Road pavement service performance decay estimation method and system and storage medium

Abstract

The invention discloses a highway pavement service performance decay estimation method. Typical highway pavement service performance historical decay data and corresponding influence characteristic data are acquired; the method comprises the following steps: constructing a typical road section database, preprocessing data, constructing a CART decision tree model, calculating principal component feature vectors, calculating similarity, and determining a decay model and parameters; and predicting decay data of the service performance of the road section. According to the method, PCA analysis and other data preprocessing technologies are adopted, the CART decision tree model is combined to calculate the importance weight of the principal component features and calculate the feature similarity degree between the road sections, and therefore the prediction model for service performance decay of the to-be-predicted road section and the calibration parameters of the prediction model are determined, and precise prediction of the highway pavement service performance is achieved. As time goes on, the sample size of the typical road section and the decay prediction model library are expanded and enriched, the prediction accuracy can be further improved, and the method provided by the invention has the characteristics of high prediction precision, expandability, self-adaption, automation and the like.

Description

Technical Field

[0001] The present invention relates to the field of highway maintenance, and in particular to a method, system and storage medium for predicting decay of highway pavement service performance. Background Art

[0002] The study of the decay prediction law of highway pavement service performance is crucial for maintenance planning and decision-making. It not only affects the judgment of maintenance timing, but also plays a decisive role in the quantitative analysis of maintenance benefits. An accurate prediction model can help managers take appropriate maintenance measures at the right time, thereby maximizing the performance and life of the pavement, saving maintenance costs, and improving the efficiency of capital use. Generally speaking, for sections with a long opening time and sufficient historical pavement inspection data, the form of the nonlinear fitting model can be determined and the decay parameters can be calculated.

[0003] However, for newly built highway projects, projects that lack historical decay data, or projects whose historical decay data are not referenceable, it is difficult to scientifically determine their pavement performance decay models and decay parameters, and thus it is impossible to accurately predict their performance decay trends, which affects the medium- and long-term maintenance decisions and planning of highway pavements. Summary of the invention

[0004] In view of the above problems, the present invention aims to provide a method, system and storage medium for predicting the service performance decay of highway pavement. By calculating the feature similarity between the predicted section and the typical section in the database, the performance decay prediction model and parameters of the newly built highway section or the highway section lacking historical data are determined.

[0005] To achieve the technical purpose, the solution of the present invention is: a method for predicting the decay of highway pavement service performance, the specific steps are as follows: S1. Obtain historical decay data of typical highway pavement service performance and corresponding impact characteristic data; S2. Build a typical road section database: including historical data on pavement service performance, influencing characteristic data, decay model, calibration parameters and its inverse life; S3, data preprocessing: Apply PCA principal component analysis method to reduce the dimension of database features, select k principal components, and project the original data of the database onto the selected principal components; S4, constructing a CART decision tree model: using a grid search algorithm to optimize the model parameters, and then calculating the importance coefficient of each principal component feature data; S5. Calculate the principal component eigenvector: weight the characteristic data of the road section to be predicted by using the importance coefficient to calculate the principal component eigenvector; S6. Calculate similarity: Calculate the spatial cosine similarity between the principal component feature vectors of the road section to be predicted and the typical road sections in the database, and sort them from large to small; S7, determining the decay model and parameters: selecting the typical road section with the highest similarity, assigning the service performance decay prediction model and calibrated parameters of the road section to the road section to be predicted; S8. Predict the decay data of the service performance of the road section; finally, formulate medium- and long-term maintenance decisions and plans based on the decay data.

[0006] Preferably, in step S1, the historical decay data of the highway pavement service performance includes one or more of the following: pavement technical condition index PQI, pavement damage condition index PCI, pavement ride quality index RQI, pavement rutting depth index RDI, and pavement skid resistance index SRI; The influencing characteristic data include one or more of the latest maintenance status of the corresponding road section unit, structure type, structural layer type, years of traffic, design deflection, number of asphalt layers, asphalt layer thickness, total pavement thickness, average daily traffic volume, number of categories three to six vehicles, and proportion of heavy vehicles.

[0007] Preferably, in step S2, the typical road section in the constructed database is a road section unit containing the road performance decay rules and characteristics under various characteristic combination conditions. Preferably, in step S3, the road section database features are subjected to principal component dimensionality reduction by a pre-set k value; in step S4, the model parameters are optimized by applying a grid search algorithm, a decision tree model is constructed and the importance coefficient of each principal component feature dimension is calculated; in step S5, the principal component features of the road section unit to be predicted are weighted using the principal component feature importance coefficient, and the principal component feature vector is calculated.

[0008] As a preferred method, the decay prediction model and its parameter library are constructed, which refers to a nonlinear model formula for fitting the pavement service performance. The specific formula is: ; The PPI in the formula is the pavement service performance index. is the initial value of the corresponding index, is the minimum value of the corresponding indicator, is the maximum value of the corresponding index, Indicates the year of pavement service performance decay, symbol are the calibration parameters of the corresponding model formulas; Calculate and store the corresponding typical road section performance indicators attenuated to the set threshold Inversion life , the formula is as follows: .

[0009] Preferably, in step S7, by calculating and sorting the weighted principal component feature similarities of the intended predicted section and each section in the typical section database by the importance coefficient, the service performance decay prediction model and calibration parameters of the typical section with the highest similarity are selected, and then the performance prediction model and decay parameters of the intended predicted section are determined, and the decay data of the service performance of the section in the corresponding year in the future is predicted.

[0010] Preferably, in step S6, the spatial cosine similarity between the principal component feature vector of the road section unit to be predicted and the principal component feature vector in the typical road section database is calculated and sorted from large to small; wherein the calculation formula of the spatial cosine similarity is as follows:

[0011] Vector in formula and They represent the principal component eigenvector of the road section unit to be predicted and the principal component eigenvector in the typical road section database, and are the L2 norms of the corresponding vectors respectively.

[0012] A prediction system, comprising: a typical road section database storage module: collecting, cleaning, and storing historical service performance index decay data and pavement impact characteristic data of typical highway sections, fitting historical decay data of different performance indicators using nonlinear fitting formulas, selecting typical sections with a goodness of fit greater than 0.9, and storing their decay models and corresponding fitting model parameters; Data preprocessing module: compile executable code programs to process missing values, encode, and center the feature data in the database of the road section to be predicted and the typical road section, and use PCA to reduce the dimension; Feature importance coefficient calculation module: build a CART decision tree model, calculate the importance coefficient of the feature after dimensionality reduction and assign it as a weight, and calculate the principal component eigenvector; Similarity calculation module: calculates the spatial cosine similarity between the feature vector of the road section to be predicted and each feature vector in the typical road section database, and calculates and returns the location number of the typical road section in the database with the highest feature similarity to the road section to be predicted; Pavement service performance prediction module: Using the typical road sections recommended by the similarity calculation module, its service performance index prediction model and calibration parameters are assigned to the road section unit to be predicted, and its performance index decay data in future years is predicted.

[0013] A computer-readable storage medium comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method for predicting the degradation of the service performance of a highway pavement is implemented.

[0014] The beneficial effects of the present invention are as follows: the present application uses historical data of highway pavement to construct a typical road section database, adopts data preprocessing techniques such as PCA analysis, and combines the CART decision tree model to calculate the importance weights of the principal component features, and calculates the degree of feature similarity between road sections, thereby determining the prediction model and calibration parameters for the service performance decay of the intended predicted road section, thereby achieving accurate prediction of the service performance of the highway pavement. With the passage of time, the expansion and enrichment of the sample size of typical road sections and the decay prediction model library, the accuracy of the prediction can be further improved. The method provided by the present application has the characteristics of high prediction accuracy, scalability, adaptability, and automation. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Flow chart of the maintenance decision-making method of the present invention.

[0016] Figure 2 It is a visualization diagram of a CART decision tree constructed with the principal component features after PCA dimensionality reduction and with the inversion life as the goal in an embodiment of the present invention.

[0017] Figure 3 It is a calculation result diagram of the highway pavement service performance decay prediction model and its parameter determination system for the preferred section in the embodiment of the present invention. DETAILED DESCRIPTION

[0018] The following is combined with Figure 1-3 The present invention is further described in detail with reference to the accompanying drawings and specific embodiments. Example

[0019] The specific embodiment of the present invention discloses a method for predicting the decay of the service performance of a highway pavement. The method for predicting the decay of the service performance of a highway pavement includes the following steps: S1, collecting historical decay data of the service performance of typical highway pavements and corresponding influencing characteristic data. The historical decay data of the service performance of a highway pavement includes at least one of the following pavement performance indicators PPI, a pavement technical condition index PQI, a pavement damage condition index PCI, a pavement ride quality index RQI, a pavement rutting depth index RDI, and a pavement skid resistance index SRI.

[0020] Table 1 shows the PCI historical decay data of some sections in the typical road section database.

[0021] Project Code Lane Starting pile number End station Year 1 PCI Year 2 PCI ... Year 10 PCI JZ B 262600 262900 100.00 98.80 86.31 GS A 2012800 2014000 100.00 99.97 80.11 GS E 1988600 1991300 100.00 100.00 57.78 GS F 1983400 1985200 100.00 99.98 86.52 TS A 96600 97200 100.00 99.74 70.20 TS A 113600 114000 100.00 99.79 87.88 TS A 131700 132300 100.00 100.00 77.57 TS A 59100 63600 100.00 100.00 75.79 TS A 68200 69700 100.00 99.98 80.20 TS B 65000 67000 100.00 100.00 66.35 TS B 133500 134400 100.00 100.00 87.61 SG A 510600 607300 100.00 99.30 87.25 SG B 595100 595900 100.00 99.75 82.93 YM B 3290000 3295000 100.00 99.92 89.48 YM B 3312035 3315000 100.00 99.64 85.23 YM B 3352000 3357000 100.00 99.99 84.75 GZ B 17100 19000 100.00 99.99 74.91 GZ A 17900 18600 100.00 100.00 80.56 GZ B 7700 8500 100.00 100.00 85.52 GS F 2135400 2136200 100.00 100.00 80.44 GS F 2136200 2137000 100.00 99.54 85.85 GS F 2137000 2137600 100.00 100.00 77.51 The corresponding influencing characteristic data include at least one of the following: the latest maintenance status of the road section unit, the type of structure, the type of structural layer, the years of traffic, the designed deflection, the number of asphalt layers, the thickness of the asphalt layer, the total thickness of the road surface, the average daily traffic volume, the number of vehicles of categories 3 to 6, and the proportion of heavy vehicles. It should be understood that the historical decay data of the road service performance and the corresponding influencing characteristic data are not limited to this, and may also include other indicators or parameters.

[0022] Table 2 shows the characteristic data corresponding to some sections in the typical road section database collected and constructed.

[0023] Structures Structure layer type Open to traffic Total pavement thickness Asphalt layer thickness Asphalt layers ... Average daily traffic volume Heavy vehicle ratio (%) Roadbed Semi-rigid base 8 88 17 3 7483 18.8 Roadbed Semi-rigid base 8 88 17 3 7483 18.8 Roadbed Semi-rigid base 8 88 17 3 13309 15.2 Roadbed Semi-rigid base 8 88 17 3 11362 14.7 Roadbed Roller compacted concrete base 10 72 12 2 11032 21.7 Roadbed Roller compacted concrete base 10 72 12 2 10302 17.6 Roadbed Semi-rigid base 8 88 17 3 11781 14.6 Roadbed Semi-rigid base 8 88 17 3 11781 14.6 Roadbed Semi-rigid base 8 88 17 3 13120 15.5 Roadbed Combined base 11 76 18 3 7099 20.4 Roadbed Combined base 11 76 18 3 7099 20.4 Roadbed Semi-rigid base 19 76 20 4 15721 18.9 Roadbed Semi-rigid base 19 76 20 4 15719 19.1 Roadbed Semi-rigid base 8 88 17 3 7483 18.8 Roadbed Semi-rigid base 8 88 17 3 12931 15.7 Roadbed Semi-rigid base 8 88 17 3 12931 15.7 Roadbed Semi-rigid base 8 88 17 3 12931 15.7 bridge Bridge Deck 11 10 10 2 11210 19.2 bridge Bridge Deck 11 10 10 2 7099 20.4 bridge Bridge Deck 11 10 10 2 7702 19.4 Roadbed Semi-rigid base 8 88 17 3 9651 16.6 Roadbed Semi-rigid base 8 88 17 3 11781 14.6 Step S2: construct a typical road section database, including historical data of pavement service performance, influencing characteristic data, decay model, calibration parameters and their inverse life; The pavement service performance decay model includes one or more of the linear or nonlinear fitting model formulas. The following is a preferred embodiment, and its calibration parameters are and , pre-set indicator threshold is 85, and the inversion life of each typical road section is calculated according to the corresponding inversion life model formula It should be understood that the pavement service performance decay model and index threshold are not limited to this, and may also include other linear or nonlinear model formulas, and other index thresholds.

[0024] ; Step S3: Data preprocessing, applying the PCA principal component analysis method to reduce the dimension of database features, selecting k principal components, and projecting the original data onto the selected principal components.

[0025] In the embodiment of the present application, the k value is pre-set to 5, and the feature data in the database of the proposed predicted road section and the typical road section are processed for missing values, discrete variable encoding, and data centralization, and then PCA is used for dimensionality reduction; it should be understood that the principal component k value is not limited to this, and other parameters may also be included.

[0026] Calculate the principal component feature matrix As follows, is the number of typical road sections in the database.

[0027] ; Step S4: Use the principal component features after PCA dimension reduction and invert the lifespan The CART decision tree constructed for the target is used, and the grid search algorithm is used to optimize the model parameters, and then the importance coefficient of each principal component feature data is calculated. Figure 2 , which is a visualization diagram of the decision tree model constructed in the embodiment of the present application.

[0028] Calculate the principal component feature importance weight coefficient vector as follows: ; Step S5: Calculate the principal component eigenvector: Use the importance coefficient to weight the characteristic data of the road section to be predicted, calculate the principal component eigenvector, and calculate the weighted principal component characteristic matrix as follows: ; Step S6: Calculate the spatial cosine similarity between the principal component feature vectors of the road section to be predicted and the typical road sections in the database, and sort them from large to small.

[0029] Among them, the principal component eigenvector of the road section unit to be predicted is calculated and the first The principal component eigenvector of the road segment The calculation formula is: ; ; Therefore, the road section in the typical road section database that is most similar to the principal component characteristics of the road section unit to be predicted The calculation formula is as follows, where and are the L2 norms of the corresponding vectors respectively.

[0030]

[0031] Step S7: Select the typical road section with the highest similarity based on the above calculation , according to its decay fitting model and parameters, the service performance decay prediction model and calibrated parameters of the section are assigned to the section to be predicted.

[0032] Step S8: predicting the decay data of the road surface service performance index of the road section in future years.

[0033] Embodiment 2 In step S2 of the method of the present application, a decay prediction model and its parameter library are constructed to fit the linear or nonlinear model formula of the pavement service performance, which can also be the following formula: ; The PPI in the formula is the pavement service performance index, is the initial value of the corresponding index, is the minimum value of the corresponding indicator, is the maximum value of the corresponding index, Indicates the year of pavement service performance decay, symbol are the parameters of the corresponding model formulas.

[0034] Calculate and store the corresponding typical road section performance indicators attenuated to the set threshold The inversion lifespan is as follows: .

[0035] Embodiment 3 In step S2 of the method of the present application, a decay prediction model and its parameter library are constructed to fit the linear or nonlinear model formula of the pavement service performance, which can also be the following formula: ; The PPI in the formula is the pavement service performance index, is the initial value of the corresponding index, is the minimum value of the corresponding indicator, is the maximum value of the corresponding index, Indicates the year of pavement service performance decay, symbol are the parameters of the corresponding model formulas.

[0036] Calculate and store the corresponding typical road section performance indicators attenuated to the set threshold The inversion lifespan is as follows: .

[0037] A prediction system includes a typical road section database storage module, a data preprocessing module, a feature importance coefficient calculation module, a similarity calculation module and a road service performance prediction module. Figure 3 As shown, it is a graphical interface of the calculation results of the highway pavement service performance decay prediction model and its parameter determination system for the preferred section in the embodiment of the present application.

[0038] The typical road section database storage module collects, stores and manages the historical service performance index decay data and road surface impact characteristic data of typical road sections, uses a variety of nonlinear fitting formulas to fit the historical decay data of different performance indicators, selects typical road sections with a goodness of fit greater than 0.9, stores their decay models and corresponding fitting model parameters, and uses pre-set indicator thresholds to calculate the decay model parameters. , calculate and save its corresponding inversion life .

[0039] The data preprocessing module performs missing value processing, discrete variable encoding processing, and data centering processing on the feature data in the database of the predicted section and the typical section, and then uses PCA to perform dimensionality reduction processing.

[0040] The feature importance coefficient calculation module uses the principal component feature data after dimensionality reduction to construct the target inversion life The CART decision tree model is used to calculate the importance coefficient of the principal component features after dimensionality reduction and assign them as weights to obtain the principal component feature vector.

[0041] The similarity calculation module calculates the spatial cosine similarity between the principal component feature vector of the intended predicted section and each feature vector in the typical section database, and calculates and returns the location number of the typical section in the database with the highest feature similarity to the intended predicted section.

[0042] The pavement service performance prediction module assigns the service performance index prediction model and calibration parameters to the predicted road section unit based on the typical road sections recommended by the similarity calculation module, and predicts the performance index decay data in future years.

[0043] The embodiment of the present application also provides a computer storage medium on which an executable computer program is stored. When the computer program is executed by a processor, the highway pavement service performance decay prediction method and computing system described in the above embodiment are implemented.

[0044] In summary, the embodiments of the present application use the historical decay data of highway pavement performance to build a typical road section database, use data preprocessing techniques such as PCA analysis, and combine the CART decision tree model to calculate the importance weight coefficient of the principal component features, and calculate the feature similarity between road sections, so as to determine the prediction model and its calibration parameters for the service performance decay of the road section to be predicted, and then achieve accurate prediction of the service performance of the highway pavement. With the passage of time, the expansion and enrichment of the sample size of typical road sections and the prediction model library, the accuracy of the prediction can be further improved. The method, system and storage medium provided by the present application have the characteristics of high prediction accuracy, scalability, adaptability, and automation.

[0045] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any slight modification, equivalent substitution and improvement made to the above embodiment based on the technical essence of the present invention should be included in the protection scope of the technical solution of the present invention.

Claims

1. A method for predicting the degradation of highway pavement service performance, characterized in that: The specific steps are as follows: S1. Obtain historical decay data of typical highway pavement service performance and corresponding impact characteristic data; S2. Build a typical road section database: including historical data on pavement service performance, influencing characteristic data, decay model, calibration parameters and its inverse life; S3, data preprocessing: Apply PCA principal component analysis method to reduce the dimension of database features, select k principal components, and project the original data of the database onto the selected principal components; S4, constructing a CART decision tree model: using a grid search algorithm to optimize the model parameters, and then calculating the importance coefficient of each principal component feature data; S5, calculating the principal component eigenvector: using the importance coefficient to weight the characteristic data of the road section to be predicted, and calculating the principal component eigenvector; S6. Calculate similarity: Calculate the spatial cosine similarity between the principal component feature vectors of the road section to be predicted and the typical road sections in the database, and sort them from large to small; S7, determining the decay model and parameters: selecting the typical road section with the highest similarity, assigning the service performance decay prediction model and calibrated parameters of the road section to the road section to be predicted; S8. Predict the decay data of the service performance of the road section; finally, formulate medium- and long-term maintenance decisions and plans based on the decay data.

2. The method for predicting the degradation of road pavement service performance according to claim 1 is characterized in that: In the step S1, the historical decay data of the highway pavement service performance includes one or more of the pavement technical condition index PQI, the pavement damage condition index PCI, the pavement ride quality index RQI, the pavement rutting depth index RDI, and the pavement skid resistance index SRI; The influencing characteristic data include one or more of the latest maintenance status of the corresponding road section unit, structure type, structural layer type, years of traffic, design deflection, number of asphalt layers, asphalt layer thickness, total pavement thickness, average daily traffic volume, number of categories three to six vehicles, and proportion of heavy vehicles.

3. The method for predicting the degradation of road pavement service performance according to claim 2 is characterized in that: In step S2, the typical road section in the constructed database is a road section unit containing the decay rules and characteristics of road performance under various characteristic combination conditions.

4. The method for predicting the degradation of road pavement service performance according to claim 3 is characterized in that: In the step S3, the road section database features are subjected to principal component dimensionality reduction by using a preset k value; In step S4, the model parameters are optimized by applying a grid search algorithm, a decision tree model is constructed, and the importance coefficient of each principal component feature dimension is calculated; In step S5, the main component features of the road section unit to be predicted are weighted using the main component feature importance coefficients to calculate and obtain the main component feature vector.

5. The method for predicting the degradation of road pavement service performance according to claim 3 is characterized in that: Constructing a decay prediction model and its parameter library refers to a nonlinear model formula used to fit the pavement service performance. The specific formula is: ; The PPI in the formula is the pavement service performance index. is the initial value of the corresponding index, is the minimum value of the corresponding indicator, is the maximum value of the corresponding index, Indicates the year of pavement service performance decay, symbol are the calibration parameters of the corresponding model formulas; Calculate and store the corresponding typical road section performance indicators attenuated to the set threshold Inversion life , the formula is as follows: .

6. The method for predicting the degradation of road pavement service performance according to claim 3 is characterized in that: In step S7, by calculating and sorting the weighted principal component feature similarities of the intended predicted section and each section in the typical section database by the importance coefficient, the service performance decay prediction model and calibration parameters of the typical section with the highest similarity are selected, and then the performance prediction model and decay parameters of the intended predicted section are determined, and the decay data of the service performance of the section in the corresponding year in the future is predicted.

7. The method for predicting the degradation of road pavement service performance according to claim 3 is characterized in that: In step S6, the spatial cosine similarity between the principal component feature vector of the road section unit to be predicted and the principal component feature vector in the typical road section database is calculated and sorted from large to small; wherein the calculation formula of the spatial cosine similarity is as follows: Vector in formula and They represent the principal component eigenvector of the road section unit to be predicted and the principal component eigenvector in the typical road section database, and are the L2 norms of the corresponding vectors respectively.

8. An estimation system, characterized in that: include: Typical road section database storage module: collects, cleans and stores the historical service performance index decay data and pavement impact characteristic data of typical highway sections, uses nonlinear fitting formula to fit the historical decay data of different performance indicators, selects typical sections with goodness of fit greater than 0.9, and stores their decay models and corresponding fitting model parameters; Data preprocessing module: compile executable code programs to process missing values, encode, and center the feature data in the database of the road section to be predicted and the typical road section, and use PCA to reduce the dimension; Feature importance coefficient calculation module: build a CART decision tree model, calculate the importance coefficient of the feature after dimensionality reduction and assign it as a weight, and calculate the principal component eigenvector; Similarity calculation module: calculates the spatial cosine similarity between the feature vector of the road section to be predicted and each feature vector in the typical road section database, and calculates and returns the location number of the typical road section in the database with the highest feature similarity to the road section to be predicted; Pavement service performance prediction module: Using the typical road sections recommended by the similarity calculation module, its service performance index prediction model and calibration parameters are assigned to the road section unit to be predicted, and its performance index decay data in future years is predicted.

9. A computer-readable storage medium, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for predicting the degradation of the service performance of a highway pavement as claimed in any one of claims 1 to 7 is implemented.

Images