Asphalt pavement performance prediction model adaptive regression method and system

Abstract

The invention discloses an asphalt pavement performance prediction model adaptive regression method and system, and relates to the technical field of pavement maintenance management, and the method comprises the steps: classifying pavement performance detection data; determining a basic form of a prediction model of the pavement performance indexes; performing parameter value interval setting on the basic form of the prediction model; continuously decayed detection data slices are obtained, and the relative road age is calculated according to the prediction model form; and for a plurality of groups of prediction equation forms containing different parameter combinations, grid search is carried out, an optimal regression parameter is obtained by using a least square method, and prediction model adaptive regression is carried out. According to the method, the complexity of traditional road age correction is simplified, and automatic calculation is realized in the whole regression process, so that the calculation efficiency is improved; the prediction effect is accurate, five common pavement performance indexes are predicted by adopting different regression model forms, and the change trend of the pavement performance can be accurately reflected; the method is suitable for road sections with different traffic loads and different pavement structures.

Description

Technical Field

[0001] The present invention relates to the technical field of pavement maintenance management, and more particularly to an adaptive regression method and system for predicting asphalt pavement performance. Background Art

[0002] At present, in the highway traffic network, understanding the decay characteristics of pavement performance, establishing a scientific pavement performance prediction model, and predicting the pavement performance indicators of roads in the future operation period are the prerequisites for determining maintenance strategies.

[0003] The concept of "pavement performance prediction" was first proposed in the American Association of State Highway and Transportation Official (AASHTO) test road research in the 1960s, and developed with the development of the pavement management system (PMS). According to the form of the pavement performance prediction model, it can be summarized into three categories: deterministic model, non-deterministic model and combined prediction model. The deterministic model is an explicit functional relationship between the pavement performance index and a series of external factors such as road structure, road age, traffic load, etc., and its prediction result is certain; the non-deterministic model has no explicit functional form, and most of them are probability distributions of pavement performance indicators; because the above two models have their own advantages and disadvantages, some scholars combine several of them to reduce the influence of some environmental factors in the prediction process of a single model. In the actual application scenarios of pavement maintenance countermeasures, deterministic models are often used by maintenance managers.

[0004] The empirical regression models applicable to the pavement technical condition indicators mentioned in the "Technical Specifications for Highway Maintenance Decision-making" (Draft for Comments) include: (1) Linear regression model: mainly used for the pavement from the initial use, which is affected by factors such as design, construction quality, and maintenance conditions; (2) Exponential regression model: used for the prediction of indicators when the pavement technical condition indicators decay faster and faster with the increase of road age on unmaintained sections; (3) Improved S-shaped regression model: used when the pavement technical condition index decays to a certain extent, and the pavement technical condition indicators are improved due to maintenance measures, and the pavement technical condition decay rate decreases; (4) Dual parameter curve regression model: comprehensively considers factors such as environment, traffic, and materials, and can be used to simulate the decay of asphalt pavement technical conditions. Among them, only the improved S-shaped regression model takes into account the factors of maintenance measures. The other three recommended prediction model forms are all related to the road age t, and are deterministic models with the initial pavement performance index PI0 and road age t as variables.

[0005] Road age generally refers to the cumulative time from the time a road is built and put into use to the time of evaluation, reflecting the number of years the road has been in service. If the road surface has undergone major or medium-sized repair and maintenance projects, such as re-laying the base or large-scale structural reconstruction, its "road age" in the technical sense needs to be recalculated from the completion of the reconstruction. In this case, road age is divided into two concepts: cumulative road age (total time since construction) and effective road age (time from the most recent restorative maintenance to the present). In many existing pavement performance prediction model studies, maintenance measures are taken as one of the influencing factors, and their role is reflected by effective road age. This process often involves the data preprocessing of sorting out historical data of maintenance projects, matching detection units with maintenance projects, and correcting cumulative road age to effective road age. The process is cumbersome and the standards are inconsistent. At the same time, the lack of historical data of maintenance projects due to untimely communication and improper storage is likely to lead to inaccurate road age correction, affecting the accuracy of model prediction.

[0006] Therefore, how to propose an adaptive regression method and system for asphalt pavement performance prediction model, oriented to asphalt pavement performance indicators, including pavement condition index (PCI), pavement ride quality index (RQI), pavement rutting depth index (RDI), pavement skid resistance index (SRI) and pavement structure strength index (PSSI), to provide technical support for scientific decision-making in highway maintenance is an urgent problem that technicians in this field need to solve. Summary of the invention

[0007] In view of this, the present invention provides an adaptive regression method and system for predicting asphalt pavement performance, which is oriented to asphalt pavement performance indicators, including pavement damage index (PCI), pavement ride quality index (RQI), pavement rutting depth index (RDI), pavement skid resistance index (SRI) and pavement structure strength index (PSSI), to provide technical support for scientific decision-making in highway maintenance. In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0008] An adaptive regression method for predicting asphalt pavement performance model, comprising:

[0009] Classify the pavement performance test data;

[0010] Determine the basic form of the prediction model for pavement performance indicators;

[0011] Setting parameter value intervals for the basic form of the prediction model;

[0012] Obtain continuous decay detection data slices and calculate relative road age based on the prediction model form;

[0013] For several groups of prediction equations containing different parameter combinations, grid search is performed, and the optimal regression parameters are obtained using the least squares method to perform adaptive regression of the prediction model.

[0014] Optionally, the classification of the pavement performance test data includes: taking traffic load and pavement structure composition as main quantitative indicators.

[0015] Optionally, the basic form of the prediction model for determining the pavement performance index includes: a dual-parameter curve regression model is used for the pavement damage condition index and the pavement ride quality index.

[0016] Optionally, the two-parameter curve regression model is:

[0017] Where PI is the pavement performance index, PI0 is the initial value of the pavement performance index, t is the road age, and α and β are model parameters.

[0018] Optionally, the basic form of the prediction model for determining the pavement performance index also includes: a linear regression model is selected for the pavement rutting depth index, the pavement anti-skid performance index and the pavement structure strength index.

[0019] Optionally, the linear regression model is: PI=PI0-αt, where PI is a pavement performance index, PI0 is an initial value of the pavement performance index, t is a road age, and α is a model parameter.

[0020] Optionally, the parameter value interval setting for the basic form of the prediction model includes: for a two-parameter curve regression model, the α parameter interval is set to [12, 25], with a step size of 0.5; the β parameter interval is set to [0.3, 2], with a step size of 0.1; for a linear regression model, the α parameter interval is set to [0.1, 5], with a step size of 0.05, and according to the parameter settings, several groups of prediction equation forms containing different parameter combinations are generated.

[0021] Optionally, the acquisition of the continuous decay detection data slice and the calculation of the relative road age according to the prediction model form include: assuming that the performance index value of the initial year in the continuous decay detection data slice is in the prediction equation form, the relative road age t of the initial year is inferred according to the determined prediction equation form and the performance index value, then the relative road age value of the second year is t+1, the subsequent points of the continuous decay slice are accumulated year by year on the basis of the relative road age of the initial year, and the performance index value of the corresponding year is calculated according to the prediction model form, and for several groups of prediction equation forms containing different parameter combinations, a grid search is performed, and the least squares method is used to determine the coefficient of determination R 2 The maximum is taken as the principle to obtain the optimal regression parameters.

[0022] Optionally, model validation is also included, and the root mean square error is selected as the evaluation indicator to quantify the difference between the predicted value and the actual value. The smaller the RMSE, the better the prediction performance of the model. The formula is as follows:

[0023] Where n is the number of samples; y i is the measured value; is the predicted value.

[0024] Optionally, an adaptive regression system for predicting asphalt pavement performance model comprises:

[0025] Classification module: used to classify road performance test data;

[0026] Determination module: used to determine the basic form of the prediction model of pavement performance indicators;

[0027] Parameter value setting module: used to set parameter value intervals for the basic form of the prediction model;

[0028] Calculation module: used to obtain the detection data slices of continuous decay and calculate the relative road age according to the prediction model form;

[0029] Adaptive regression module: It is used to perform grid search for several groups of prediction equations containing different parameter combinations, use the least squares method to obtain the optimal regression parameters, and perform adaptive regression of the prediction model.

[0030] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses an adaptive regression method and system for predicting asphalt pavement performance, which has the following beneficial effects:

[0031] The present invention proposes an adaptive regression method for predicting asphalt pavement performance, comprising: classifying pavement performance test data; determining the basic form of the predicting model of pavement performance indicators; setting the parameter value interval for the basic form of the predicting model; obtaining a slice of the continuously decaying test data, and calculating the relative road age according to the predicting model form; for a plurality of groups of predicting equation forms containing different parameter combinations, performing a grid search, using the least square method to obtain the optimal regression parameters, and performing adaptive regression of the predicting model. The present invention realizes the adaptive regression process of the predicting model of asphalt pavement performance, (1) high computational efficiency, the modeling method based on relative road age effectively simplifies the complexity of traditional road age correction, and the whole regression process realizes computer automatic calculation, thereby improving computational efficiency. (2) accurate prediction effect, using different regression model forms to predict five commonly used pavement performance indicators respectively, the results show that the model root mean square error (RMSE) is between 0 and 2, the prediction accuracy is high, and the model can accurately reflect the changing trend of pavement performance, meeting the actual engineering application requirements. (3) wide application range, suitable for sections with different traffic loads and different pavement structures, and the examples cover 22 sections of national and provincial roads in Shanghai. In the future, it can be used for highway research in more provinces and cities, and will be suitable for a wide range of engineering scenarios and management needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0033] Figure 1 A schematic flow chart of an adaptive regression method for predicting asphalt pavement performance provided by the present invention.

[0034] Figure 2 This is a structural framework diagram of an adaptive regression system for an asphalt pavement performance prediction model provided by the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] The embodiment of the present invention discloses an adaptive regression method for predicting asphalt pavement performance. Figure 1 As shown, including:

[0037] Classify the pavement performance test data;

[0038] Determine the basic form of the prediction model for pavement performance indicators;

[0039] Setting parameter value intervals for the basic form of the prediction model;

[0040] Obtain continuous decay detection data slices and calculate relative road age based on the prediction model form;

[0041] For several groups of prediction equations containing different parameter combinations, grid search is performed, and the optimal regression parameters are obtained using the least squares method to perform adaptive regression of the prediction model.

[0042] Furthermore, the classification of the pavement performance test data includes: taking traffic load and pavement structure composition as main quantitative indicators.

[0043] Furthermore, the basic form of the prediction model for determining the pavement performance index includes: a double-parameter curve regression model is selected for the pavement damage condition index and the pavement ride quality index.

[0044] Furthermore, the two-parameter curve regression model is:

[0045] Where PI is the pavement performance index, PI0 is the initial value of the pavement performance index, t is the road age, and α and β are model parameters.

[0046] Furthermore, the basic form of the prediction model for determining the pavement performance index also includes: a linear regression model is selected for the pavement rutting depth index, the pavement anti-skid performance index and the pavement structure strength index.

[0047] Furthermore, the linear regression model is: PI=PI0-αt, where PI is the pavement performance index, PI0 is the initial value of the pavement performance index, t is the road age, and α is the model parameter.

[0048] Furthermore, the parameter value interval setting for the basic form of the prediction model includes: for the dual-parameter curve regression model, the α parameter interval is set to [12, 25], with a step size of 0.5; the β parameter interval is set to [0.3, 2], with a step size of 0.01; for the linear regression model, the α parameter interval is set to [0.1, 5], with a step size of 0.05. According to the parameter settings, several groups of prediction equation forms containing different parameter combinations are generated.

[0049] Furthermore, the acquisition of the continuous decay detection data slice and the calculation of the relative road age according to the prediction model form include: assuming that the performance index value of the initial year in the continuous decay detection data slice is in the prediction equation form, the relative road age t of the initial year is inferred according to the determined prediction equation form and the performance index value, then the relative road age value of the second year is t+1, the subsequent points of the continuous decay slice are accumulated year by year on the basis of the relative road age of the initial year, and the performance index value of the corresponding year is calculated according to the prediction model form, and for several groups of prediction equation forms containing different parameter combinations, a grid search is performed, and the least squares method is used to determine the coefficient of determination R 2 The maximum is taken as the principle to obtain the optimal regression parameters.

[0050] Furthermore, the model is validated and the root mean square error is selected as the evaluation indicator to quantify the difference between the predicted value and the actual value. The smaller the RMSE, the better the prediction performance of the model. The formula is as follows:

[0051] Where n is the number of samples; y i is the measured value; is the predicted value.

[0052] In a specific embodiment, an adaptive regression system for predicting asphalt pavement performance is provided. Figure 2 As shown, including:

[0053] Classification module: used to classify road performance test data;

[0054] Determination module: used to determine the basic form of the prediction model of pavement performance indicators;

[0055] Parameter value setting module: used to set parameter value intervals for the basic form of the prediction model;

[0056] Calculation module: used to obtain the detection data slices of continuous decay and calculate the relative road age according to the prediction model form;

[0057] Adaptive regression module: It is used to perform grid search for several groups of prediction equations containing different parameter combinations, use the least squares method to obtain the optimal regression parameters, and perform adaptive regression of the prediction model.

[0058] In a specific implementation, the present invention proposes an adaptive regression method for predicting asphalt pavement performance, and constructs a relative road age modeling method, which avoids the inaccuracy that may occur in the traditional road age correction process, thereby effectively reducing the risk of modeling errors. This method makes the prediction model more accurate by considering the relative change of road age, and can better reflect the actual road performance changes.

[0059] First, the premise assumptions required for the adaptive regression method of asphalt pavement performance prediction model implemented in the present invention are:

[0060] (1) The increase in the pavement index (PI) means that maintenance works have been carried out;

[0061] (2) PI decays naturally over successive years;

[0062] (3) When PI is 100, the road age is 0;

[0063] (4) The decay of PI is only related to the initial PI, pavement structure, and the average daily equivalent single axle load (ESAL).

[0064] Specifically, the prediction model adaptive regression method includes the following steps:

[0065] (1) Classify the pavement performance test data, using traffic load and pavement structure composition (surface thickness) as the main quantitative indicators. The classification standards for the predicted model surface thickness and predicted model traffic load are shown in Tables 1 and 2.

[0066] Table 1 Classification of surface thickness of prediction model (cm)

[0067] Serial number Surface thickness classification illustrate 1 Thin 0<Thickness<=12 2 middle 12<Thickness<=18 3 thick Thickness>18

[0068] Table 2 Traffic load classification of the prediction model (times / days / lanes)

[0069] Serial number Traffic load classification illustrate 1 Extra light 0<ESAL<=500 2 light 500<ESAL<=1500 3 medium 1500<ESAL<=2500 4 Heavy 2500<ESAL<=4000 5 overweight 4000<ESAL<=6000 6 Extra heavy ESAL>6000

[0070] (2) Determine the basic form of the prediction model.

[0071] The pavement surface condition index (PCI) and the pavement riding quality index (RQI) use a dual-parameter curve regression model, as shown in formula (1);

[0072] The pavement rutting depth index (RDI), pavement skidding resistance index (SRI) and pavement structure strength index (PSSI) use a linear regression model, as shown in formula (2).

[0073]

[0074] Where PI is the pavement performance index, PI0 is the initial value of the pavement performance index, t is the road age, α and β are model parameters, the same below.

[0075] PI=PI0-αt(2).

[0076] (3) Set parameter values ​​within a reasonable range.

[0077] For the two-parameter curvilinear regression model, the α parameter interval is set to [12, 25], the step size is 0.5, and the β parameter interval is set to [0.3, 2], the step size is 0.1; for the linear regression model, the α parameter interval is set to [0.1, 5], the step size is 0.05. According to the parameter settings, several groups of prediction equations containing different parameter combinations are generated.

[0078] (4) Obtain the detection unit slices of continuous decay and calculate the relative road age.

[0079] Assuming that the performance index value of the initial year in the continuous decay detection unit slice is on the prediction equation, the relative road age t of the initial year is inferred based on the determined prediction equation form and the performance index value. Then the relative road age value of the second year is t+1. The subsequent points of the continuous decay detection unit slice are accumulated year by year based on the road age of the initial year, and the performance index value of the corresponding year is calculated according to the prediction model form.

[0080] (5) For several sets of prediction equations containing different parameter combinations, a grid search is performed using the least squares method with the coefficient of determination R 2 Based on the maximum principle, the optimal regression parameters are obtained to realize the adaptive regression of the prediction model.

[0081] In a specific embodiment, the method is described by taking the national and provincial trunk roads in Shanghai as an example:

[0082] (1) Data classification:

[0083] According to the classification standards of asphalt surface thickness and standard axle load times ESAL, the existing national and provincial trunk line facilities in Shanghai can be classified into the following combinations, as shown in Table 3 and Table 4:

[0084] Table 3 Combination types of national trunk roads in Shanghai

[0085] Serial number Surface thickness classification Traffic load classification 1 middle Extra heavy 2 middle overweight 3 thick Extra heavy 4 Thin Extra heavy 5 thick Heavy 6 middle Heavy 7 middle medium 8 Thin Heavy

[0086] Table 4 Combination types of provincial trunk roads in Shanghai

[0087]

[0088]

[0089] (2) Modeling results

[0090] The road performance test data of Shanghai from 2016 to 2021 were used to adaptively model the prediction model for the above 22 groups of road sections. The highway prediction models suitable for different road performance technical condition indicators are shown in Tables 5, 6, 7, and 8:

[0091] Table 5 Results of the prediction model for national trunk roads (dual parameters)

[0092]

[0093] Table 6 Results of the prediction model for national trunk roads (single parameter)

[0094]

[0095] Table 7 Results of provincial trunk road prediction model (dual parameters)

[0096]

[0097] Table 8 Results of provincial trunk road prediction model (single parameter)

[0098]

[0099]

[0100] (3) Model verification

[0101] The prediction effect of the model is verified using the Shanghai pavement performance test data from 2022 to 2023. The root mean square error (RMSE) is selected as the evaluation indicator to quantify the difference between the predicted value and the actual value. The smaller the RMSE, the better the prediction performance of the model. The calculation formula is as follows:

[0102]

[0103] Where n is the number of samples; y i is the measured value; is the predicted value.

[0104] The prediction model validation results are shown in Table 9:

[0105] Table 9 Prediction model validation results

[0106]

[0107] According to the above results, the prediction model verification of national and provincial highways shows that the root mean square error (RMSE) is in the range of 0 to 2. This result fully demonstrates that the prediction accuracy of the model meets the needs of actual engineering and shows good prediction performance. It can be seen that the prediction model can not only provide effective support for the current pavement performance evaluation, but also lay a solid foundation for future pavement maintenance and management decisions.

[0108] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0109] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An adaptive regression method for predicting asphalt pavement performance, characterized in that: include: Classify the pavement performance test data; Determine the basic form of the prediction model for pavement performance indicators; Setting parameter value intervals for the basic form of the prediction model; Obtain continuous decay detection data slices and calculate relative road age based on the prediction model form; For several groups of prediction equations containing different parameter combinations, grid search is performed, and the optimal regression parameters are obtained using the least squares method to perform adaptive regression of the prediction model.

2. The adaptive regression method for predicting asphalt pavement performance according to claim 1 is characterized in that: The classification of the pavement performance test data includes: taking traffic load and pavement structure composition as main quantitative indicators.

3. The adaptive regression method for predicting asphalt pavement performance according to claim 1 is characterized in that: The basic form of the prediction model for determining the pavement performance index includes: a double-parameter curve regression model is selected for the pavement damage condition index and the pavement ride quality index.

4. The adaptive regression method for predicting asphalt pavement performance according to claim 3 is characterized in that: The two-parameter curvilinear regression model is: Where PI is the pavement performance index, PI0 is the initial value of the pavement performance index, t is the road age, and α and β are model parameters.

5. The adaptive regression method for predicting asphalt pavement performance according to claim 1 is characterized in that: The basic form of the prediction model for determining the pavement performance index also includes: a pavement rutting depth index, a pavement anti-skid performance index and a pavement structure strength index use a linear regression model.

6. The adaptive regression method for predicting asphalt pavement performance according to claim 5 is characterized in that: The linear regression model is: PI=PI0-αt, where PI is the pavement performance index, PI0 is the initial value of the pavement performance index, t is the road age, and α is the model parameter.

7. The adaptive regression method for predicting asphalt pavement performance according to claim 1 is characterized in that: The parameter value interval setting for the basic form of the prediction model includes: for the dual-parameter curve regression model, the α parameter interval is set to [12, 25], with a step size of 0.5; the β parameter interval is set to [0.3, 2], with a step size of 0.1; for the linear regression model, the α parameter interval is set to [0.1, 5], with a step size of 0.

05. According to the parameter setting, several groups of prediction equation forms containing different parameter combinations are generated.

8. The adaptive regression method for predicting asphalt pavement performance according to claim 1 is characterized in that: The method of obtaining the continuous decay detection data slice and calculating the relative road age according to the prediction model form includes: assuming that the performance index value of the initial year in the continuous decay detection data slice is in the prediction equation form, inverting the relative road age t of the initial year according to the determined prediction equation form and the performance index value, then the relative road age value of the second year is t+1, the subsequent points of the continuous decay slice are accumulated year by year on the basis of the relative road age of the initial year, and the performance index value of the corresponding year is calculated according to the prediction model form, and for several groups of prediction equation forms containing different parameter combinations, grid search is performed, and the least square method is used to determine the coefficient of determination R 2 The maximum is taken as the principle to obtain the optimal regression parameters.

9. The adaptive regression method for predicting asphalt pavement performance according to claim 1, characterized in that: It also includes model validation, selecting the root mean square error as the evaluation indicator to quantify the difference between the predicted value and the actual value. The smaller the RMSE, the better the prediction performance of the model. The formula is as follows: Where n is the number of samples; y i is the measured value; is the predicted value.

10. An adaptive regression system for predicting asphalt pavement performance, characterized in that: include: Classification module: used to classify road performance test data; Determination module: used to determine the basic form of the prediction model of pavement performance indicators; Parameter value setting module: used to set parameter value intervals for the basic form of the prediction model; Calculation module: used to obtain the detection data slices of continuous decay and calculate the relative road age according to the prediction model form; Adaptive regression module: It is used to perform grid search for several groups of prediction equations containing different parameter combinations, use the least squares method to obtain the optimal regression parameters, and perform adaptive regression of the prediction model.

Images