Traffic load historical characteristic evaluation method in long-term service process of road

Abstract

The invention discloses a traffic load historical characteristic evaluation method in a long-term service process of a road. The method comprises the following steps: collecting the existing road surface condition of a long-term service and non-maintenance historical road section and year-by-year historical traffic, road surface and weather condition data; the method comprises the following steps: determining half of a given standard axle load action frequency (ESAL) value and overload rates in road section service time corresponding to the ESAL respectively, and calculating a ratio (LOI) of the ESAL value and the overload rates; grouping the LOIs, and comparing the recession condition of the pavement damage condition index (PCI) along with the ESAL under different LOIs; and constructing a PCI prediction model, and comparing the prediction precision change of the model when the LOI is used as the independent variable and the importance sequence of the LOI in the independent variable set. According to the method, the LOI index is provided for quantitative evaluation through the historical loading characteristics of the traffic load of the long-term service road section, and the method has important guiding significance for traffic load control, pavement performance prediction and maintenance decision.

Description

Technical Field

[0001] The invention belongs to the technical field of road maintenance, and in particular relates to a method for evaluating historical characteristics of traffic loads during the long-term service of a road. Background Art

[0002] Traffic load is one of the key factors affecting pavement performance degradation. Accurately evaluating the impact of traffic load on pavement performance is crucial to making reasonable maintenance decisions. Commonly used traffic load measurement indicators include the number of standard axle load actions (ESAL) and the annual average daily traffic volume (AADT). These indicators are usually treated as cumulative values ​​or average values ​​when evaluating the traffic conditions of long-serving roads. However, this method ignores the specific loading history of traffic loads and cannot fully reflect the complexity of the actual stress conditions of the road. Previous studies have shown through indoor tests that different loading sequences will produce different failure modes for asphalt mixtures, affecting their service life and degree of damage. In addition, field observation data also show that even under the same ESAL, pavement structural materials and climatic conditions, the performance degradation rates of different sections may still be significantly different, resulting in different maintenance needs. This shows that traditional traffic load variables cannot fully capture the subtle differences in traffic load history in pavement performance prediction, which may lead to inaccurate maintenance decisions. Therefore, it is particularly important to accurately evaluate the historical characteristics of traffic loads for long-serving roads. This can not only provide support for accurate prediction of pavement performance, but also help optimize maintenance plans and ensure that roads maintain a high service level throughout their life cycle. To solve this problem, this method proposes an evaluation method for the historical characteristics of traffic loads during long-term service of roads. Based on many years of measured traffic axle load spectra and pavement performance test data, this method proposes quantitative indicators for measuring the historical characteristics of traffic loads. These indicators can better reflect the complexity of traffic loads and are used to analyze the decay law of pavement damage conditions, and then construct a prediction model for damage conditions. Through this method, the important contribution of traffic load history to the development and prediction of pavement damage conditions can be clarified, providing a more scientific basis for road performance prediction and maintenance decisions. This not only helps to extend the service life of roads and reduce maintenance costs, but also improves the overall service quality of roads, and has important practical application value and guiding significance. Summary of the invention

[0003] The present invention aims to solve the shortcomings of the existing evaluation methods of historical characteristics of traffic loads on long-term service roads, and innovatively designs a quantitative index to accurately measure the historical characteristics of traffic loads. It can not only be used to deeply analyze the decay law of pavement damage conditions, but also to build a prediction model for damage conditions, thereby clarifying the important role of historical characteristics of traffic loads on pavement damage conditions and their prediction. It provides a more scientific and reliable basis for accurate prediction of road performance and maintenance decisions, and helps improve the maintenance efficiency and service level of roads throughout their life cycle.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for evaluating historical characteristics of traffic loads during long-term service of a road, characterized by comprising the following steps:

[0006] Step 1: Collect data on the basic characteristics, pavement structure materials, traffic load conditions, pavement damage conditions, and weather conditions of roads that have been in service for ten years or more without maintenance;

[0007] Step 2: Determine the overload rate generated during the service time of the road section corresponding to half of the given standard axle load (ESAL) value and ESAL, and calculate their ratio, which is defined as the loading order index (LOI);

[0008] Step 3: Group the LOI indicators and use the box plot to compare the decay of the pavement condition index (PCI) with ESAL under different LOIs;

[0009] Step 4: Use random forest and artificial neural network algorithms to build PCI prediction models, and compare the changes in the prediction accuracy of the models when LOI is used as an independent variable and the importance ranking of LOI in the independent variable set.

[0010] Furthermore, the data items that need to be collected in the step one include: basic characteristics of the road section: road age, lane, road surface or bridge deck; pavement structural materials: modified asphalt layer thickness, asphalt surface layer thickness, asphalt surface layer gradation type, base layer thickness, base layer material; year-by-year traffic load conditions: ESAL, axle load spectrum; year-by-year pavement damage conditions: PCI; weather conditions: proportion of high temperature days, proportion of low temperature days, proportion of rainy days, annual average daily temperature, annual average daily precipitation.

[0011] Furthermore, the calculation method of the loading order index LOI in step 2 is as follows:

[0012] 1) For a road section with a service time of more than 10 years T, determine its corresponding ESAL and find the service time corresponding to ESAL / 2: T half=f(ESAL / 2),f:ESAL→t, where f represents a mapping from ESAL to service time t;

[0013] 2) Using the axle load spectrum information, calculate the T and T of the road section from opening to service half The corresponding overload rates for each year are:

[0014] Where OR(t) represents the overload rate from the opening of the road to year t, A represents the axle type set, including single axle, double axle and triple axle, a represents a specific axle type, Indicates time t ′ When the shaft type a is overloaded, Indicates time t ′ When , the total number of axes of axis type a;

[0015] 3) Calculate the loading order index corresponding to the service time T of the road section: LOI(T) = OR(T half ) / OR(T).

[0016] Further, the graphic display method in step three is as follows:

[0017] 1) For each road section with a service time of more than ten years, the LOI value was calculated and divided into three groups according to its value: <1.0, 1.0-1.2, and ≥1.2;

[0018] 2) For each LOI group, the PCI set corresponding to every 1 million ESAL increment from the opening of the road section to the specific service time is counted;

[0019] 3) Draw box plots and stacked plots in groups to determine the decline of PCI with ESAL under different LOIs.

[0020] Furthermore, the methods for constructing the PCI prediction model, comparing the prediction accuracy and evaluating the importance of LOI in step 4 are as follows:

[0021] 1) Analyze the factors affecting the pavement damage and determine the independent variable set of the PCI prediction model, including road age, lane, pavement or bridge deck, modified asphalt layer thickness, asphalt surface layer thickness, asphalt surface layer gradation type, base thickness, base material, ESAL, overload rate, LOI, proportion of high temperature days, proportion of low temperature days, proportion of rainy days, annual average daily temperature and annual average daily precipitation. Based on this, a data set for establishing the PCI prediction model is constructed, with 80% of the data used as a training set and 20% of the data used as a test set;

[0022] 2) The nominal variables and numerical variables in the independent variables were standardized by one-hot encoding and Z-score methods, and the neural network model and random forest model were constructed without considering LOI and with LOI respectively. The hyperparameters in the model were tuned by random search method. The hyperparameters of the neural network model included the number of hidden layers, the number of neurons in the hidden layer, the optimizer, the learning rate, the number of training rounds, and the batch size. The hyperparameters of the random forest model included the number of trees, the number of sampled features, the maximum depth of the tree, the minimum number of samples required to split the internal nodes, the minimum number of samples required for the leaf nodes, and whether the sample set was sampled with replacement.

[0023] 3) The coefficient of determination R2 was used to evaluate the prediction accuracy of the model on the test set, and the prediction accuracy of the two machine learning models before and after LOI was compared;

[0024] 4) Evaluate the importance of each independent variable for PCI prediction, focus on the importance ranking of LOI, and determine the importance of traffic load history characteristics on pavement performance degradation.

[0025] The historical characteristics of traffic loads on long-term service roads, especially the impact of loading sequence on pavement performance degradation, are fully considered, and quantitative indicators are innovatively proposed to overcome the limitations of existing traffic load indicators that only focus on cumulative values ​​or average values ​​and cannot effectively measure loading history. This quantification method is simple and easy to use, with good repeatability. Through graphical display and machine learning modeling verification, the key role of loading history characteristics in the development and prediction of pavement damage conditions is clarified, and the prediction accuracy is significantly improved. This invention provides a more scientific and accurate basis for road performance prediction and maintenance decisions, and has important practical application value and broad guiding significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A flow chart for evaluating the historical characteristics of traffic loads during the long-term service of roads;

[0027] Figure 2 It is the distribution map of LOI indicator;

[0028] Figure 3 The decline of PCI with ESAL at different LOI levels;

[0029] Figure 4 Rank the importance of independent variables in the PCI prediction model. DETAILED DESCRIPTION

[0030] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0031] A method for evaluating the historical characteristics of traffic loads during the long-term service of a road, such as Figure 1As shown, the following steps are included:

[0032] S1. Collect the basic characteristics of sections that have been in service for ten years or more without maintenance, the pavement structure materials, and the traffic load conditions, pavement damage conditions and weather conditions during the service period. Divide the highway into multiple sub-sections according to the pavement type (such as pavement or bridge deck), number of lanes, driving direction, structure and traffic characteristics of the highway. For sections longer than 1.5 kilometers, further subdivide them at intervals of 1 kilometer. Match the time nodes of all pavement performance tests over the years for each section of about 1 kilometer in length, and associate the corresponding traffic flow and weather condition variables to calculate the service time;

[0033] S2, determine half of the given standard axle load action times (ESAL) value and the overload rate generated during the service time of the road section corresponding to the ESAL, and calculate their ratio, which is defined as the loading order index (LOI). Figure 2 The figure shows the distribution of the LOI index. It can be found that a considerable proportion of road sections had a high overload rate in the early stage, resulting in a LOI greater than 1, that is, the initial overload rate was high, but it was effectively controlled later and the overload rate decreased.

[0034] S3. Group the LOI indicators. For each road section with a service time of more than ten years, divide the LOI into three groups: <1.0, 1.0-1.2, and ≥1.2. For each LOI group, count the PCI set corresponding to each 1 million ESAL increment from the opening of the road section to the specific service time, and draw a box plot for each group, such as Figure 3 As shown, the decay of PCI with ESAL under different LOI is analyzed. It can be seen from the figure that the larger the LOI value, the higher the decay rate of PCI, which indicates that different loading sequences will affect the damage condition of asphalt pavement.

[0035] S4. Analyze the factors affecting the pavement damage and determine the independent variable set of the PCI prediction model, including road age, lane, pavement or bridge deck, modified asphalt layer thickness, asphalt surface layer thickness, asphalt surface layer gradation type, base thickness, base material, ESAL, overload rate, LOI, proportion of high temperature days, proportion of low temperature days, proportion of rainy days, annual average daily temperature and annual average daily precipitation. Based on this, construct a data set for establishing the PCI prediction model. Take 80% of the data as the training set and 20% of the data as the test set. Use the one-hot encoding and Z-score methods to standardize the nominal variables and numerical variables in the independent variables, respectively. Without considering L and considering L In the case of OI, a neural network model and a random forest model were constructed respectively, and the random search method was used to tune the hyperparameters in the model. The hyperparameters of the neural network model include the number of hidden layers, the number of neurons in the hidden layer, the optimizer, the learning rate, the number of training rounds, and the batch size. The hyperparameters of the random forest model include the number of trees, the number of sampled features, the maximum depth of the tree, the minimum number of samples required to split internal nodes, the minimum number of samples required for leaf nodes, and whether to perform replacement sampling on the sample set. The determination coefficient R2 was used to evaluate the prediction accuracy of the model on the test set, and the prediction accuracy of the two machine learning models before and after LOI was compared, as shown in Table 1.

[0036] Table 1 Model evaluation results

[0037]

[0038] It can be seen that after adding LOI, the performance of both the neural network and random forest models are slightly improved, indicating that LOI has a certain contribution to the prediction of pavement damage conditions, and the performance of random forest is slightly better than that of the neural network model. Therefore, the random forest model is used, and the importance of each independent variable to PCI prediction is evaluated. As shown in the figure, it can be seen that among the three indicators related to traffic load, LOI contributes the most to the prediction of pavement damage conditions, even exceeding common indicators such as ESAL and overload rate, which illustrates the importance of considering the historical characteristics of traffic load in pavement performance prediction.

Claims

1. A method for evaluating historical characteristics of traffic loads during long-term service of a road, characterized in that: The following steps are involved: Step 1: Collect data on the basic characteristics, pavement structure materials, traffic load conditions, pavement damage conditions, and weather conditions of roads that have been in service for ten years or more without maintenance; Step 2: Determine the overload rate generated during the service time of the road section corresponding to half of the given standard axle load (ESAL) value and ESAL, and calculate their ratio, which is defined as the loading order index (LOI); Step 3: Group the LOI indicators and use box plots and other graphic display methods to compare the decline of the pavement condition index (PCI) with ESAL under different LOIs; Step 4: Use random forest and artificial neural network algorithms to build PCI prediction models, and compare the changes in the prediction accuracy of the models when LOI is used as an independent variable and the importance ranking of LOI in the independent variable set.

2. A method for evaluating historical characteristics of traffic load during long-term service of a road as claimed in claim 1, characterized in that: The data items that need to be collected in the step 1 include: basic characteristics of the road section: road age, lane, road surface or bridge deck; pavement structural materials: modified asphalt layer thickness, asphalt surface layer thickness, asphalt surface layer gradation type, base layer thickness, base layer material; year-by-year traffic load conditions: ESAL, axle load spectrum; year-by-year pavement damage conditions: PCI; weather conditions: proportion of high temperature days, proportion of low temperature days, proportion of rainy days, annual average daily temperature, annual average daily precipitation.

3. A method for evaluating historical characteristics of traffic load during long-term service of a road as claimed in claim 1, characterized in that: The calculation method of the loading order index LOI in step 2 is as follows: 1) For a road section with a service time of more than 10 years T, determine its corresponding ESAL and find the service time corresponding to ESAL / 2: T half =f(ESAL / 2),f:ESAL→t, where f represents a mapping from ESAL to service time t; 2) Using the axle load spectrum information, calculate the T and T of the road section from opening to service half The corresponding overload rates for each year are: Where OR(t) represents the overload rate from the opening of the line to year t, A represents a set of axis types, including single axis, double axis and triple axis, a represents a specific axis type, Indicates time t ′ When the shaft type a is overloaded, Indicates time t ′ When , the total number of axes of axis type a; 3) Calculate the loading order index corresponding to the service time T of the road section: LOI(T) = OR(T half ) / OR(T).

4. A method for evaluating historical characteristics of traffic load during long-term service of a road as claimed in claim 1, characterized in that: The graphical display method in step 3 is as follows: 1) For each road section with a service time of more than ten years, the LOI value was calculated and divided into three groups according to its value: <1.0, 1.0-1.2, and ≥1.2; 2) For each LOI group, the PCI set corresponding to every 1 million ESAL increment from the opening of the road section to the specific service time is counted; 3) Draw box plots and stacked plots in groups to determine the decline of PCI with ESAL under different LOIs.

5. A method for evaluating historical characteristics of traffic load during long-term service of a road as claimed in claim 1, characterized in that: The methods for constructing the PCI prediction model, comparing the prediction accuracy and evaluating the importance of LOI in step 4 are as follows: 1) Analyze the factors affecting the pavement damage and determine the independent variable set of the PCI prediction model, including road age, lane, pavement or bridge deck, modified asphalt layer thickness, asphalt surface layer thickness, asphalt surface layer gradation type, base thickness, base material, ESAL, overload rate, LOI, proportion of high temperature days, proportion of low temperature days, proportion of rainy days, annual average daily temperature and annual average daily precipitation. Based on this, a data set for establishing the PCI prediction model is constructed. Use 80% of the data as a training set and 20% of the data as a test set; 2) Use one-hot encoding and Z-score methods to standardize the nominal variables and numerical variables in the independent variables respectively. A neural network model and a random forest model were constructed without or with LOI, and the hyperparameters in the model were tuned using a random search method. The hyperparameters of the neural network model included the number of hidden layers, the number of neurons in the hidden layer, the optimizer, the learning rate, the number of training rounds, and the batch size. The hyperparameters of the random forest model included the number of trees, the number of sampled features, the maximum depth of the tree, the minimum number of samples required to split an internal node, the minimum number of samples required for a leaf node, and whether to perform sampling with replacement on the sample set. 3) The coefficient of determination R2 was used to evaluate the prediction accuracy of the model on the test set, and the prediction accuracy of the two machine learning models before and after LOI was compared; 4) Evaluate the importance of each independent variable for PCI prediction, focus on the importance ranking of LOI, and determine the importance of traffic load history characteristics on pavement performance degradation.

Images