A new-old pavement interlayer fatigue estimation method and system

Abstract

The application provides a new and old pavement interlayer fatigue prediction method and system based on CT and AVIZO pore characteristics, which comprises the following steps: CT scanning of test pieces under different conditions; introducing the two-dimensional images obtained after CT scanning into AVIZO software for visual processing, and reconstructing the three-dimensional images of the test pieces; extracting the pore structure in the test pieces by threshold segmentation, three-dimensionally reconstructing the pores, and counting the pore quantity in the test pieces, calculating the pore volume and area; sequentially extracting the open and closed pores and the connected and isolated pores in the test pieces, and displaying the pores and pore throats in the test pieces in the form of equivalent diameter display ratio to obtain the equivalent pore structure of the test pieces; extracting the characteristic information of the interlayer position pores to obtain the pore distribution characteristics at the interlayer position; and completing the interlayer fatigue prediction by using an ANFIS adaptive network fuzzy model, revealing the mechanism of interlayer damage, and the model has high precision, the interlayer damage condition is evaluated according to the prediction data, which is helpful to improve the road performance of the thin layer overlay technology, the service life of the thin layer overlay technology can be prolonged for the pavement using the thin layer overlay preventive maintenance technology, the arrival of the large and medium maintenance of the pavement is delayed, and the energy consumption and maintenance cost are reduced.

Description

Technical Field

[0001] This invention belongs to the field of road engineering pavement maintenance technology, and in particular relates to an interlayer damage evaluation model based on CT scanning and AVIZO visualization, with pore curvature as the evaluation index. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] After years of explosive growth in road construction, the overall performance of asphalt pavements has been declining due to increasing traffic loads and natural factors, leading to frequent pavement defects and making maintenance extremely challenging. Thin-layer overlay technology is widely used in pavement maintenance due to its thinness and low cost.

[0004] When the Road Condition Index (PCI) drops to a certain level and structural damage occurs, major and medium-scale road repairs are necessary. These repairs result in increased energy consumption and maintenance costs. According to the Statistical Bulletin on Transportation Development, the total mileage of highways requiring major and medium-scale repairs was 57,265 kilometers, with an investment of 132.475 billion yuan. The cost of major and medium-scale repairs is 4 to 10 times that of preventative maintenance. Therefore, extending the service life of preventative maintenance technologies and delaying the need for major and medium-scale repairs has been a continuous research topic for those skilled in the art. However, the inventors discovered that existing thin-layer overlay technologies are prone to cracking, shoving, and peeling under the combined effects of load and natural factors. Therefore, improving the road performance of thin-layer overlay technologies urgently needs to be addressed. Summary of the Invention

[0005] To overcome the shortcomings of the existing technology, this invention provides a method and system for predicting fatigue between new and old road layers. The purpose is to reveal the interlayer failure mechanism by analyzing the interlayer porosity characteristics under different conditions, thereby overcoming the deficiencies of thin overlay technology. By simulating conditions such as original, immersion, and freeze-thaw cycles, the porosity characteristics of the specimens are analyzed from a microscopic perspective to reveal the interlayer failure mechanism, which is beneficial to improving the road performance of thin overlays and delaying the need for major and medium road repairs.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0007] In a first aspect, the present invention discloses a method for predicting fatigue between old and new road layers, comprising:

[0008] New and old road surfaces under different conditions were taken as test specimens, and CT scans were performed on the test specimens under different conditions to obtain two-dimensional images after scanning.

[0009] The two-dimensional images obtained after CT scanning are imported into AVIZO software for visualization processing, and the three-dimensional images of the specimen are reconstructed.

[0010] Threshold segmentation is used to extract the pore structure from the 3D image of the specimen, the pores are reconstructed in 3D, the number of pores inside the specimen is counted, and the pore volume and area are calculated.

[0011] The open and closed pores, as well as the connected and isolated pores inside the specimen, are extracted in sequence. The pores and throats in the specimen are displayed using the equivalent diameter as the display ratio to obtain the equivalent pore structure of the specimen.

[0012] Extract the feature information of pores at interlayer locations to obtain the pore distribution characteristics at interlayer locations;

[0013] Interlayer fatigue prediction was performed using the ANFIS adaptive network fuzzy model.

[0014] As an alternative implementation method, the two-dimensional image obtained after CT scanning is imported into AVIZO software, and the image is denoised using median filtering before the three-dimensional image of the specimen is reconstructed.

[0015] As an alternative implementation method, after obtaining the equivalent pore structure of the specimen, the coordination number of each pore is calculated.

[0016] As an alternative implementation, the pore distribution characteristics at the interlayer location are obtained, including the pore radius and pore area at the interlayer location, and a spatial distribution map of the interlayer pore volume is generated based on the equivalent radius of the pores.

[0017] As an alternative implementation method, the ANFIS adaptive network fuzzy model uses pore tortuosity as the evaluation index, with equivalent diameter, volume shape factor, and porosity as input variables and tortuosity as the output variable.

[0018] As an alternative implementation method, training the ANFIS adaptive network fuzzy model includes:

[0019] S1: The input x and y values ​​are fuzzified using a membership function to obtain a membership degree μ in the range [0,1]. 0 represents that x is not an element of the fuzzy set, 1 represents that x is an element of the fuzzy set, and values ​​between 0 and 1 indicate that the element partially conforms to the fuzzy set.

[0020]

[0021] Where x and y are the input values ​​of node i, A i B i It is a fuzzy set That is A i B i The membership function value indicates that x and y belong to A.i B i The degree of membership function The shape is entirely determined by the parameters of the predecessor;

[0022] S2: Multiply the membership degrees of each variable to obtain the trigger strength of each rule. This layer is called the rule strength release layer. The output of each node represents the credibility of that rule.

[0023]

[0024] S3: Normalize the trigger strength of each rule in the second layer. This result will represent the trigger weight of that rule in the entire rule base.

[0025]

[0026] S4: In the fourth layer, the variable x needs to be input again. Each node i in this layer is an adaptive node, and its output is:

[0027]

[0028] In the formula: For the output of the third layer, {p i ,q i ,r i} represents the parameter set of this node, called the consequent parameters;

[0029] S5: Deblurring yields the exact result, which is the weighted average of the system's output for each rule.

[0030]

[0031] Given the antecedent parameters, the output value of the adaptive fuzzy inference system can be expressed as:

[0032]

[0033] The parameter update is performed during both forward and backward propagation. During the forward learning process of the network, n sets of training data input values ​​are used and the forward parameters are fixed. When the fourth layer is reached, the n input values ​​are calculated according to the least squares principle, and the original expected error value of the training data is backpropagated to update the backward parameters.

[0034] S6: In reverse learning, the parameters in each rule are defined. Using the chain rule and the loss function, the partial derivative of the measurement error with respect to each parameter is obtained, thus calculating the partial derivatives of the forward parameters. The parameters are then updated from the opposite direction of the gradient, where the measurement error E... P The sum of mean square errors,

[0035]

[0036]

[0037]

[0038] In the formula: T m,p It is the m-th component of the output of the p-th target. It is the m-th component of the L-th layer of the p-th actual output. The sum of the mean square errors of all P outputs is the root mean square error in one training session, denoted as E.

[0039] S7: Using equivalent diameter, volumetric shape factor, and porosity as input variables and tortuosity as output variable, import the data into the training model for simulation and prediction, and compare the error between the predicted pore tortuosity and the actual calculation results.

[0040] As an alternative implementation method, the specific operation process of the ANFIS adaptive network fuzzy model is as follows:

[0041] In MATLAB, input code to read the equivalent diameter, volumetric shape factor, and porosity of the specimen obtained from the visualization analysis of AVIZO software in Excel, and load the training, testing, and checking data. When generating the FIS, input the three variables: equivalent diameter, volumetric shape factor, and porosity. Enter 3 3 3 for fuzzification in the Number of MFs field. When training the FIS, in the Train FIS dialog box, enter 0.005 for the stop error and 25 for the number of training epochs, and then train. To view the error of the dataset, in the Test FIS dialog box, select Training data, Testing data, and then click Test Now. You can see the training error and the test data error. Then export the pore tortuosity prediction data and compare it with the calculated data. Similarly, complete the model training under different conditions.

[0042] Secondly, a fatigue prediction system for the interlayer of new and old road surfaces is disclosed, including:

[0043] The CT scanning module is configured to: take new and old road surfaces under different conditions as test specimens, perform CT scans on the test specimens under different conditions, and obtain two-dimensional images after scanning;

[0044] The AVIZO visualization module is configured to import two-dimensional images obtained after CT scanning into AVIZO software for visualization processing and to reconstruct three-dimensional images of the specimen.

[0045] The threshold segmentation module is configured to: extract the pore structure in the three-dimensional image of the specimen using threshold segmentation, reconstruct the pores in three dimensions, count the number of pores inside the specimen, and calculate the pore volume and area;

[0046] The equivalent pore structure module is configured to sequentially extract open and closed pores and connected and isolated pores inside the specimen, and display the pores and pore throats in the specimen using the equivalent diameter as the display ratio to obtain the equivalent pore structure of the specimen.

[0047] The interlayer porosity distribution feature module is configured to: extract feature information of pores at interlayer locations to obtain pore distribution features at interlayer locations;

[0048] The interlayer fatigue prediction module is configured to use the ANFIS adaptive network fuzzy model to perform interlayer fatigue prediction.

[0049] The above one or more technical solutions have the following beneficial effects:

[0050] 1. This invention uses CT scanning and AVIZO visualization to obtain the internal pore structure of specimens under original, immersion, and freeze-thaw conditions. By analyzing the pore characteristics at the interlayer, the mechanism of interlayer failure is revealed.

[0051] 2. The present invention constructs an interlayer damage evaluation model with pore tortuosity as the evaluation index. The error between the predicted pore tortuosity data and the calculated data obtained from the training is between 0.1 and 0.15, indicating high model accuracy. The interlayer damage is evaluated based on the predicted data.

[0052] 3. The detailed analysis of the interlayer failure mechanism of asphalt mixtures in this invention helps to improve the road performance of thin overlay technology. For pavements using thin overlay preventive maintenance technology, it can extend the service life of thin overlay technology, delay the arrival of major and medium road repairs, and reduce energy consumption and maintenance costs.

[0053] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0054] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0055] Figure 1 The specimen model is obtained by AVIZO through three-dimensional reconstruction of two-dimensional CT scan images;

[0056] Figure 2 To extract the pore structure of the specimen using threshold segmentation;

[0057] Figure 3 This is a spatial distribution diagram of connected pores and isolated pores in the specimen;

[0058] Figure 4 This is a schematic diagram of the equivalent pore structure of the specimen, shown proportionally to the equivalent diameter.

[0059] Figure 5 This is a characteristic diagram of the pore radius distribution at the interlayer location;

[0060] Figure 6 This is a diagram showing the characteristic distribution of pore area at interlayer locations;

[0061] Figure 7 To generate a spatial distribution map of interlayer pore volume based on the equivalent radius of the pores;

[0062] Figure 8 A diagram illustrating the code input for model building in MATLAB;

[0063] Figure 9 This is a screenshot of the interface when loading Checking data;

[0064] Figure 10 A schematic diagram of the FIS data model generation and training operation interface;

[0065] Figure 11 This is a schematic diagram of the dataset error interface after model training. Detailed Implementation

[0066] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0067] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0068] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0069] Example 1

[0070] This embodiment discloses a method for predicting fatigue between new and old road layers, including:

[0071] S101: CT scans were performed on the original specimens, after freeze-thaw cycles, and after immersion in water. The scan area was expanded during the scan to capture the overall morphological characteristics of the specimens, facilitating a more comprehensive analysis.

[0072] S102: As Figure 1 As shown, the two-dimensional images obtained after CT scanning are imported into AVIZO software for visualization processing, and the three-dimensional images of the specimens are reconstructed. By reconstructing the three-dimensional specimen structure, the changes between the phases of the interlayer asphalt mixture can be observed from multiple perspectives, breaking the limitations of analyzing data changes from a two-dimensional perspective.

[0073] Furthermore, the two-dimensional images obtained from the CT scan are imported into the software, and then median filtering is selected to denoise the images and reconstruct the three-dimensional images to avoid changing the edge information in the images.

[0074] Furthermore, the Volume Edit module was selected to cut the specimen to obtain a specimen structure of 50mm×50mm×90mm, which ensured that the superimposed images completely overlapped.

[0075] S103: As Figure 2 As shown, threshold segmentation is used to extract the pore structure in the specimen, the pores are reconstructed in three dimensions, and the number of pores inside the specimen is counted. The pore volume and area are calculated. The calculation of pore volume and area is to visualize the changes in the internal structure of the specimen after the original sample, freeze-thaw, and immersion in water, and to lay a data foundation for further analysis of the effects of freeze-thaw and immersion in water on the specimen.

[0076] Furthermore, threshold segmentation was used to extract the pore structure in the specimen. To analyze the distribution and changes in pore volume in the specimen, each pore in the specimen was divided into 9 ranges, in mm. 3 (0.1≤V; 0.1≤V<0.5; 0.5≤V<1; 1≤V<5; 5≤V<10; 10≤V<20; 20≤V<30; 30≤V<50; 50≤V); The number of pores in each range is statistically analyzed, and the maximum and average values ​​of the volume and area of ​​the pores in the specimen are calculated. Because the structure of asphalt mixture specimens is diverse, comparing the changes in overall porosity or individual pores cannot directly reveal the mechanism of environmental factors. Therefore, the porosity of the specimens is divided according to the volume size, and the changes in the internal structure of the asphalt mixture specimens are analyzed from the changes in pore distribution.

[0077] S104: As Figure 3 As shown, open and closed pores and connected and isolated pores inside the specimen are extracted in sequence, and the pores and pore throats in the specimen are displayed with the equivalent diameter as the display ratio to obtain the equivalent pore structure of the specimen.

[0078] Furthermore, to analyze the distribution of pores inside freeze-thaw and immersion specimens, open and closed pores, as well as connected and isolated pores, were extracted sequentially from the specimens. Connected pores were extracted using the Axis Connective module, and then the And Not Image module was used to remove connected pores from all pores, obtaining isolated pores in the specimens. The internal pores of asphalt mixtures can be divided into open and closed pores, connected and isolated pores. Since the formation conditions of different pores and their effects on performance are different, pores in asphalt mixture specimens were extracted according to category to reveal the mechanism by which freeze-thaw and immersion affect the performance of asphalt mixtures.

[0079] Furthermore, such as Figure 4 As shown, the Separate module is used to display the pore network structure in the material, and the pores and pore throats in the specimen are displayed with the equivalent diameter as the display scale. The equivalent pore structure diagrams of the three specimens are obtained, and the coordination number of each pore is calculated, which allows for a more intuitive observation of the pore distribution.

[0080] S105: As Figure 5 — Figure 7 The display shows the porosity features at the interlayer location. The feature information of 1mm pores at the interlayer location is extracted. The feature information is the area and volume of the pores at the interlayer location, and the porosity distribution features at the interlayer location are obtained.

[0081] Furthermore, the pore distribution characteristics at the interlayer location include: pore radius and pore area at the interlayer location, and based on the equivalent radius of the pore, i.e., the equivalent radius of the pore is half of the equivalent diameter, a spatial distribution map of the interlayer pore volume is generated.

[0082] S106: Using the ANFIS adaptive network fuzzy model to complete interlayer fatigue prediction;

[0083] Furthermore, the training of the ANFIS adaptive network fuzzy model includes:

[0084] S1: The input x and y values ​​are fuzzified using a membership function to obtain a membership degree μ in the range [0,1]. 0 represents that x is not an element of the fuzzy set, 1 represents that x is an element of the fuzzy set, and values ​​between 0 and 1 indicate that the element partially conforms to the fuzzy set.

[0085]

[0086] Where x and y are the input values ​​of node i, A i B i It is a fuzzy set That is A i B i The membership function value indicates that x and y belong to A. i Bi The degree of membership function The shape is entirely determined by the parameters of the predecessor;

[0087] S2: Multiply the membership degrees of each variable to obtain the trigger strength of each rule. This layer is called the rule strength release layer. The output of each node represents the credibility of that rule.

[0088]

[0089] S3: Normalize the trigger strength of each rule in the second layer. This result will represent the trigger weight of that rule in the entire rule base.

[0090]

[0091] S4: In the fourth layer, the variable x needs to be input again. Each node i in this layer is an adaptive node, and its output is:

[0092]

[0093] In the formula: For the output of the third layer, {p i ,q i ,r i} represents the parameter set of this node, called the consequent parameters;

[0094] S5: Deblurring yields the exact result, which is the weighted average of the system's output for each rule.

[0095]

[0096] Given the antecedent parameters, the output value of the adaptive fuzzy inference system can be expressed as:

[0097]

[0098] The parameter update is performed during both forward and backward propagation. During the forward learning process of the network, n sets of training data input values ​​are used and the forward parameters are fixed. When the fourth layer is reached, the n input values ​​are calculated according to the least squares principle, and the original expected error value of the training data is backpropagated to update the backward parameters.

[0099] S6: In reverse learning, the parameters in each rule are defined. Using the chain rule and the loss function, the partial derivative of the measurement error with respect to each parameter is obtained, thus calculating the partial derivatives of the forward parameters. The parameters are then updated from the opposite direction of the gradient, where the measurement error E... P The sum of mean square errors,

[0100]

[0101]

[0102]

[0103] In the formula: T m,p It is the m-th component of the output of the p-th target. It is the m-th component of the L-th layer of the p-th actual output. The sum of the mean square errors of all P outputs is the root mean square error in one training session, denoted as E.

[0104] S7: Using equivalent diameter, volumetric shape factor, and porosity as input variables, the equivalent diameter, volumetric shape factor, and porosity can be exported from the AVIZO software. The tortuosity is the output variable. The data is imported into the training model for simulation and prediction. The error comparison analysis is performed between the predicted pore tortuosity and the actual calculation results.

[0105] Furthermore, the specific operation process of the ANFIS adaptive network fuzzy model is as follows:

[0106] like Figure 8 As shown, input the code in MATLAB to read the equivalent diameter, volume shape factor, porosity and other relevant data of the specimen under the original conditions in Excel;

[0107] like Figure 9 As shown, in MATLAB, enter the command anfisedit to bring up the Neuro-Fuzzy Designer dialog box. In the Load data dialog box, select Training for Type and Worsp for From. Then click Load Data. In the pop-up dialog box, enter the variable name training to load the testing and checking data in sequence.

[0108] like Figure 10 As shown, when generating FIS, in the Generate FIS dialog box, select Grid partition, click Generate FIS, and in the pop-up dialog box, keep 3 3 unchanged in Number of MFs. This means that the input data is divided into three fuzzy categories: "small", "medium", and "large". Since the input variables are equivalent diameter, volume shape factor, and porosity, you need to enter 3 3 3. In MF Type, select the membership function form, and keep the original state unchanged here.

[0109] When training FIS, in the Train FIS dialog box, enter 0.005 for the stop error (Error Tolerance), enter 25 for the number of training epochs, change the Type to Training, and then start training;

[0110] like Figure 11 As shown, when viewing the dataset error, in the Test FIS dialog box, select Training data, Test data, and then click Test Now. The training error is 0.11901, and the test data error is 0.1294. Then, export the predicted data and compare it with the calculated data. Similarly, complete the model training under freeze-thaw and immersion conditions. After freeze-thaw, the training error is 0.102, and the test data error is 0.102; after immersion, the training error is 0.134, and the test data error is 0.115. This demonstrates that the predicted data obtained from the trained model has high accuracy.

[0111] Example 2

[0112] This embodiment discloses a fatigue prediction system between new and old road layers, including:

[0113] The CT scanning module is configured to: take new and old road surfaces under different conditions as test specimens, perform CT scans on the test specimens under different conditions, and obtain two-dimensional images after scanning;

[0114] The AVIZO visualization module is configured to import two-dimensional images obtained after CT scanning into AVIZO software for visualization processing and to reconstruct three-dimensional images of the specimen.

[0115] The threshold segmentation module is configured to: extract the pore structure in the three-dimensional image of the specimen using threshold segmentation, reconstruct the pores in three dimensions, count the number of pores inside the specimen, and calculate the pore volume and area;

[0116] The equivalent pore structure module is configured to sequentially extract open and closed pores and connected and isolated pores inside the specimen, and display the pores and pore throats in the specimen using the equivalent diameter as the display ratio to obtain the equivalent pore structure of the specimen.

[0117] The interlayer porosity distribution feature module is configured to: extract feature information of pores at interlayer locations to obtain pore distribution features at interlayer locations;

[0118] The interlayer fatigue prediction module is configured to use the ANFIS adaptive network fuzzy model to perform interlayer fatigue prediction.

[0119] Example 3

[0120] The purpose of this embodiment is to provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0121] Example 4

[0122] The purpose of this embodiment is to provide a computer-readable storage medium.

[0123] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above-described method.

[0124] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0125] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0126] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for predicting fatigue between new and old road surface layers, characterized in that, New and old road surfaces under different conditions were taken as test specimens, and CT scans were performed on the test specimens under different conditions to obtain two-dimensional images after scanning. The two-dimensional images obtained after CT scanning are imported into AVIZO software for visualization processing, and the three-dimensional images of the specimen are reconstructed. Threshold segmentation is used to extract the pore structure from the 3D image of the specimen, the pores are reconstructed in 3D, the number of pores inside the specimen is counted, and the pore volume and area are calculated. The open and closed pores, as well as the connected and isolated pores inside the specimen, are extracted in sequence. The pores and throats in the specimen are displayed using the equivalent diameter as the display ratio to obtain the equivalent pore structure of the specimen. Extract the feature information of pores at interlayer locations to obtain the pore distribution characteristics at interlayer locations; Interlayer fatigue prediction was performed using the ANFIS adaptive network fuzzy model. In the ANFIS adaptive network fuzzy model, pore tortuosity was used as the evaluation index, and equivalent diameter, volume shape factor, and porosity were used as input variables, while tortuosity was used as the output variable.

2. The method for predicting fatigue between new and old road layers as described in claim 1, characterized in that, After importing the two-dimensional images obtained from the CT scan into AVIZO software, the images are denoised using median filtering, and then the three-dimensional images of the specimen are reconstructed.

3. The method for predicting fatigue between new and old road layers as described in claim 1, characterized in that, After obtaining the equivalent pore structure of the specimen, the coordination number of each pore is calculated.

4. The method for predicting fatigue between new and old road layers as described in claim 1, characterized in that, The pore distribution characteristics at interlayer locations are obtained, including the pore radius and pore area at the interlayer locations, and a spatial distribution map of the interlayer pore volume is generated based on the equivalent radius of the pores.

5. The method and system for predicting fatigue between old and new road layers as described in claim 1, characterized in that, The training of the ANFIS adaptive network fuzzy model includes: S1: The input x and y values ​​are fuzzified using a membership function to obtain a membership degree μ in the range [0,1]. 0 represents that x is not an element of the fuzzy set, 1 represents that x is an element of the fuzzy set, and values ​​between 0 and 1 indicate that the element partially conforms to the fuzzy set. Where x and y are the input values ​​of node i, A i B i It is a fuzzy set That is A i B i The membership function value indicates that x and y belong to A. i B i The degree of membership function , The shape is entirely determined by the parameters of the predecessor; S2: Multiply the membership degrees of each variable to obtain the trigger strength of each rule. This layer is called the rule strength release layer. The output of each node represents the credibility of that rule. ； S3: Normalize the trigger strength of each rule in the second layer. This result will represent the trigger weight of that rule in the entire rule base. i=1,2 ； S4: In the fourth layer, the variable x needs to be input again. Each node i in this layer is an adaptive node, and its output is: In the formula: For the output of the third layer, The parameter set for this node is called the consequent parameter; S5: Deblurring yields the exact result, which is the weighted average of the system's output for each rule. i=1,2； Given the antecedent parameters, the output value of the adaptive fuzzy inference system can be expressed as: ； The parameter update is performed during both forward and backward propagation. During the forward learning process of the network, n sets of training data input values ​​are used and the forward parameters are fixed. When the fourth layer is reached, the n input values ​​are calculated according to the least squares principle, and the original expected error value of the training data is backpropagated to update the backward parameters. S6: In the back propagation, the parameters in each rule are defined, the partial derivative of the measurement error to each parameter is obtained by using the chain rule and according to the LOSS function, the partial derivative of the forward parameter is obtained, and the parameters are updated from the opposite direction of the gradient direction, wherein the measurement error E P is the sum of the mean square errors, In the formula: T m,p It is the m-th component of the output of the p-th target. It is the m-th component of the L-th layer of the p-th actual output. The sum of the mean square errors of all P outputs is the root mean square error in one training session, denoted as E. S7: Using equivalent diameter, volumetric shape factor, and porosity as input variables and tortuosity as output variable, import the data into the training model for simulation and prediction, and compare the error between the predicted pore tortuosity and the actual calculation results.

6. The method for predicting fatigue between new and old road layers as described in claim 1, characterized in that, The specific operation process of the ANFIS adaptive network fuzzy model is as follows: In MATLAB, input code to read the equivalent diameter, volumetric shape factor, and porosity of the specimen obtained from the visualization analysis of AVIZO software in Excel, and load the training, testing, and checking data. When generating the FIS, input the three variables: equivalent diameter, volumetric shape factor, and porosity. Enter 3 3 3 for fuzzification in the Number of MFs field. When training the FIS, in the Train FIS dialog box, enter 0.005 for the stop error and 25 for the training epochs, and then train. To view the error of the dataset, in the Test FIS dialog box, select Training data, Testing data, and then click Test Now. You can see the training error and the test data error. Then export the pore tortuosity prediction data and compare it with the calculated data. Similarly, complete the model training under different conditions.

7. A fatigue prediction system for the interlayer of new and old road surfaces, characterized in that, include: The CT scanning module is configured to: take new and old road surfaces under different conditions as test specimens, perform CT scans on the test specimens under different conditions, and obtain two-dimensional images after scanning; The AVIZO visualization module is configured to import two-dimensional images obtained after CT scanning into AVIZO software for visualization processing and to reconstruct three-dimensional images of the specimen. The threshold segmentation module is configured to: extract the pore structure in the three-dimensional image of the specimen using threshold segmentation, reconstruct the pores in three dimensions, count the number of pores inside the specimen, and calculate the pore volume and area; The equivalent pore structure module is configured to sequentially extract open and closed pores and connected and isolated pores inside the specimen, and display the pores and pore throats in the specimen using the equivalent diameter as the display ratio to obtain the equivalent pore structure of the specimen. The interlayer porosity distribution feature module is configured to: extract feature information of pores at interlayer locations to obtain pore distribution features at interlayer locations; The interlayer fatigue prediction module is configured to use the ANFIS adaptive network fuzzy model to perform interlayer fatigue prediction. In the ANFIS adaptive network fuzzy model, pore tortuosity is used as the evaluation index, and equivalent diameter, volume shape factor, and porosity are used as input variables, while tortuosity is used as the output variable.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a fatigue prediction method between new and old road layers as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements a fatigue prediction method between new and old road layers as described in any one of claims 1-6.

Images