Road network level highway pavement structure strength prediction method

Abstract

The invention discloses a road network level highway pavement structure strength prediction method, and relates to the technical field of road engineering detection, and the method comprises the steps: S1, collecting and building a road network level highway pavement technical condition detection data set comprising a pavement damage condition index PCI, a pavement driving quality index RQI and a pavement structure strength index PSSI; s2, calculating feature weights of the PCI and the RQI in the detection data set; s3, on the basis of the feature weights of the PCI and the RQI, performing parameter optimization on the KNN algorithm by using a mucus algorithm, and constructing a final pavement structure strength prediction model; s4, the final pavement structural strength prediction model is used for predicting the structural strength of the pavement to be predicted, prediction accuracy is improved, and the method has the advantages of being complete in theory, high in optimizing capacity and convenient to implement.

Description

Technical Field

[0001] The present invention relates to the technical field of road engineering detection, and in particular to a method for predicting the strength of a road network-level highway pavement structure. Background Art

[0002] Pavement structure strength is an important indicator for implementing pavement maintenance decisions, and is related to the accuracy and rationality of maintenance plan preparation and maintenance fund arrangements.

[0003] At present, the pavement structure strength index PSSI is either calculated point by point through a drop-weight deflectometer, automatic deflectometer and other equipment, which has defects such as slow detection speed, low efficiency and high safety risks, or is tested by a laser high-speed pavement deflection measuring instrument, which has defects such as high detection cost and low data accuracy. When the above two types of deflection detection equipment are used for pavement structure strength detection of ordinary national and provincial trunk roads at the provincial road network level, their defects of high safety risks and high detection costs are particularly obvious.

[0004] In addition, from the perspective of pavement structure strength prediction, the reasonable determination of the gradient threshold of the grouped PSSI gradient value calculation method remains to be discussed. Secondly, the use of linear regression to process data within the PSSI gradient threshold range will weaken the sensitivity of the data and reduce the accuracy of the prediction. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a road network-level highway pavement structure strength prediction method which solves the problem that it is difficult to achieve high-precision prediction of the road pavement structure strength of ordinary national and provincial trunk highways at the road network level due to the actual situation that the road network mileage is long, there are many characteristic factors, but data collection is difficult and the data regularity is poor.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a method for predicting the strength of a road network-level highway pavement structure, comprising: S1. Collect and establish a road damage index PCI , Road Ride Quality Index RQI and pavement structure strength index PSSI A road network-level highway pavement technical condition detection dataset; S2. Calculate the detection data set PCI and RQI The feature weights of S3, based on PCI and RQI The feature weights of KNN The algorithm performs parameter optimization and constructs the final pavement structure strength prediction model; S4. Use the final pavement structure strength prediction model to predict the structural strength of the pavement to be predicted.

[0007] Further: S2 includes: S21. Calculate and detect data set PCI and PSSI The mutual information and RQI and PSSI The mutual information of S22, according to PCI and PSSI The mutual information and RQI and PSSI The mutual information of the detection data set is calculated PCI and RQI The feature weights of .

[0008] Further: In S21, the detection data set is calculated PCI and PSSI The mutual information and RQI and PSSI The expression of mutual information is:

[0009]

[0010] in, express PCI and PSSI The mutual information of express RQI and PSSI The mutual information of To detect the data set PCI and PSSI The joint probability distribution of To detect the data set RQI and PSSI The joint probability distribution of , and The detection data set is PCI , RQI and PSSI The marginal probability distribution of .

[0011] Further: In S22, the detection data set is calculated PCI and RQI The expression of feature weight is:

[0012]

[0013] in, for PCIThe feature weights of for RQI The feature weights of .

[0014] Further: S3 includes: S31. Construct the distance parameter between any two data samples based on the feature weights of PCI and RQI. p The distance equation of S32, using KNN algorithm, distance parameter based on distance equation p and neighbor parameters K value, construct a pavement structure strength prediction model; S33, using slime mold algorithm KNN The algorithm performs parameter optimization and obtains the final pavement structure strength prediction model.

[0015] Further: In S31, regarding the distance parameter p The distance equation is:

[0016] in, For data samples With data samples The distance between For data samples PCI, For data samples PCI, For data samples RQI, For data samples RQI.

[0017] Further: S33 includes: S331, setting the basic parameters of the slime mold algorithm, initializing the population, and calculating the initial value of the fitness function; The basic parameters include population size, individual dimension, fitness function, maximum number of iterations, upper boundary of individual dimension, lower boundary of individual dimension and position update parameter; S332, calculate slime mold weight and first iteration parameters a and the second iteration parameters b ; S333, generate random numbers r 1. Determine the random number r Is 1 less than the position update parameter: If yes, the individual position is updated according to the position update parameter and the process proceeds to S336; If not, proceed to S334; S334, according to the first iteration parameter a and the second iteration parametersb Generate first iteration random parameters va and the second iteration random parameters vb , and enter S335; S335, generate random numbers r 2 and normalization parameters f , determine the random number r Is 2 less than the normalization parameter? f : If so, then according to the slime mold weight and the first iteration random parameter va Update individual position and enter S336; If not, then according to the second iteration random parameter vb Update individual position and enter S336; S336, update the fitness value and the global optimal solution, and determine whether the end condition is met: If satisfied, the global optimal solution is output and parameter optimization is completed; If not satisfied, return to S332.

[0018] Further: In S331, the fitness function fitn The expression is:

[0019] in, Indicates i The distance parameter of each individual, Indicates i The neighbor parameter of each individual K value, The actual test data set PSSI value, Represents the mean square error of the predicted value of the KNN model, g is a sample of the test data set, h is the total number of samples in the test data set.

[0020] Further: In S336, the end condition is to minimize the fitness function fitn or the maximum number of iterations is reached.

[0021] The beneficial effects of the present invention are: 1. Based on the actual situation that the pavement structure strength of ordinary national and provincial trunk roads at the road network level is difficult to collect data due to the long road network mileage and many characteristic factors, and the data regularity is poor, the K nearest neighbor (KNN) algorithm is used to achieve high-precision prediction of the pavement structure strength of ordinary national and provincial trunk roads at the road network level; 2. Use the slime mold algorithm (SMA) to optimize the distance parameter p and neighbor parameters K, which improves the prediction accuracy of the algorithm and solves the problem that the algorithm factor values ​​in the traditional KNN algorithm are artificially set or default values ​​are used, resulting in poor prediction accuracy of the algorithm; 3. The mutual information theory is used to calculate the feature weights of the two feature factors PCI and RQI, and the calculation method of the distance between data samples is updated, which characterizes the information dependence and sharing between the feature factors and improves the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of the road network-level highway pavement structure strength prediction method.

[0023] Figure 2 This is the fitness calculation curve of the Slime Mold Algorithm (SMA).

[0024] Figure 3 This is a scatter plot of the test set data. DETAILED DESCRIPTION

[0025] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0026] like Figure 1 As shown, in one embodiment of the present invention, a method for predicting the strength of a road pavement structure at a road network level is provided, comprising: S1. Collect and establish a road damage index PCI , Road Ride Quality Index RQI and pavement structure strength index PSSI A road network-level highway pavement technical condition detection dataset; S2. Calculate the detection data set PCI and RQI The feature weights of S3, based on PCI and RQI The feature weights of KNN The algorithm performs parameter optimization and constructs the final pavement structure strength prediction model; S4. Use the final pavement structure strength prediction model to predict the structural strength of the pavement to be predicted.

[0027] In this embodiment, the detection data set uses a 1000m road section length as the basic data unit. Based on the Highway Technical Condition Assessment Standard (JTG 5210-2018), the pavement damage condition index PCI and the pavement ride quality index RQI in the detection data set should cover five quality levels: excellent, good, medium, poor, and poor. That is, the value range of PCI and RQI is [0, 100]. According to the specification, the excellent value range is 90-100, the good value is 80-90, the medium value is 70-80, the poor value is 60-70, and the poor value is 0-60.

[0028] In this embodiment, a data set of technical condition detection of pavement of ordinary national and provincial trunk roads in a certain place accumulated between 2019 and 2023, totaling 19020 basic data units, is selected as the detection data set.

[0029] Specifically, S2 includes: S21. Calculate and detect data set PCI and PSSI The mutual information and RQI and PSSI The mutual information of S22, according to PCI and PSSI The mutual information and RQI and PSSI The mutual information of the detection data set is calculated PCI and RQI The feature weights of .

[0030] Mutual information is used to measure the information dependence between two random variables and describe the degree of information sharing between two random variables. From the perspective of information entropy, mutual information is equivalent to information gain, which can be expressed as information the degree to which the uncertainty of information is reduced; In S21, the detection data set is calculated PCI and PSSI The mutual information and RQI and PSSI The expression of mutual information is:

[0031]

[0032] in, express PCI and PSSI The mutual information of express RQI and PSSI The mutual information of To detect the data set PCI and PSSIThe joint probability distribution of To detect the data set RQI and PSSI The joint probability distribution of , and The detection data set is PCI , RQI and PSSI The marginal probability distribution of .

[0033] In S22, the detection data set is calculated PCI and RQI The expression of feature weight is:

[0034]

[0035] in, for PCI The feature weights of for RQI The feature weights of .

[0036] Specifically, S3 includes: S31. Construct the distance parameter between any two data samples based on the feature weights of PCI and RQI. p The distance equation is expressed as:

[0037] in, For data samples With data samples The distance between For data samples PCI, For data samples PCI, For data samples RQI, For data samples RQI; S32, using KNN algorithm, distance parameter based on distance equation p and neighbor parameters K value, construct a pavement structure strength prediction model; S33, using slime mold algorithm KNN The algorithm performs parameter optimization and obtains the final pavement structure strength prediction model.

[0038] In some embodiments, S33 includes: S331. Set the basic parameters of the slime mold algorithm and initialize the population , and calculate the initial fitness value ; The basic parameters of the slime mold algorithm include the population size pop , individual dimension dim , fitness function fitn , maximum number of iterations M， Upper boundary of individual dimension , the lower boundary of the individual dimension and position update parameters z ; Fitness function fitn The expression is:

[0039] in, Indicates i The distance parameter of each individual, Indicates i The neighbor parameter of each individual K value, The actual test data set PSSI value, Represents the mean square error of the predicted value of the KNN model, i ≤ pop represents a slime mold individual, j ≤ dim represents the individual dimension, g is a sample of the test data set, h is the total number of samples in the test data set; S332, iterate according to the slime mold algorithm, sort the individual fitness, and calculate the slime mold weight and the first iteration parameter a and the second iteration parameters b , whose expression is:

[0040]

[0041]

[0042] in, m is the number of iterations, For the m The individual i The weight of slime mold, represents individuals whose individual fitness is in the top 50%. Indicates individuals whose individual fitness is in the bottom 50%. BF(m) For the m The optimal fitness of the population at the iteration, WF(m) For the m The worst fitness of the population at the iteration, Fi(m)For the m The individual i The fitness value of is a random number vector with dimension [1, dim]; S333, generate random numbers r 1. Determine the random number r Is 1 less than the position update parameter: If yes, the individual position z is updated according to the position update parameter, and the process proceeds to S336; If not, proceed to S334; S334, according to the first iteration parameter a and the second iteration parameters b Generate first iteration random parameters va and the second iteration random parameters vb , and enter S335; Among them, the first iteration random parameter va and the second iteration random parameters vb They are the iteration parameters a and iteration parameters b About Vector A random function: , ; S335, generate random numbers r 2 and normalization parameters f , determine the random number r Is 2 less than the normalization parameter? f : in,

[0043] If so, then according to the slime mold weight and the first iteration random parameter va Update individual position and enter S336; At this time, the individual position update method is: ; If not, then according to the second iteration random parameter vb Update individual position and enter S336; At this time, the individual position update method is: ; Among them, the subscript A and B are two individuals selected at random, For the m The position of the individual with the best fitness at the iteration, is the updated individual position, is the individual position before updating, For the mThe individual A location, For the m The individual B location; S336, update the fitness value and the global optimal solution, and determine whether the end condition is met: The end condition is to minimize the fitness function fitn or the maximum number of iterations is reached M ; If satisfied, the global optimal solution is output and parameter optimization is completed; If not satisfied, return to S332.

[0044] In one embodiment of the present invention, an example test is performed using the road surface technical condition detection data of a portion of a road network-level highway shown in Table 1: Table 1 Technical condition test data of some road surfaces of a certain road network level highway

[0045] The test set results are as follows Figure 2 and Figure 3 As shown, =1, =5, =1, =100, M =100, z=0.03; the detection data set is divided into a training data set and a test data set in a ratio of 0.7:0.3. Figure 3 In the figure, the blue circle represents the test data set, the red circle represents the KNN prediction set, and the slime mold algorithm (SMA) is used to find the global optimal solution. , corresponding to the test data set .

[0046] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting the strength of a road pavement structure at a road network level, characterized in that: include: S1. Collect and establish a road damage index PCI , Road Ride Quality Index RQI and pavement structure strength index PSSI A road network-level highway pavement technical condition detection dataset; S2. Calculate the detection data set PCI and RQI The feature weights of S3, based on PCI and RQI The feature weights of KNN The algorithm performs parameter optimization and constructs the final pavement structure strength prediction model; S4. Use the final pavement structure strength prediction model to predict the structural strength of the pavement to be predicted.

2. The method for predicting the road pavement structure strength of a road network level highway according to claim 1, characterized in that S2 include: S21. Calculate and detect data set PCI and PSSI The mutual information and RQI and PSSI The mutual information of S22, according to PCI and PSSI The mutual information and RQI and PSSI The mutual information of the detection data set is calculated PCI and RQI The feature weights of .

3. The method for predicting the road pavement structure strength of a road network level highway according to claim 2, characterized in that: In S21, the detection data set is calculated PCI and PSSI The mutual information and RQI and PSSI The expression of mutual information is: in, express PCI and PSSI The mutual information of express RQI and PSSI The mutual information of To detect the data set PCI and PSSI The joint probability distribution of To detect the data set RQI and PSSI The joint probability distribution of , and The detection data set is PCI , RQI and PSSI The marginal probability distribution of .

4. The method for predicting the road pavement structure strength of a road network level highway according to claim 3, characterized in that: In S22, the detection data set is calculated PCI and RQI The expression of feature weight is: in, for PCI The feature weights of for RQI The feature weights of .

5. The method for predicting the road pavement structure strength of a road network level highway according to claim 1, characterized in that S3 include: S31. Construct the distance parameter between any two data samples based on the feature weights of PCI and RQI. p The distance equation of S32, using KNN algorithm, distance parameter based on distance equation p and neighbor parameters K value, construct a pavement structure strength prediction model; S33, using slime mold algorithm KNN The algorithm performs parameter optimization and obtains the final pavement structure strength prediction model.

6. The method for predicting the road pavement structure strength of a network-level highway according to claim 5, characterized in that: In S31, the distance parameter p The distance equation is: in, For data samples With data samples The distance between For data samples PCI, For data samples PCI, For data samples RQI, For data samples RQI.

7. The method for predicting the road pavement structure strength of a road network level highway according to claim 5, characterized in that: S33 includes: S331, setting the basic parameters of the slime mold algorithm, initializing the population, and calculating the initial value of the fitness function; The basic parameters include population size, individual dimension, fitness function, maximum number of iterations, upper boundary of individual dimension, lower boundary of individual dimension and position update parameter; S332, calculate slime mold weight and first iteration parameters a and the second iteration parameters b ; S333, generate random numbers r 1. Determine the random number r Is 1 less than the position update parameter: If yes, the individual position is updated according to the position update parameter and the process proceeds to S336; If not, proceed to S334; S334, according to the first iteration parameter a and the second iteration parameters b Generate first iteration random parameters va and the second iteration random parameters vb , and enter S335; S335, generate random numbers r 2 and normalization parameters f , determine the random number r Is 2 less than the normalization parameter? f : If so, then according to the slime mold weight and the first iteration random parameter va Update individual position and enter S336; If not, then according to the second iteration random parameter vb Update individual position and enter S336; S336, update the fitness value and the global optimal solution, and determine whether the end condition is met: If satisfied, the global optimal solution is output and parameter optimization is completed; If not satisfied, return to S332.

8. The method for predicting the road pavement structure strength of a road network level highway according to claim 7, characterized in that: In S331, the fitness function fitn The expression is: in, Indicates i The distance parameter of each individual, Indicates i The neighbor parameter of each individual K value, The actual test data set PSSI value, Represents the mean square error of the predicted value of the KNN model, g is a sample of the test data set, h is the total number of samples in the test data set.

9. The method for predicting the road pavement structure strength of a road network level highway according to claim 7, characterized in that: In S336, the end condition is to minimize the fitness function fitn or the maximum number of iterations is reached.

Images