a model parameter; From parameters The two-dimensional model is jointly formed and used for realizing nonlinear conversion of the area breakage rate in the model so as to accurately reflect the breakage characteristics of different road surface types; The said Road surface breakage rate passing through each detection point The weight calculation is obtained by the calculation formula: wherein K represents the total number of regions; Indicating the detection point Is a road surface investigation area; The said The area of each road surface damage at each detection point and the corresponding dynamic weight are calculated, and the calculation formula is as follows: Wherein, the Indicating the detection point Is the first of (2) Disease-like area; Is shown at the detection point Location No The dynamic weight of road surface damage is similar to that of road surface damage, Is the type of road surface damage; Is the total number of damage types.
8. The method for predicting the pavement performance of the rural simple pavement according to claim 1, wherein the implementation manner of the LASSO regression in the step S4 comprises: construction of a Linear regression model Wherein the method comprises the steps of In response to the vector of variables, As a matrix of the independent variables, As a vector of the regression coefficients, Is a random error term vector; by introducing L1 norm penalty term Solving such that the objective function The minimum regression coefficient estimated value enables the regression coefficient of partial independent variables to be compressed to 0, thereby realizing variable selection, In order to adjust the parameters of the device, As an independent variable matrix The total number of independent variables in the set, Is an index of the argument.
9. The method for predicting the pavement performance of the rural simple pavement according to claim 1, wherein the implementation manner of the random forest in the step S4 comprises the following steps: For each decision tree in a random forest Wherein the method comprises the steps of For the purpose of the decision tree indexing, Representing the number of decision trees in the random forest, performing the following operations: Matrix of OOB data input into the decision tree Predicting to obtain a prediction result And calculates a predicted value Mean square error with true value B Holding matrix Other characteristics of (a) are unchanged, and the (b) is disturbed The characteristic value sequence of each characteristic is obtained to obtain a rearranged matrix Matrix pairs using the same decision tree Predicting to obtain a prediction result And calculates a predicted value Mean square error with true value B Calculating an importance score of the feature on the decision tree Traversing all decision trees in the random forest, and calculating the average importance score of the feature And evaluating the importance of the features according to the average importance score and selecting.
10. The method for predicting the pavement performance of a rural simple pavement according to claim 1, wherein the step S5 comprises: Calculate the first Entropy value of disease-like Wherein, the Normalizing the coefficients for entropy values; The total number of the evaluation index items; Is the first Under the disease category Specific gravity of index value of individual item, wherein Is that Disease of the first kind The specific calculation mode of the evaluation values of the individual items is shown in the step S1; Based on the entropy value Calculate the first Entropy weight of disease-like Wherein, the Is the total number of disease types; Is the information utility value of the ith disease.

Description

Pavement performance prediction method for rural simple pavement Technical Field The invention belongs to the technical field of road engineering and traffic transportation, and particularly relates to a pavement performance prediction method of a rural simple pavement. Background In order to meet the travel demands of rural residents, simple pavement materials such as gravel pavement, elasto-stone pavement, block pavement and the like with relatively low manufacturing cost are generally adopted in the road construction of small traffic road sections. The selection not only reduces the construction cost, but also improves the traffic capacity of the road to a certain extent, so as to more effectively meet the travel demands of local residents and promote the sustainable development of economy. However, simple paved roads also face challenges in terms of durability and road maintenance. At present, aiming at simple pavement surfaces such as block and the like, systematic damage type definition, calculation methods and special technical condition assessment models are lacked, so that the detection and assessment of the simple pavement surfaces are difficult to scientifically and standardize. Therefore, in order to scientifically and reasonably manage rural highways, research on a method for detecting and evaluating technical conditions of simple pavement of rural highways is required. The current research focuses on highways and urban roads, and lacks systematic attention on simple pavement surfaces with various types, such as sand stones, precast blocks, boulders and the like which are widely existing in rural areas, and the structural complexity and disease mechanism of the simple pavement surfaces do not form unified standards and long-term data support, so that quantitative management and scientific maintenance are difficult to realize. Meanwhile, the existing method generally ignores the intervention effect of actual maintenance measures on performance change in the road service period, so that the predicted result has deviation from the actual road condition, and the practicability and decision support capability of the model are weakened. In addition, the prior art lacks of fine treatment on the weight distribution of disease types, and is difficult to accurately reflect the differential influence of different diseases on the road performance by adopting a unified treatment mode, and the model generalization capability is weaker, so that the method is difficult to adapt to complicated and changeable environmental conditions in rural areas. The current road condition classification mainly depends on manual experience, and lacks an objective and unified classification method, so that the accuracy of an evaluation result is insufficient, and the reasonable distribution of maintenance resources is further influenced. The problems limit the scientificity and practicability of the technical condition assessment of the simple pavement of the rural highways, and development of a method capable of meeting diversified requirements, quantitative maintenance intervention, refined disease weight distribution and objective grading assessment is needed to promote high-quality management and maintenance of the rural highways. Disclosure of Invention The invention aims to provide a pavement performance prediction method for rural simple pavement, which solves the problems of lack of systematic research, neglecting maintenance intervention influence, unscientific disease weight distribution and poor model generalization caused by grading dependence on manual experience of rural simple pavement materials in the prior art, so as to improve the scientificity and the accuracy of rural road maintenance. The technical scheme adopted by the invention is that the pavement performance prediction method of the rural simple pavement comprises the following steps: Step S1, data standardization, namely constructing an original data matrix based on various disease evaluation values of a project to be evaluated, and carrying out standardization processing on elements in the matrix; Step S2, formalized quantitative analysis, namely constructing a road evaluation system based on a multi-level sampling structure, collecting disease data, climate environment factors and traffic data of detection points, calculating weights of various diseases through a dynamic weight function, further calculating pavement damage rate and pavement damage condition index, and carrying out multi-level summarization and statistical analysis; Step S3, constructing a rural highway pavement performance evaluation system, namely evaluating by adopting a pavement technical condition index PQI based on sampling detection data; S4, performing systematic analysis on various disease types of the rural highway simple pavement, and screening out main disease types affecting pavement condition indexes through comprehensive analysis of three methods of optimal subset regr