Comprehensive Identification Method for Pavement Maintenance and Outliers in Condition Data Based on Data Driving

Abstract

The present invention belongs to the technical field of road surface condition data monitoring, and specifically relates to a comprehensive identification method for road surface maintenance and abnormal values of condition data based on data-driven. This method uses normal data sets to train a maintenance measure probability prediction model based on an artificial neural network and a road surface condition prediction model based on a Bayesian neural network; first, the trained maintenance measure probability prediction model based on the artificial neural network is used to identify and correct abnormal maintenance data, and then the road surface condition prediction model based on the Bayesian neural network is used to identify and correct abnormal road surface condition index data, and the 2σ principle is combined to dynamically adjust the abnormal thresholds of maintenance data and road surface condition index for different road sections. The present invention can be used to improve the data quality of the road surface management system and help managers make scientific and reasonable maintenance decisions.

Description

Technical Field

[0001] The present invention belongs to the technical field of pavement condition data monitoring, and particularly relates to a comprehensive identification method for pavement maintenance and outlier of condition data based on data driving. Background Art

[0002] With the continuous expansion of the highway network and the continuous growth of traffic flow, the contradiction between maintenance requirements and limited financial investment has become increasingly prominent. In this context, the scientific nature and effectiveness of the pavement management system are crucial for ensuring road performance, service life, and the reasonable allocation of maintenance funds. The core of the pavement management system lies in accurately evaluating the pavement condition and formulating a reasonable maintenance plan. However, there are still many limitations in the existing pavement management system in terms of data processing and model construction, which restrict the reliability of the system and the scientific nature of decision-making.

[0003] Firstly, the pavement management system highly depends on the quality of pavement condition data and maintenance data. These data reflect the physical state, performance, maintenance history, and effects of the pavement, and are the basis for making maintenance decisions and reasonably allocating maintenance funds. However, due to possible errors in the data collection, transmission, and storage processes, the pavement condition data and maintenance data often contain outliers, which will significantly reduce the reliability of the system.

[0004] Secondly, there are obvious deficiencies in feature extraction, model generalization ability, and uncertainty quantification in the existing methods. Traditional models are mostly based on static or single-time-point data. For example, the performance changes of the same road section are significantly different between the rainy season and the dry season, but the existing methods are difficult to capture such dynamic features. In addition, the pavement state is affected by multiple random factors (such as sudden traffic loads, extreme weather), and the existing deterministic models cannot quantify the uncertainty of the prediction results, which restricts the reliability of decision-making.

[0005] Finally, the existing research on pavement management data identification mostly focuses on a single type of data, that is, either only focuses on pavement condition data or only focuses on maintenance data, lacking a method that can comprehensively identify and correct outliers in pavement condition and maintenance data. For example, some studies identify outliers in pavement condition data through statistical methods or machine learning models, while other studies focus on the outlier detection of maintenance data. These methods fail to fully consider the internal relationship between pavement condition data and maintenance data, resulting in the inability to comprehensively identify outliers in the two types of data.

[0006] In view of the above problems, there is an urgent need for a comprehensive identification method that can integrate multi-source data, dynamically optimize the model structure, and quantify uncertainty to improve the scientific nature and adaptability of pavement maintenance decision-making. Summary of the Invention

[0007] The purpose of the present invention is to provide a comprehensive identification method for pavement maintenance and condition data outliers based on data-driven to solve the problems raised in the background art.

[0008] To achieve the above technical objectives, the technical solutions adopted by the present invention are as follows:

[0009] A comprehensive identification method for pavement maintenance and condition data outliers based on data-driven, comprising the following steps:

[0010] S1: Identify outliers in pavement management data and divide the pavement management data into a normal data set and an abnormal data set;

[0011] S2: Use the normal data set to train a maintenance measure probability prediction model based on an artificial neural network and a pavement condition prediction model based on a Bayesian neural network respectively;

[0012] S3: First, use the trained maintenance measure probability prediction model based on an artificial neural network (ANN) to identify and correct abnormal maintenance data, then use the pavement condition prediction model based on a Bayesian neural network (BNN) to identify and correct abnormal pavement condition index data, and combine the 2σ principle to dynamically adjust the abnormal thresholds of maintenance data and pavement condition index for different road sections. Finally, comprehensively identify and correct the data set.

[0013] Further preferably, the pavement management data includes pavement condition data, maintenance data, road attribute data, traffic data, climate and environment data.

[0014] Further preferably, the pavement condition data includes: the pavement condition index (PCI) of the previous year and the pavement condition index of the current year; the maintenance data includes: no maintenance, thin layer anti-skid surface layer, microsurfacing, fine surface treatment, ultra-viscous wearing course, ultra-thin wearing course, in-situ hot recycling, milling and replacing one layer, milling and replacing two layers; the road attribute data includes: surface layer thickness, surface layer material, base layer thickness, base layer material, road age; the traffic data includes: annual average daily traffic volume, annual average daily truck traffic volume.

[0015] Further preferably, the maintenance measure probability prediction model based on an artificial neural network (ANN) includes:

[0016] 13 input variables, including: the pavement condition index (PCI) of the previous year, surface layer thickness, surface layer material, base layer thickness, base layer material, road age, annual average daily traffic volume, annual average daily truck traffic volume, number of high temperature days, number of consecutive high temperature days, number of low temperature days, number of consecutive low temperature days, precipitation;

[0017] Hidden layer: 2 layers, using the ReLU activation function;

[0018] Output layer: 9 output nodes, corresponding to the probabilities of 9 maintenance measures. The 9 maintenance measures include: no maintenance, thin anti-skid surface layer, microsurfacing, fine surface treatment, ultra-thin bonding wearing course, ultra-thin wearing course, in-situ hot recycling, milling and replacing one layer, and milling and replacing two layers;

[0019] Use the Softmax function to output the probability distribution.

[0020] Further preferably, the pavement condition prediction model based on the Bayesian neural network (BNN) includes:

[0021] Input layer: 14 input variables, including: pavement condition index (PCI) of the previous year, surface layer thickness, surface layer material, base layer thickness, base layer material, road age, average annual daily traffic volume, average annual daily truck traffic volume, number of high-temperature days, consecutive high-temperature days, number of low-temperature days, consecutive low-temperature days, precipitation, and maintenance measures;

[0022] Hidden layer: 2 layers, using the ReLU activation function;

[0023] Output layer: 1 output node, corresponding to the predicted value of the pavement condition index for the current year.

[0024] Further preferably, the loss value of the maintenance measure probability prediction model based on the artificial neural network (ANN) is the cross-entropy loss.

[0025] Further preferably, the loss value of the pavement condition prediction model based on the Bayesian neural network (BNN) is the mean square error (MSE), and the accuracy rate is the coefficient of determination R 2 。

[0026] Further preferably, in step S3, use the trained maintenance measure probability prediction model based on the artificial neural network (ANN) to calculate the probability of each maintenance measure for each section, and identify and correct abnormal maintenance data; according to the 2σ principle, if the probability of a certain maintenance measure is less than 0.05, then determine that this maintenance measure is abnormal and replace it with the maintenance measure with the highest probability.

[0027] Further preferably, in step S3, calculate the reasonable range of the pavement condition index for each section according to the 2σ principle; if the pavement condition index for the current year exceeds this reasonable range, then determine that this pavement condition index is abnormal and replace it with the mean value of the pavement condition index predicted by the pavement condition prediction model based on the Bayesian neural network (BNN).

[0028] By analyzing the internal rules between road surface condition data and maintenance data, the present invention classifies the data into a normal data set and an abnormal data set; constructs a maintenance measure probability prediction model based on an artificial neural network (ANN) and a road surface condition prediction model based on a Bayesian neural network (BNN), which are respectively used to identify and correct abnormal maintenance data and road surface condition data; through actual case verification, the proportion of abnormal data is reduced from 47.79% to 6.35%, proving the effectiveness of the method. By integrating multi-source data, constructing a phased single-channel neural network model architecture, and combining a dynamic threshold optimization mechanism, the present invention realizes the comprehensive identification of outliers in road maintenance and condition data, significantly improves the data quality, helps managers make scientific and reasonable maintenance decisions, and can also promote the rational allocation of road maintenance funds. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The present invention can be further illustrated by the non-limiting embodiments given in the accompanying drawings.

[0030] Figure 1 is the method flow chart of the present invention;

[0031] Figure 2 is the change diagram of the loss value and accuracy of the maintenance measure probability prediction model based on the artificial neural network;

[0032] Figure 3 is the change diagram of the loss value and accuracy of the road surface condition prediction model based on the Bayesian neural network (BNN);

[0033] Figure 4 is the probability value of all maintenance measures in the maintenance measure probability prediction model based on the artificial neural network;

[0034] Figure 5 is the percentage comparison diagram of abnormal data before and after the implementation of the comprehensive identification and correction method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0035] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0036] As Figure 1 shown, a data-driven comprehensive identification method for outliers in road maintenance and condition data of the present invention includes the following steps;

[0037] S1: Identify outliers in road management data and divide the road management data into a normal data set and an abnormal data set;

[0038] S2: Use the normal data set to train a maintenance measure probability prediction model based on an artificial neural network and a road surface condition prediction model based on a Bayesian neural network respectively;

[0039] S3: First, use the trained probability prediction model of maintenance measures based on artificial neural network (ANN) to identify and correct abnormal maintenance data. Then, use the pavement condition prediction model based on Bayesian neural network (BNN) to identify and correct abnormal pavement condition index data, and dynamically adjust the abnormal thresholds of maintenance data and pavement condition index for different road sections in combination with the 2σ principle. Finally, integrate the dataset after identification and correction.

[0040] In step S1 of this embodiment, the pavement management data includes pavement condition data, maintenance data, road attribute data, traffic data, climate and environment data; the pavement condition data includes: the pavement condition index (PCI) of the previous year and the pavement condition index of the current year; the maintenance data includes: no maintenance, thin anti-skid surface layer, microsurfacing, fine surface treatment, ultra-viscous wearing course, ultra-thin wearing course, in-situ hot recycling, milling and replacing one layer, milling and replacing two layers; the road attribute data includes: surface layer thickness, surface layer material, base layer thickness, base layer material, road age; the traffic data includes: annual average daily traffic volume, annual average daily truck traffic volume; the climate and environment data includes: number of high-temperature days, consecutive high-temperature days, number of low-temperature days, consecutive low-temperature days, precipitation.

[0041] In this embodiment, the daily average temperature > 25°C and < 0°C are defined as high temperature and low temperature respectively.

[0042] (1) Total number of high-temperature days: The total number of days in a year when the daily average temperature > 25°C.

[0043] (2) Consecutive high-temperature days: The number of days in a year when the daily average temperature > 25°C for 3 consecutive days or more.

[0044] (3) Total number of low-temperature days: The total number of days in a year when the daily average temperature < 0°C.

[0045] (4) Consecutive low-temperature days: The number of days in a year when the daily average temperature < 0°C for 3 consecutive days or more.

[0046] (5) Precipitation: The total amount of precipitation in a specific area within a year in the form of rain, snow, sleet or hail.

[0047] In step S1 of this embodiment, data standardization processing is also performed, and then outlier identification is carried out. Standardize the input variables (such as PCI, road attributes, traffic data, climate data, etc.) to ensure that the data is on the same dimension, which is convenient for subsequent model training.

[0048] Based on the inherent law between pavement maintenance measures and the Pavement Condition Index (PCI): In the case of no maintenance, the pavement condition shows a gradually deteriorating trend over time; while when maintenance measures are taken, the pavement condition index of the current year will increase to a certain extent. Outlier identification is performed on pavement management data, including three abnormal situations, as shown in Table 1:

[0049] Table 1 Judgment of data abnormal situations

[0050]

[0051] The maintenance measure probability prediction model based on the Artificial Neural Network (ANN) includes:

[0052] 13 input variables, including: the pavement condition index (PCI) of the previous year, surface layer thickness, surface layer material, base layer thickness, base layer material, road age, annual average daily traffic volume, annual average daily truck traffic volume, number of high-temperature days, consecutive high-temperature days, number of low-temperature days, consecutive low-temperature days, precipitation. Hidden layer: 2 layers, using the ReLU activation function. Output layer: 9 output nodes, corresponding to the probabilities of 9 maintenance measures. The 9 maintenance measures include: no maintenance, thin anti-skid surface layer, micro-surfacing, fine surface treatment, ultra-viscous wearing course, ultra-thin wearing course, in-situ hot recycling, milling and replacing one layer, milling and replacing two layers. The Softmax function is used to output the probability distribution.

[0053] Based on experience and multiple attempts, the hyperparameters of the maintenance measure probability prediction model based on the Artificial Neural Network (ANN) are determined: including the loss function (cross-entropy loss), learning rate (0.001), batch size (500), number of neurons in the hidden layer (36, 36), optimizer (Adam), activation function (ReLU), number of training epochs (1000).

[0054] The pavement condition prediction model based on the Bayesian Neural Network (BNN) includes:

[0055] Input layer: 14 input variables, including: the pavement condition index (PCI) of the previous year, surface layer thickness, surface layer material, base layer thickness, base layer material, road age, annual average daily traffic volume, annual average daily truck traffic volume, number of high-temperature days, consecutive high-temperature days, number of low-temperature days, consecutive low-temperature days, precipitation, maintenance measures. Hidden layer: 2 layers, using the ReLU activation function. Output layer: 1 output node, corresponding to the predicted value of the pavement condition index of the current year. Applying Monte Carlo Dropout: Random dropout (probability 0.3) is introduced in the hidden layer, and the uncertainty of the pavement condition index prediction is simulated through multiple forward propagations to increase the robustness of the model.

[0056] Based on experience and multiple attempts, the hyperparameters of the pavement condition prediction model based on the Bayesian neural network (BNN) are determined: including the loss function (mean squared error MSE), learning rate (0.001), batch size (500), number of neurons in the hidden layer (36, 36), optimizer (Adam), activation function (ReLU), number of training epochs (1000), number of simulations (20), and dropout probability (0.3).

[0057] The loss value of the maintenance measure probability prediction model based on the artificial neural network (ANN) is the cross-entropy loss:

[0058] ;

[0059] The prediction accuracy formula is as follows:

[0060] ;

[0061] Among them, is the loss value of the maintenance measure probability prediction model based on the artificial neural network (ANN), i is the category, represents the true label, with a value of 1 only at the position corresponding to the target category and 0 at other positions; is the predicted probability output by the maintenance measure probability prediction model based on the artificial neural network (ANN), that is, the output value after being processed by the softmax function; is the number of correct predictions of the model; is the total number, is the prediction accuracy of the maintenance measure probability prediction model based on the artificial neural network.

[0062] The loss value of the pavement condition prediction model based on the Bayesian neural network (BNN) is the mean squared error (MSE), and the formula is:

[0063] ;

[0064] The accuracy is the coefficient of determination R 2 and the formula is:

[0065] ;

[0066] Among them, is the loss value of the pavement condition prediction model based on the Bayesian neural network (BNN), is the total number of predicted values or true values; represents the number of simulations; represents the th true value; represents the th predicted value in the th run; is the average value of all values, and is the accuracy rate of the pavement condition prediction model based on the Bayesian neural network (BNN).

[0067] According to the 2σ principle, it is judged whether the current value is abnormal, that is, 95% of the data will fall within the interval of the mean plus or minus twice the standard deviation, otherwise it is regarded as an outlier. The normal interval formula is as follows:

[0068] ;

[0069] where, is the mean value, is the standard deviation.

[0070] In step S3 of this embodiment, the trained probability prediction model of maintenance measures based on the artificial neural network (ANN) is used to calculate the probability of each maintenance measure for each road section, and abnormal maintenance data is identified and corrected. According to the 2σ principle, if the probability of a certain maintenance measure is lower than 0.05, it is determined that the maintenance measure is abnormal and is replaced by the maintenance measure with the highest probability. This process realizes the dynamic adjustment of the abnormal threshold of maintenance data for different road sections.

[0071] The trained pavement condition prediction model based on the Bayesian neural network (BNN) is used to identify and correct abnormal pavement condition index data, and the reasonable range of the pavement condition index of each road section in the current year is calculated according to the 2σ principle. If the pavement condition index in the current year exceeds this reasonable range, it is determined that the pavement condition index is abnormal and is replaced by the mean value of the pavement condition index predicted by the pavement condition prediction model based on the Bayesian neural network (BNN). This operation realizes the dynamic adjustment of the abnormal threshold of the pavement condition index for different road sections.

[0072] For the first two abnormal situations in Table 1, if the maintenance data is normal, the pavement condition index is abnormal; if the maintenance data is abnormal, it is still necessary to continue to judge whether the pavement condition index is abnormal. The abnormal maintenance measure is replaced by the maintenance measure with the highest predicted probability value of the probability prediction model of maintenance measures based on the artificial neural network (ANN).

[0073] For the third abnormal situation in Table 1, it is regarded that the pavement condition index situation in the current year must be abnormal. Then it is judged whether the maintenance measure is abnormal. If it is abnormal, it is replaced by the maintenance measure with the highest probability value.

[0074] In this embodiment, taking a certain expressway as an example, according to the judgment rules in Table 1, the abnormal data set and the normal data set are distinguished. There are 12,632 (47.79%) abnormal data and 13,802 (52.21%) normal data. Subsequently, using the normal data set, a maintenance measure probability prediction model based on artificial neural network (ANN) and a pavement condition prediction model based on Bayesian neural network (BNN) are constructed respectively.

[0075] The values of 13 input features are linearly transformed to follow a standard normal distribution (i.e., Gaussian distribution) with a mean of 0 and a standard deviation of 1. One-hot encoding is performed on 9 maintenance measures, and each maintenance measure is represented as a vector of length 9, where the corresponding position is 1 and the rest are 0.

[0076] The maintenance measure probability prediction model based on artificial neural network (ANN) is trained using the normal data set, and the prediction accuracy finally converges to 93.6%, and the loss value converges to 0.156, as Figure 2 shown; the pavement condition prediction model based on Bayesian neural network (BNN) is trained using the normal data set, and the prediction accuracy finally converges to 0.733, and the loss value converges to 0.266, as Figure 3 shown. It shows that the established model has good accuracy.

[0077] To facilitate understanding of how the present invention uses the maintenance measure probability prediction model based on artificial neural network (ANN) and the pavement condition prediction model based on Bayesian neural network (BNN) for comprehensive identification, four sections are selected as case studies. Table 2 lists the detailed data of these four sections.

[0078] Table 2 Detailed data of sections

[0079]

[0080] First, it is necessary to determine the type of abnormal section data. For section 1, maintenance measures are applied but the pavement condition deteriorates. For section 2, no maintenance is carried out. For section 3, maintenance measures are taken, but its pavement condition remains unchanged. For section 4, no maintenance measures are taken, but the pavement condition coefficient improves.

[0081] Subsequently, the values of the first 13 variables of the four sections in Table 2 are input into the maintenance measure probability prediction model based on artificial neural network (ANN) to obtain the probability values of all maintenance measures, as Figure 4 shown.

[0082] For section 1, its maintenance measure is a super sticky wearing course. Corresponding to Figure 4, the probability value of this maintenance measure is less than the threshold of 0.05, indicating that this maintenance measure is abnormal. Then, this abnormal maintenance measure is replaced with the maintenance measure with the highest probability value. Although the maintenance measure is corrected, it is still necessary to determine whether the pavement condition index (PCI) for the current year is abnormal.

[0083] For Sections 2 and 3, the PCI data for the current year must be abnormal, and the maintenance measures are no maintenance and ultra-thin bonded wearing course respectively. In Figure 4 , the probability value of no maintenance for Section 2 is greater than 0.05, indicating that no action is normal. The probability value of the ultra-thin bonded wearing course for Section 3 is less than 0.05, indicating that the ultra-thin bonded wearing course is abnormal. Then, this abnormal maintenance measure (ultra-thin bonded wearing course) is replaced with the maintenance measure with the highest probability value (no maintenance).

[0084] For Section 4, the maintenance measure is no maintenance. In Figure 4 , the probability value of no maintenance is greater than 0.05, indicating that this maintenance measure is normal and does not need to be replaced. This also further indicates that the PCI data for the current year must be abnormal.

[0085] Finally, the abnormal PCI data is identified based on the pavement condition prediction model of Bayesian neural network (BNN) and the 2σ principle. The first 13 variables in Table 1 and the corrected maintenance data are input into the pavement condition prediction model of Bayesian neural network (BNN). Each set of input data is repeatedly input into the pavement condition prediction model of Bayesian neural network (BNN) 1000 times to generate 1000 PCI prediction values. Then, the mean (μ) and standard deviation (σ) of these prediction values are calculated to obtain the reasonable range of PCI values , as shown in Table 3.

[0086] Table 3 Reasonable PCI value (pavement condition index) range for sections

[0087]

[0088] Referring to Table 3, for Section 1, the PCI value for the current year is 89.25, which is between (56.94, 97.65), indicating that the PCI value for the current year is reasonable; the current-year PCI values for Sections 2 and 3 are abnormal and are corrected to the corresponding means in Table 3, which are 94.55 and 89.67 respectively; for Section 4, since the maintenance measure is normal, the current-year PCI value must be abnormal and is corrected to 72.12.

[0089] According to the above comprehensive recognition process, data of other road sections are also recognized and corrected. With the help of the abnormal judgment criteria in Table 1, it is found that there are still 1,678 pieces of abnormal road section data, accounting for 6.35% of the total data. This may be because the normal data used for training the model is not very appropriate or the classification method for normal data and abnormal data is relatively simple, and some hidden abnormal situations may be inevitably included. As Figure 5 shown, the proportion of abnormal data before correction is 47.79%, and after correction is 6.35%, which proves the effectiveness of the comprehensive recognition method proposed by the present invention.

[0090] Compared with the prior art, the present invention adopts the above technical solutions and has the following technical effects: The present invention follows the inherent characteristics and change rules of asphalt pavement maintenance and condition data, and constructs a comprehensive recognition method for abnormal values of pavement maintenance and condition data based on multi-source data fusion and a phased neural network model. Moreover, the artificial neural network model and the Bayesian neural network model are coupled, and through reasonable model architecture design and dynamic threshold adjustment mechanism, efficient and accurate recognition and correction of abnormal values of pavement maintenance and condition data are realized. The present invention provides a more accurate and reliable decision-making support tool for the field of asphalt pavement management technology, which is of great significance for improving the scientificity and rationality of pavement maintenance management and ensuring the operation quality of the highway network, and is a powerful support for promoting the development of pavement maintenance work towards intelligence and refinement.

[0091] The above embodiments are only used to exemplarily illustrate the principles and effects of the present invention, rather than to limit the present invention. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or changes completed by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed by the present invention should still be covered by the claims of the present invention.

Claims

1. A comprehensive identification method for pavement maintenance and abnormal values of condition data based on data driving, characterized in that It includes the following steps: S1: Identify outliers in the pavement management data and divide the pavement management data into a normal data set and an abnormal data set; S2: Use the normal data set to train a maintenance measure probability prediction model based on an artificial neural network and a pavement condition prediction model based on a Bayesian neural network respectively; S3: First, use the trained maintenance measure probability prediction model based on an artificial neural network to calculate the probability of each maintenance measure for each section, and identify and correct abnormal maintenance data. The abnormal maintenance data includes abnormalities with maintenance records and abnormalities without maintenance records; According to the 2σ principle, if the probability of a certain maintenance measure is lower than the set abnormal threshold of section maintenance data, then determine that the maintenance measure is abnormal and replace it with the maintenance measure with the highest probability; Then, use the pavement condition prediction model based on a Bayesian neural network to identify and correct abnormal pavement condition index data, and calculate the reasonable range of the pavement condition index for each section in the current year according to the 2σ principle; If the pavement condition index in the current year exceeds this reasonable range, then determine that the pavement condition index is abnormal and replace it with the average value of the pavement condition index predicted by the pavement condition prediction model based on a Bayesian neural network; Combine the 2σ principle to dynamically adjust the abnormal thresholds of maintenance data and pavement condition index for different sections, and at the same time identify the types of abnormalities. The types of abnormalities include: abnormal maintenance measures, abnormal pavement condition index, and both abnormal maintenance measures and pavement condition index; Finally, comprehensively identify and correct the data set.

2. The data-driven comprehensive identification method for pavement maintenance and abnormal values of condition data according to claim 1, characterized in that The pavement management data includes pavement condition data, maintenance data, road attribute data, traffic data, climate and environmental data.

3. The comprehensive identification method of pavement maintenance and condition data outliers based on data driving according to claim 2, characterized in that The pavement condition data includes: the pavement condition index of the previous year and the pavement condition index of the current year; The maintenance data includes: no maintenance, thin layer anti-skid surface layer, microsurfacing, fine surface treatment, ultra-viscous wearing course, ultra-thin wearing course, in-situ hot recycling, milling and replacing one layer, milling and replacing two layers; The road attribute data includes: surface layer thickness, surface layer material, base layer thickness, base layer material, road age; The traffic data includes: annual average daily traffic volume, annual average daily truck traffic volume.

4. The method for comprehensively identifying outliers in pavement maintenance and condition data based on data driving according to claim 1, wherein The maintenance measure probability prediction model based on an artificial neural network includes: 13 input variables, including: the pavement condition index of the previous year, surface layer thickness, surface layer material, base layer thickness, base layer material, road age, annual average daily traffic volume, annual average daily truck traffic volume, number of high temperature days, consecutive high temperature days, number of low temperature days, consecutive low temperature days, precipitation; Hidden layer: 2 layers, using the ReLU activation function; Output layer: 9 output nodes, corresponding to the probabilities of 9 maintenance measures. The 9 maintenance measures include: no maintenance, thin layer anti-skid surface layer, microsurfacing, fine surface treatment, ultra-viscous wearing course, ultra-thin wearing course, in-situ hot recycling, milling and replacing one layer, milling and replacing two layers; Use the Softmax function to output the probability distribution.

5. The method for comprehensively identifying outliers in pavement maintenance and condition data based on data-driven according to claim 1, characterized in that The pavement condition prediction model based on a Bayesian neural network includes: Input layer: 14 input variables, including: pavement condition index of the previous year, surface layer thickness, surface layer material, base layer thickness, base layer material, road age, average annual daily traffic volume, average annual daily truck traffic volume, number of high-temperature days, consecutive high-temperature days, number of low-temperature days, consecutive low-temperature days, precipitation, maintenance measures; Hidden layer: 2 layers, using ReLU activation function; Output layer: 1 output node, corresponding to the predicted value of the pavement condition index for the current year.

6. The comprehensive identification method for pavement maintenance and abnormal values of condition data based on data-driven according to claim 1, characterized in that, The loss value of the maintenance measure probability prediction model based on artificial neural network is cross-entropy loss.

7. The data-driven comprehensive identification method for pavement maintenance and abnormal values of condition data according to claim 1, wherein The loss value of the pavement condition prediction model based on the Bayesian neural network is the mean square error, and the accuracy rate is the coefficient of determination R 2 .

Images