Method and system for predicting fatigue life and rut life of asphalt pavement

Abstract

The invention provides a method and a system for predicting the fatigue life and the rut life of an asphalt pavement, and the method comprises the steps: inputting the road structure data of a to-be-predicted asphalt pavement into a trained neural network model, and obtaining the prediction results of the fatigue life and the rut life of the asphalt pavement; the road structure data of the to-be-predicted asphalt pavement comprises nine types of data including the thickness of an upper surface layer, the thickness of a middle surface layer, the thickness of a lower surface layer, the thickness of a base layer, the modulus of the upper surface layer, the modulus of the middle surface layer, the modulus of the lower surface layer, the modulus of the base layer and the modulus of a soil base layer. The invention relates to the technical field of road engineering, can directly predict the fatigue life and rut life of the asphalt pavement through the road structure data of the asphalt pavement, is simple and practical, and saves time and labor; during prediction, required data are easy to obtain, and the prediction result is high in precision; the method is not only suitable for predicting the service life of the asphalt pavement paved by brand new asphalt, but also suitable for predicting the service life of the asphalt pavement added with the RAP, especially high-parameter RAP.

Description

Technical Field

[0001] The present invention relates to the technical field of road engineering, and particularly relates to a method and system for predicting the fatigue life and rutting life of asphalt pavements. Background Art

[0002] The fatigue life of asphalt pavement is a key factor in measuring road safety and economic benefits. Insufficient road fatigue life will lead to premature appearance of diseases such as cracks and potholes on the road surface, increasing the risk of traffic accidents and maintenance costs. Accurate fatigue life prediction helps to timely discover potential safety hazards on the road, such as cracks and settlements, so as to formulate reasonable maintenance plans and take necessary maintenance measures to ensure driving safety; it also helps to ensure that the road operates in the best state, extend the service life of the road, avoid unnecessary road repairs and reconstructions, and save funds.

[0003] Rutting is the wheel rut left after vehicles drive on the road surface, and its depth directly reflects the usage condition of the road. By predicting the rutting life, the wear degree and service life of the road can be understood. Rutting life prediction helps to discover design defects of the road surface during the design stage, and then optimize the road surface structure design by adjusting the material ratio, surface layer thickness, etc. of the road surface; it also helps to customize targeted maintenance plans, such as local repair and overlay, to extend the service life of the road and improve driving comfort and safety.

[0004] With the continuous increase in traffic volume and the improvement of environmental awareness, the application of Reclaimed Asphalt Pavement (RAP) in road engineering is becoming more and more extensive. The incorporation of RAP can not only effectively reduce the use of new materials, reduce resource waste, but also significantly reduce the project cost, with significant economic and environmental benefits. However, the RAP content has an important impact on the fatigue life and rutting life of asphalt pavements, and the domestic and foreign research on asphalt pavements with RAP content mainly focuses on material properties, mixture design, construction technology, etc. Foreign research shows that the incorporation of RAP can significantly improve the rutting resistance performance of asphalt mixtures, but the impact on its fatigue resistance performance is relatively complex. Generally, as the RAP content increases, the fatigue life will decrease. Domestic research shows that the RAP content has a significant impact on the mechanical properties and durability of asphalt mixtures, but there is a lack of systematic methods for predicting fatigue life and rutting life. Currently, when building roads, it is the general trend to incorporate RAP into new asphalt mixtures. However, there is still a lack of effective methods for predicting the fatigue life and rutting life of roads incorporated with RAP, mainly because the types and contents of RAP have a complex impact on the fatigue life and rutting life of roads.

[0005] At present, the commonly used prediction methods for the fatigue life and rutting life of asphalt pavements mainly include: (1) Prediction methods based on experimental data, such as the asphalt linear amplitude sweep test. In this test, a dynamic shear rheometer is used to test the fatigue performance of asphalt binders under medium-temperature conditions. Based on the viscoelastic continuum damage mechanics analysis of the test data, the fatigue life of asphalt binders under different deformations is predicted. Although this test is simple to operate and has a short test time, the prediction results are often affected by factors such as experimental conditions and sample preparation; (2) Prediction methods based on material mechanical property parameters, such as the stress-life curve method (S-N curve method). This method is applicable to the prediction of high-cycle fatigue with a low stress level and the fatigue life of notch-free structures. It can be predicted through a simple stress-life relationship, which is easy to understand and apply. However, its disadvantage is that it does not consider the local plasticity at the notch root and the influence of the cyclic loading on fatigue, and it takes a lot of time and cost to obtain material fatigue performance data (S-N curve method). When calculating the fatigue life of structures with stress concentration, the calculation error is relatively large; (3) The local stress method. The disadvantage of this method is that the calculation result of the fatigue life is very sensitive to the fatigue notch factor K value and requires complex mathematical calculations and model establishment; (4) Prediction methods based on damage accumulation. This method predicts the fatigue life of materials through a damage accumulation model. The damage accumulation model converts the load history into a damage variable and predicts the fatigue life based on the accumulation of the damage variable. This method is applicable to situations where there is a clear relationship between the fatigue life of materials and damage accumulation. However, the establishment of its damage accumulation model requires a large amount of experimental data and parameter fitting, and there may be limitations when dealing with complex loading histories and multiaxial fatigue problems. At the same time, none of the above methods can be well applied to the life prediction of asphalt pavements containing RAP. Summary of the Invention

[0006] To solve at least one of the above technical problems, the present invention provides a prediction method and system for the fatigue life and rutting life of asphalt pavements, which can directly predict the fatigue life and rutting life of the asphalt pavement through the road structure data of the asphalt pavement.

[0007] The first aspect of the present invention provides a prediction method for the fatigue life and rutting life of asphalt pavements, including: inputting the road structure data of the asphalt pavement to be predicted into a trained neural network model to obtain the prediction results of the fatigue life and rutting life of the asphalt pavement.

[0008] Preferably, the road structure data of the asphalt pavement to be predicted includes nine types of data: the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, the thickness of the base layer, the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer.

[0009] Preferably, any of the above solutions is such that the trained neural network model is determined through the following steps: Step 1: Collect existing asphalt pavement data and form a road structure dataset based on it; Step 2: For each piece of data in the road structure dataset, establish a finite element model, substitute the corresponding road structure data into the finite element model, and calculate the corresponding maximum tensile and compressive strains; Step 3: Calculate the road fatigue life and road rutting life corresponding to the road structure data through the maximum tensile and compressive strains, and correspond the road fatigue life and road rutting life with the corresponding road structure data to form a training dataset; Step 4: Train a neural network with the training dataset to obtain a neural network model for predicting the fatigue life and rutting life of asphalt pavements.

[0010] Preferably, in Step 1, each piece of data in the road structure dataset includes nine types of data: the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, the thickness of the base layer, the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer.

[0011] Preferably in any of the above solutions, in Step 1, for the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, and the thickness of the base layer, determine multiple specific values according to the collected maximum and minimum values and typical values; for the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer, according to the collected maximum and minimum values, and at the same time according to the vehicle speed of the traffic load and the size of the pavement temperature, determine multiple specific values through the dynamic modulus master curve.

[0012] Preferably in any of the above solutions, in Step 1, combine the multiple specific values of various types of data to form a road structure dataset including multiple pieces of road structure data.

[0013] Preferably in any of the above solutions, the maximum value of the modulus of the upper layer, the modulus of the middle layer, and the modulus of the lower layer is set to 30000 MPa.

[0014] Preferably in any of the above solutions, in Step 2, when establishing the finite element model, the thickness of the soil base layer is set to infinity.

[0015] Preferably in any of the above solutions, in Step 2, establish an ABAQUS finite element model.

[0016] Preferably in any of the above solutions, in Step 3, according to the formula: , Calculate the road fatigue life N f , where k f1 , k f2 , k f3They are the first fatigue global field correction coefficient, the second fatigue global field correction coefficient, and the third fatigue global field correction coefficient; β f1 , β f2 , β f3 are the first fatigue local field correction coefficient, the second fatigue local field correction coefficient, and the third fatigue local field correction coefficient respectively; ɛ t is the maximum tensile strain; C is the thickness correction coefficient, which is used to correct the influence of crack load transfer capacity on fatigue damage, and its value depends on the crack type; C H is the thickness adjustment coefficient, which is used to correct the influence of subgrade support conditions on the stress of the pavement slab, and its value depends on the crack type; E is the modulus of the lower layer.

[0017] Preferably, in any of the above solutions, in step 3, according to the formula: , calculate the rutting life N r of the road, where k r1 , k r2 , k r3 are the first rutting global field correction coefficient, the second rutting global field correction coefficient, and the third rutting global field correction coefficient respectively; β r1 , β r2 , β r3 are the first rutting local field correction coefficient, the second rutting local field correction coefficient, and the third rutting local field correction coefficient respectively; Δp is the accumulated permanent deformation of the asphalt layer; T is the pavement temperature; K Z is the depth confining pressure coefficient, ɛ r is the maximum tensile strain.

[0018] Preferably, in any of the above solutions, according to the formula: K z = (C1 + C2I depth ) × 0.328196I depth calculate the depth confining pressure coefficient K Z , where C1 = -0.1039h 2 + 2.4868h - 17.342, C2 = 0.0172h 2 - 1.7331h + 27.428, I depth is the depth from the road surface to the calculation point, and h is the thickness of the asphalt layer.

[0019] Preferably, in any of the above solutions, in step 4, the neural network adopts a BP neural network.

[0020] Preferably, in any of the above solutions, in step 4, the training data set is divided into a training set and a test set according to a ratio of 8:2.

[0021] Preferably, in any of the above solutions, in step 4, the road structure data is used as the input feature of the BP neural network, and the corresponding road fatigue life and road rutting life are used as the output features of the BP neural network.

[0022] The second aspect of the present invention provides a prediction system for the fatigue life and rutting life of an asphalt pavement. The system includes a processor on which a computer program runs. The processor runs the computer program to execute the prediction method for the fatigue life and rutting life of the asphalt pavement.

[0023] The prediction method and system for the fatigue life and rutting life of the asphalt pavement of the present invention have the following beneficial effects: 1. The fatigue life and rutting life of the asphalt pavement can be directly predicted through the road structure data of the asphalt pavement, which is simple, practical, time-saving and labor-saving; 2. The data required for prediction is simple to obtain, and the prediction result has high accuracy; 3. It is not only applicable to predicting the life of asphalt pavements newly paved with asphalt, but also applicable to predicting the life of asphalt pavements added with reclaimed asphalt mixture (RAP), especially asphalt pavements with high RAP content; 4. The complex influence of the type and content of RAP on the life of asphalt pavements is transformed into the influence of the thickness and modulus of each surface layer on the tensile and compressive strains under standard loads, and the life is calculated through the tensile and compressive strains, and a neural network model for life prediction is formed by training the sample set, which greatly reduces the difficulty of predicting the life of asphalt pavements incorporated with RAP, and at the same time has prediction accuracy; 5. It can provide data support for road departments, which is convenient for road departments to optimize road design, improve road durability and reduce road maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of a preferred embodiment of the prediction method for the fatigue life and rutting life of an asphalt pavement according to the present invention.

[0025] Figure 2 It is for the prediction method and system for the fatigue life and rutting life of an asphalt pavement according to the present invention as Figure 1 shown in the schematic diagram of the determination process of the trained neural network model of the embodiment.

[0026] Figure 3 It is for the prediction method for the fatigue life and rutting life of an asphalt pavement according to the present invention as Figure 1 shown in the schematic diagram of the fatigue life prediction result of the embodiment.

[0027] Figure 4 It is for the prediction method for the fatigue life and rutting life of an asphalt pavement according to the present invention as Figure 1Schematic diagram of rutting life prediction results of the illustrated embodiment. Detailed implementation manners

[0028] To better understand the present invention, the present invention will be described in detail below in conjunction with specific embodiments.

[0029] Embodiment 1 As Figure 1 shown, a method for predicting the fatigue life and rutting life of an asphalt pavement includes: inputting the road structure data of the asphalt pavement to be predicted into a trained neural network model to obtain the prediction results of the fatigue life and rutting life of the asphalt pavement.

[0030] The road structure data of the asphalt pavement to be predicted includes nine types of data: the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, the thickness of the base layer, the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer.

[0031] As Figure 2 shown, the trained neural network model is determined through the following steps: Step 1: Collect existing asphalt pavement data and form a road structure data set according to it; Step 2: For each piece of data in the road structure data set, establish a finite element model, substitute the corresponding road structure data into the finite element model, and calculate the corresponding maximum tensile and compressive strains; Step 3: Calculate the road fatigue life and road rutting life corresponding to the road structure data through the maximum tensile and compressive strains, and correspond the road fatigue life and road rutting life with the corresponding road structure data to form a training data set; Step 4: Train a neural network with the training data set to obtain a neural network model for predicting the fatigue life and rutting life of the asphalt pavement.

[0032] In Step 1, each piece of data in the road structure data set includes nine types of data: the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, the thickness of the base layer, the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer. For the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, and the thickness of the base layer, multiple specific values are determined according to the maximum and minimum values and typical values collected thereof; for the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer, multiple specific values are determined according to the maximum and minimum values collected thereof and according to the vehicle speed of the driving load and the size of the road surface temperature through the dynamic modulus master curve. The multiple specific values of various data are combined to form a road structure data set including multiple pieces of road structure data.

[0033] It should be noted that in step 1, when collecting existing asphalt pavement data, it is not limited to collecting domestic data, but can also include foreign data; it is not limited to collecting actual road data, but can also include design / theoretical data; it is not limited to roads paved with brand-new asphalt, but can also include roads with RAP added. It should be further noted that when determining the specific values of various types of data, the maximum value, minimum value, and typical value collected should be included, and when necessary, the maximum value, minimum value, and typical value should be rounded as the specific value.

[0034] Preferably in this embodiment, considering that the moduli of the upper, middle, and lower layers of the road with RAP added are larger than those of the upper, middle, and lower layers of the road paved with brand-new asphalt, and different RAP parameters have an impact on the moduli of the upper, middle, and lower layers, the maximum values of the upper layer modulus, middle layer modulus, and lower layer modulus are set to 30,000 MPa. For each type of data, the specific values determined are shown in Table 1.

[0035] Table 1 Specific values of various types of data

[0036] Further preferably in this embodiment, when combining multiple specific values of various types of data to form multiple road structure data, considering the actual situation of domestic asphalt pavements, the following thickness combinations of upper layer thickness / middle layer thickness / lower layer thickness / base layer thickness are formed: 5 / 7 / 11 / 45, 4 / 6 / 10 / 36, 4 / 6 / 10 / 45, 4 / 6 / 10 / 54, 4 / 5 / 8 / 45. Then, the above thickness combinations are combined with the specific values of the moduli of each layer to form a road structure dataset including 5*9*9*9*3*2 = 21,870 road structure data.

[0037] It should be noted that it is also possible to arbitrarily combine the specific values of the upper layer thickness, middle layer thickness, lower layer thickness, and base layer thickness to form 2*3*3*3 = 54 thickness combinations, and then combine them with the specific values of the moduli of each layer to form a road structure dataset including 54*9*9*9*3*2 = 236,196 road structure data. It should be understood that the more data included in the road structure dataset, the more data in the formed training dataset, and the better the generalization ability of the trained neural network model. However, it also means that more costs are required when establishing the ABAQUS finite element model, and more computing resources are required when training the neural network model.

[0038] In Step 2, for each road structure data in the road structure dataset, a finite element model is established, and the corresponding road structure data is substituted into the finite element model to calculate the corresponding maximum tensile and compressive strains. It should be noted that when establishing the finite element model, the thickness of the soil base layer is set to be infinite. It should be understood that after Step 2, 21,870 maximum tensile and compressive strain values corresponding to the road structure data can be obtained. In this embodiment, preferably, the finite element model is established through ABAQUS software. It should be understood that the finite element model can also be established through other software, as long as the corresponding maximum tensile and compressive strains can be obtained based on the established finite element model and the road structure data.

[0039] In Step 3, according to the formula: , calculate the road fatigue life N f , where k f1 , k f2 , k f3 are the first fatigue global field correction coefficient, the second fatigue global field correction coefficient, and the third fatigue global field correction coefficient respectively; β f1 , β f2 , β f3 are the first fatigue local field correction coefficient, the second fatigue local field correction coefficient, and the third fatigue local field correction coefficient respectively; ɛ t is the maximum tensile strain; C is the thickness correction coefficient used to correct the influence of crack load transfer ability on fatigue damage, and its value depends on the crack type; C H is the thickness adjustment coefficient used to correct the influence of the base support condition on the pavement slab stress, and its value depends on the crack type; E is the modulus of the bottom layer. In this embodiment, preferably, K f1 = 0.007566, k f2 = -3.949200, k f3 = -1.281000, β f1 = 1, β f2 = 1, β f3 = 1.

[0040] According to the formula: , calculate the road rutting life N r , where k r1 , k r2 , k r3 are the first rutting global field correction coefficient, the second rutting global field correction coefficient, and the third rutting global field correction coefficient respectively; β r1 , β r2 , β r3They are the first rut local field correction coefficient, the second rut local field correction coefficient, and the third rut local field correction coefficient respectively; Δp is the cumulative permanent deformation of the asphalt layer; T is the pavement temperature; K Z is the depth confining pressure coefficient, ɛ r is the maximum tensile strain. According to the formula: K z = (C1 + C2I depth ) × 0.328196I depth Calculate the depth confining pressure coefficient K Z , where C1 = -0.1039h 2 + 2.4868h - 17.342, C2 = 0.0172h 2 - 1.7331h + 27.428, I depth is the depth from the road surface to the calculation point, and h is the thickness of the asphalt layer. It should be noted that the calculation point refers to the location of the maximum compressive strain at the middle depth of the asphalt layer, and the asphalt layer thickness refers to the sum of the thicknesses of the upper, middle, and lower layers. In this embodiment, preferably, k r1 = -3.35412, k r2 = 0.49710, k r3 = 1.56060, β r1 = 1, β r2 = 1, β r3 = 1.

[0041] It should be noted that the formulas for calculating the road fatigue life and road rut life based on the maximum tensile and compressive strains are the formulas determined according to the American AASHTO specification, and the effectiveness of these formulas has been verified. It should be further noted that through step 3, 21,870 road fatigue lives and road rut lives corresponding to the road structure data can be obtained, and then a training data set including 21,870 training data can be obtained.

[0042] In step 4, the neural network uses a BP neural network. The training data set is divided into a training set and a test set according to a ratio of 8:2. The road structure data is used as the input feature of the BP neural network, and the corresponding road fatigue life and road rut life are used as the output features of the BP neural network.

[0043] It should be noted that the training of the BP neural network can refer to the methods already disclosed in the prior art, and this application does not make special limitations. It should be further noted that before using the training data set, the training data set can be processed first, such as denoising, enhancement, etc. Since the road structure data in the training data set uses a unified expression method, and the road fatigue life and road rut life are also calculated using a unified calculation formula, therefore, there is no need to perform standardization processing on the training data set.

[0044] In step 4, after training the BP neural network with the training set, the trained BP neural network is tested with the test set, and some of the test results are shown in Tables 2 to 6.

[0045] Table 2 Test Results (Thickness of the upper layer / middle layer / lower layer / base layer: 5 / 7 / 11 / 45 (unit: cm))

[0046] Table 3 Test Results (Thickness of the upper layer / middle layer / lower layer / base layer: 4 / 6 / 10 / 36 (unit: cm))

[0047] Table 4 Test Results (Thickness of the upper layer / middle layer / lower layer / base layer: 4 / 6 / 10 / 45 (unit: cm))

[0048] Table 5 Test Results (Thickness of the upper layer / middle layer / lower layer / base layer: 4 / 6 / 10 / 54 (unit: cm))

[0049] Table 6 Test Results (Thickness of the upper layer / middle layer / lower layer / base layer: 4 / 5 / 8 / 45 (unit: cm))

[0050] From the above Tables 2 to 6, it can be found that the minimum error between the fatigue life calculated by the BP neural network model and the fatigue life calculated by the formula is -3.382497846%, and the maximum value is 2.659147921%; the minimum error between the rutting life calculated by the BP neural network model and the rutting life calculated by the formula is -2.55823875%, and the maximum value is 2.872266098%. This shows that the trained BP neural network model has high accuracy.

[0051] Table 7 shows the coefficient of determination R of the trained BP neural network for fatigue life prediction and rutting life prediction 2 , from Table 7, it can be found that the coefficient of determination R of the trained BP neural network for fatigue life prediction and rutting life prediction 2 is very close to 1, that is, the trained BP neural network model is effective.

[0052] Table 7 Model Coefficient of Determination for Fatigue Life and Rutting Life (R 2 )

[0053] Figure 3 and Figure 4 respectively show the test results of fatigue life and rutting life. According to Figure 3 and Figure 4 it can be found that the predicted values of the BP neural network model and the values calculated by the formula are distributed near the straight line with a slope of 1, indicating that the BP neural network model has a high fitting degree. At the same time, it can be found that the predicted results of fatigue life are more distributed at higher values compared to the predicted results of rutting life, which is consistent with "the value of rutting life usually does not get too high and will be concentrated within a certain range, while the upper limit of the value of fatigue life is usually higher and the distribution range is wider".

[0054] Example 2 A prediction system for the fatigue life and rutting life of an asphalt pavement, the system includes a processor, and a computer program runs on the processor, and the processor runs the computer program to execute the prediction method for the fatigue life and rutting life of the asphalt pavement.

[0055] Example 3 When using the prediction method and system for the fatigue life and rutting life of the asphalt pavement, the asphalt pavement to be predicted is an asphalt pavement in the design stage. By inputting the specific design values of the road structure data designed for it, including the thickness of the upper layer, the middle layer, the lower layer, the base layer, the modulus of the upper layer, the middle layer, the lower layer, the base layer, and the soil base layer, into the trained BP neural network model, the predicted values of the fatigue life and rutting life of the asphalt pavement can be obtained, and thus provide guidance for the optimization design of the asphalt pavement.

[0056] Example 4 When using the prediction method and system for the fatigue life and rutting life of the asphalt pavement, the asphalt pavement to be predicted is an asphalt pavement that has been built and is in use. Core samples are taken on the asphalt pavement and the road structure data of the core samples are measured, including the thickness of the upper layer, the middle layer, the lower layer, the base layer, the modulus of the upper layer, the middle layer, the lower layer, the base layer, and the soil base layer. By inputting the specific measured values of the measured road structure data into the trained BP neural network model, the predicted values of the fatigue life and rutting life of the asphalt pavement can be obtained, and thus provide guidance for the maintenance of the asphalt pavement, improve the durability, safety and comfort of the asphalt pavement, and reduce the maintenance cost.

[0057] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the foregoing embodiments have described the present invention in detail, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features, and these replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the present invention.

Claims

1. A prediction method for the fatigue life and rutting life of an asphalt pavement, characterized in that: It includes: inputting the road structure data of the asphalt pavement to be predicted into the trained neural network model to obtain the prediction results of the fatigue life and rutting life of the asphalt pavement; the road structure data of the asphalt pavement to be predicted includes nine types of data: the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, the thickness of the base layer, the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer. The trained neural network model is determined through the following steps: Step 1: Collect existing asphalt pavement data and form a road structure data set according to it. Step 2: For each piece of data in the road structure data set, establish a finite element model, substitute the corresponding road structure data into the finite element model, and calculate the corresponding maximum tensile and compressive strains. Step 3: Calculate the road fatigue life and road rutting life corresponding to the road structure data through the maximum tensile and compressive strains, and correspond the road fatigue life and road rutting life with the corresponding road structure data to form a training data set. Step 4: Train the neural network with the training data set to obtain a neural network model for predicting the fatigue life and rutting life of the asphalt pavement.

2. The prediction method of the fatigue life and rutting life of the asphalt pavement according to claim 1, wherein: In Step 1, each piece of data in the road structure data set includes nine types of data: the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, the thickness of the base layer, the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer.

3. The prediction method for the fatigue life and rutting life of an asphalt pavement according to claim 2, wherein: In Step 1, for the thickness of the upper layer, the thickness of the middle layer, the thickness of the lower layer, and the thickness of the base layer, determine multiple specific values according to their maximum and minimum values and typical values collected; for the modulus of the upper layer, the modulus of the middle layer, the modulus of the lower layer, the modulus of the base layer, and the modulus of the soil base layer, determine multiple specific values according to their maximum and minimum values collected, and at the same time, according to the vehicle speed of the traffic load and the size of the pavement temperature, determine them through the dynamic modulus master curve.

4. The prediction method of the fatigue life and rutting life of an asphalt pavement according to claim 3, characterized in that: In Step 1, combine the multiple specific values of various types of data to form a road structure data set including multiple pieces of road structure data.

5. The prediction method of the fatigue life and rutting life of the asphalt pavement according to claim 3, characterized in that: The maximum value of the modulus of the upper layer, the modulus of the middle layer, and the modulus of the lower layer is set to 30000 MPa.

6. The prediction method for the fatigue life and rutting life of an asphalt pavement according to claim 1, characterized in that: In Step 2, when establishing the finite element model, the thickness of the soil base layer is set to infinity.

7. The prediction method for the fatigue life and rutting life of an asphalt pavement according to claim 1, characterized in that: In Step 3, according to the formula: ， Calculate the road fatigue life N f , where k f1 , k f2 , k f3 are the first fatigue global field correction coefficient, the second fatigue global field correction coefficient, and the third fatigue global field correction coefficient respectively; β f1 , β f2 , β f3 are the first fatigue local field correction coefficient, the second fatigue local field correction coefficient, and the third fatigue local field correction coefficient respectively; ɛ t is the maximum tensile strain; C is the thickness correction coefficient, used to correct the influence of crack load transfer ability on fatigue damage, and its value depends on the crack type; C H is the thickness adjustment coefficient, used to correct the influence of subbase support conditions on the stress of the pavement slab, and its value depends on the crack type; E is the modulus of the lower layer.

8. The prediction method for the fatigue life and rutting life of an asphalt pavement according to claim 1, characterized in that: In Step 3, according to the formula: ， Calculate the rutting life N of the road r , where k r1 , k r2 , k r3 are the first rut global field correction coefficient, the second rut global field correction coefficient, and the third rut global field correction coefficient respectively; β r1 , β r2 , β r3 are the first rut local field correction coefficient, the second rut local field correction coefficient, and the third rut local field correction coefficient respectively; Δp is the accumulated permanent deformation of the asphalt layer; T is the pavement temperature; K Z is the depth confining pressure coefficient, ɛ r is the maximum compressive strain.

9. The prediction method of the fatigue life and rutting life of an asphalt pavement according to claim 8, characterized in that: According to the formula: K z = (C1 + C2I depth ) × 0.328196I depth Calculate the coefficient of depth confining pressure K Z , where C1 = -0.1039h 2 + 2.4868h - 17.342, C2 = 0.0172h 2 - 1.7331h + 27.428, I depth is the depth from the road surface to the calculation point, and h is the thickness of the asphalt layer.

10. A prediction system for the fatigue life and rutting life of an asphalt pavement, including a processor, and a computer program is running on the processor, characterized in that: The processor runs the computer program to execute the method for predicting the fatigue life and rutting life of the asphalt pavement according to any one of claims 1-9.

Images