Intelligent design method for low-temperature high-toughness asphalt pavement material

Abstract

The application discloses an intelligent design method for low-temperature high-toughness asphalt pavement materials, and belongs to the field of asphalt pavement materials. The application aims to solve the problem that the traditional design method for low-temperature toughness-oriented asphalt pavement materials cannot meet the requirements of service environments in cold regions. A cross-scale database is constructed to cover low-temperature toughness data of asphalt, microstructure images of mixtures and low-temperature performance data. Through a data-physical dual-driven model, CNN and MLP are combined to process images and physical parameters, and low-temperature fracture performance of the asphalt mixture is predicted. The method realizes intelligent mapping from material composition, microstructure to macro performance, provides a digital design method for high-toughness asphalt pavement materials in cold regions, and significantly improves the research and development efficiency. The application realizes full-process digitalization and intelligentization from material modeling to performance prediction. The method of the application significantly shortens the research and development cycle of asphalt pavement materials in cold regions, and provides key technical support for digital design of asphalt pavement materials and intelligent research and development based on the material genome concept.

Description

Technical Field

[0001] This invention belongs to the field of asphalt pavement materials, and more specifically, relates to an intelligent design method for low-temperature, high-toughness asphalt pavement materials. Background Technology

[0002] In frigid regions such as Qinghai-Tibet and Northeast China, prolonged periods of low temperatures in winter pose a severe challenge to the service performance and lifespan of asphalt pavements. Traditional asphalt pavement material development relies primarily on experimental and trial-and-error methods, which suffer from long development cycles and high costs, making it difficult to meet the requirements for material toughness under complex low-temperature environments.

[0003] With the informatization of materials simulation technology, data-driven materials design methods, by constructing materials performance databases and combining them with machine learning algorithms for material optimization, offer new insights for the research and development of high-toughness transportation infrastructure materials. This method establishes a mapping relationship between material composition and macroscopic properties based on existing experimental data, enabling accurate prediction of material properties.

[0004] However, existing methods are mostly limited to the macroscopic physical parameters of materials and fail to fully consider the influence of the heterogeneity of the microstructure of asphalt mixtures on macroscopic low-temperature toughness. Asphalt mixtures are porous multiphase composite materials composed of asphalt, aggregates, and mineral powders. Their internal microstructure exhibits significant heterogeneity due to the randomness of the molding process and the complexity of the material composition. Among these, the packing pattern and contact morphology of the aggregates directly determine the structural characteristics of the mixture, thus having a key impact on macroscopic low-temperature toughness. Summary of the Invention

[0005] To address the problem that traditional design methods for asphalt pavement materials, which are oriented towards low-temperature toughness, cannot meet the requirements of cold-region service environments, a smart design method for low-temperature, high-toughness asphalt pavement materials is proposed.

[0006] The method of this invention includes a digital design method using numerical simulation and a data-driven intelligent prediction method for the low-temperature performance of asphalt-based materials. It realizes the full-process digitalization and intelligentization from modeling to performance prediction, significantly shortens the R&D cycle of asphalt pavement materials in cold regions, and provides key technical support for the digital design of asphalt pavement materials and the intelligent R&D of genome-based materials.

[0007] To address the aforementioned technical problems, the present invention adopts the following technical solution: The purpose of this invention is to provide an intelligent design method for low-temperature, high-toughness asphalt pavement materials, which specifically includes the following steps: I. Construction of a multi-scale database for the low-temperature toughness properties of asphalt mixtures: Microstructural images and low-temperature performance of asphalt mixtures. (1) Establishment of a dataset on the low-temperature toughness of asphalt materials: Step 1: Select base asphalt of different grades to prepare modified asphalt with different styrene-butadiene (SBS) content and different rubber powder (CR) content; Step 2: Based on the single-sided notched beam bending (SENB) test and bending beam rheology (BBR) test, test the fracture characteristics and rheological characteristics of each asphalt type under different temperature and heat preservation time conditions; Step 3: Establish a dataset with a one-to-one correspondence between asphalt grade, modifier type, modifier dosage, insulation time, fracture characteristics, and rheological characteristics; remove outlier data from the dataset; and mean the dataset to reduce the influence of dimensions. (2) Establishment of image dataset of asphalt mixture skeleton structure: Step 4: Prepare asphalt mixtures with different gradations, namely AC, SMA, and OGFC, according to the specifications. Step 5: Use an industrial CT scanner to perform a cross-sectional scan of the asphalt mixture, and perform image enhancement and segmentation on the CT images. Step 6: Reconstruct the longitudinal section of the asphalt mixture using 3D reconstruction software; Step 7: According to the function and size of the aggregates, divide the aggregates in the asphalt mixture, extract the microstructural features of the asphalt mixture, and simplify the aggregate contact structure using a ball-and-stick model. Step 8: Establish a quantitative microstructure characterization method based on topological theory, and statistically analyze the skeleton contact structure characteristics of asphalt mixtures with different gradations; Step 9: Based on topological statistical properties, develop a two-dimensional digital model of asphalt mixture and generate images of the skeleton structure of asphalt mixture with real microstructure for different gradations. (3) Establishment of a dataset on the low-temperature toughness of asphalt mixtures: Step 10: Use Python to perform secondary development on the finite element software Abaqus, conduct low-temperature freezing fracture numerical simulation tests on the microscale of asphalt mixtures, and obtain simulation data of each sample under different modified asphalt and different gradation combinations. (4) Construction of cross-scale datasets: Step 11: Use software to annotate the image of the asphalt mixture skeleton structure and ensure that the image is effectively input into the deep learning model; Step 12: Construct a cross-scale database of low-temperature toughness of asphalt materials, contact structure images of asphalt mixture skeleton, and low-temperature fracture performance of asphalt mixtures. II. Construction of the Data-Physics Dual-Driven Model: Step 13: Divide the dataset from the previous steps into a training set and a test set. The training set is used to train the model, and the test set is used to verify the accuracy of the model. Step 14: Select a suitable convolutional neural network (CNN) and use a multilayer perceptron (MLP) to process the physical parameters, and process the relationship between asphalt physical parameters, skeleton structure images and macroscopic properties of asphalt mixtures; Step 15: Validate and compare the models to determine their goodness of fit. R 2 As an indicator; Step 16: Based on the comparison results of the previous steps, select the best convolutional neural network, build a user interface, and take the skeleton structure image, asphalt grade, modifier type, modifier dosage, fracture characteristics, and rheological characteristics as inputs, and the low-temperature fracture characteristics of asphalt mixture as output.

[0008] Further specifying, in step 1, the SBS doping amount is 4%, 5%, and 6%; the CR doping amount is 18%, 20%, and 22%. Further specifying, in step 2, the temperature is -12℃, -18℃, and -24℃.

[0009] Further specifying, in step 2, the heat preservation time is: 3h, 6h, 12h, or 24h.

[0010] To further define, in step 2, the fracture characteristic refers to the fracture energy obtained based on the SENB test.

[0011] Further specifying, in step 2, the rheological characteristics refer to the stiffness modulus obtained based on the BBR test. S and creep rate m .

[0012] Further specifying, in step 4, the asphalt mixture gradation is AC-16, SMA-16 and OGFC-16.

[0013] Further specifying, in step 5, the image enhancement method is a brightness uniformization method, a histogram equalization enhancement method, and a median filtering noise reduction method.

[0014] Furthermore, in step 5, the image segmentation method is an image segmentation method based on the U-Net network structure, which performs image segmentation training on CT images.

[0015] Further specifying, in step 6, the reconstruction method involves rotating the asphalt mixture specimen around its center every 10°, from the initial position to 170°, resulting in 18 asphalt mixture specimens with different rotation angles, and obtaining longitudinal section images of the asphalt mixture with a fixed slicing angle.

[0016] Further specifying, in step 7, the aggregate is divided into central large particles (9.5~19mm), skeleton particles (4.75~9.5mm), interference particles (1.18~4.75mm) and filler particles (<1.18mm).

[0017] Further specifying, in step 8, the topological index is degree distribution characteristic, isomatch coefficient and average clustering coefficient.

[0018] Further specifying, in step 10, the numerical simulation experiment is a constrained freeze-break test (TSRST), with the initial ambient temperature set at 20℃ and the final temperature at -20℃; the cooling method is a linear cooling method, and the cooling rate is simulated by setting the analysis step duration; To further define the data, the simulated data refers to the indicators for evaluating low-temperature toughness, namely the total energy consumed and the freezing point temperature during the entire test. The total energy is calculated as follows:

[0019] in, Let be the strain energy density function. The critical strain. For stress components, For strain components.

[0020] Further specifying, in step 11, the image processing method is: modifying the original image of size 246×369 to 224×224 by proportional scaling and background filling.

[0021] To further specify, in step 13, the division ratio is 80% of the dataset as the training set and 20% as the validation set.

[0022] Further specifying, in step 15, the... R 2 for

[0023] In the formula: This represents the true value of the low-temperature toughness performance (freezing temperature) of the asphalt mixture. This represents the predicted low-temperature toughness (freezing temperature) of asphalt mixtures. It is the average value of the data.

[0024] The present invention can also replace steps 14 and 15 with the following operations: Step 14: Using 2DCNN+MLP as the network structure, train and test the divided dataset; Step 15: Validate the model, using the true values ​​of the low-temperature fracture characteristics in the validation set as the independent variable and the model predictions as the dependent variable. Analyze the performance of the 2DCNN+MLP model, and the fitting results are as follows: R 2 =0.913 The present invention can also replace steps 14, 15 and 16 with the following operation: Step 14, using ResNet50+ MLP as the network structure, train and test the divided dataset; Step 15: Validate the model using the true values ​​of the low-temperature fracture characteristics in the validation set as the independent variable and the model predictions as the dependent variable. Analyze the performance of the ResNet50+MLP model. The fitting results are as follows: R 2 =0.942; Step 16: The convolutional neural network is a ResNet50+MLP network structure.

[0025] Compared with the prior art, the present invention has the following beneficial effects: This invention enables intelligent research and development of asphalt-based materials, overcomes the limitations of traditional design methods such as poor performance and high cost, and significantly shortens the research and development cycle of asphalt pavement materials in cold regions.

[0026] For a deeper understanding of the features and technical content of this invention, please refer to the accompanying detailed description and drawings. It should be noted that the drawings are provided for illustrative purposes only and are not intended to limit the scope of the invention. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the intelligent design method for low-temperature, high-toughness asphalt pavement materials of the present invention. Figure 2 This is a flowchart illustrating the digital model of asphalt mixtures according to the present invention. Detailed Implementation

[0028] The present invention will be described in detail below with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but should not be considered as limiting the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0029] Example 1: The intelligent design method for low-temperature, high-toughness asphalt pavement materials in this example is implemented through the following steps: I. Construction of a multi-scale database for the low-temperature toughness properties of asphalt mixtures: Microstructural images and low-temperature performance of asphalt mixtures. (1) Establishment of a dataset on the low-temperature toughness of asphalt materials: Step 1: Select 70# and 90# base asphalt, and name them BA-1 and BA-2 respectively, to prepare modified asphalt with different styrene-butadiene (SBS) content and different rubber powder (CR) content; The SBS doping amounts are 4%, 5%, and 6%, and are named S-1, S-2, and S-3, respectively. The CR doping amounts are 18%, 20%, and 22%, and are named CR-1, CR-2, and CR-3, respectively. Step 2: Based on the single-sided notched beam bending (SENB) test and the bending beam rheology (BBR) test, the fracture characteristics and rheological characteristics of each asphalt type under different temperature and heat preservation time conditions were tested. The temperatures mentioned are -12℃, -18℃, and -24℃. The heat preservation time refers to the heat preservation time of the asphalt specimens under the aforementioned temperature conditions, which are 3h, 6h, 12h, and 24h, respectively. The fracture characteristics refer to the fracture energy obtained based on the SENB test; The rheological characteristics refer to the stiffness modulus obtained based on the BBR test. S and creep rate m ; Step 3: Establish a dataset with a one-to-one correspondence between asphalt grade, modifier dosage, insulation time, fracture characteristics, and rheological properties, remove outliers from the dataset, and average the dataset to reduce the influence of dimensions.

[0030] (2) Establishment of image dataset of asphalt mixture skeleton structure: Step 4: Prepare asphalt mixtures with different gradations, namely AC, SMA, and OGFC. The asphalt mixture has a gradation of AC-16, SMA-16 and OGFC-16; The gradation of each asphalt mixture consists of its upper limit, median, and lower limit, as shown in the table below: Table 1

[0031] Step 5: Perform cross-sectional scanning of the above-mentioned asphalt mixture using industrial CT, and perform image enhancement and segmentation on the CT images. The image enhancement methods include brightness uniformization, histogram equalization enhancement, and median filtering noise reduction. The image segmentation method is an image segmentation method based on the U-Net network structure, which is used to train image segmentation on CT images; Step 6: Reconstruct the longitudinal section of the asphalt mixture using 3D reconstruction software. That is, rotate the asphalt mixture specimen around the center of the specimen every 10°, from the initial position to 170°, to obtain 18 asphalt mixture specimens with different rotation angles. Obtain the longitudinal section image of the asphalt mixture by fixing the slice angle. Step 7: Based on the function and size of the aggregates, divide the aggregates in the asphalt mixture into sections, extract the microstructural features of the asphalt mixture, and simplify the aggregate contact structure using a ball-and-stick model. The aggregate is divided into central large particles (9.5~19mm), skeleton particles (4.75~9.5mm), interference particles (1.18~4.75mm) and filler particles (<1.18mm). Step 8: Establish a quantitative microstructure characterization method based on topological theory, and statistically analyze the skeleton contact structure characteristics of asphalt mixtures with different gradations; The topological indices are degree distribution characteristics, isomatch coefficient, and average clustering coefficient. Step 9: Based on topological statistical properties, develop a two-dimensional digital model of asphalt mixture and generate skeleton structure images of asphalt mixtures with different gradations and real microstructure.

[0032] (3) Establishment of a dataset on the low-temperature toughness of asphalt mixtures: Step 10: Use Python to perform secondary development on the finite element software Abaqus, conduct low-temperature freeze-thaw numerical simulation tests on the microscale of asphalt mixtures, and obtain simulation data of each sample under different modified asphalt and different gradation combinations.

[0033] The numerical simulation experiment is a constrained freeze-break test (TSRST), with an initial ambient temperature of 20°C and a final temperature of -20°C; the cooling method is a linear cooling method, and the cooling rate is simulated by setting the analysis step duration. Furthermore, the simulated data refers to the indicators for evaluating low-temperature toughness, namely the total energy consumed throughout the entire test and the freezing point. The total energy is calculated as follows:

[0034] in, Let be the strain energy density function. The critical strain. For stress components, For strain components; (4) Construction of cross-scale datasets: Step 11: Use software to annotate the image of the asphalt mixture skeleton structure and ensure that the image is effectively input into the deep learning model; The image processing method is as follows: by proportional scaling and background filling, the original image size of 246×369 is modified to 224×224; Step 12: Construct a multi-scale database of low-temperature toughness of asphalt materials, contact structure images of asphalt mixture skeleton, and low-temperature fracture performance of asphalt mixtures.

[0035] II. Construction of the Data-Physics Dual-Driven Model: Step 13: Divide the dataset from the previous steps into an 80% training set and a 20% test set. The training set is used to train the model, and the test set is used to verify the accuracy of the model. Step 14: Select a suitable convolutional neural network (CNN) and use a multilayer perceptron (MLP) to process the physical parameters, and process the relationship between the asphalt physical parameters, skeleton structure image and macroscopic properties of asphalt mixture.

[0036] Step 15: Validate and compare the models to determine their goodness of fit. R 2 As an indicator; The R 2 for

[0037] In the formula: This represents the true value of the low-temperature toughness performance (freezing temperature) of the asphalt mixture. This represents the predicted low-temperature toughness (freezing temperature) of asphalt mixtures. It is the average value of the data.

[0038] III. Intelligent Design of High-Toughness Asphalt Pavement Materials for Cold Regions: Step 16: Based on the comparison results of the previous steps, select the best convolutional neural network, build a user interface, and take the skeleton structure image, asphalt grade, modifier type, modifier dosage, fracture characteristics, and rheological characteristics as inputs, and the low-temperature fracture characteristics of asphalt mixture as output.

[0039] Example 2: Steps 1 to 13 in this embodiment are the same as in embodiment 1. The remaining steps are as follows: Step 14: Using 2DCNN+MLP as the network structure, train and test the divided dataset; Step 15: Validate the model, using the true values ​​of the low-temperature fracture characteristics in the validation set as the independent variable and the model predictions as the dependent variable. Analyze the performance of the 2DCNN+MLP model, and the fitting results are as follows: R 2 =0.913.

[0040] Example 3: Steps 1 to 13 in this embodiment are the same as in embodiment 1. The remaining steps are as follows: Step 14: Using ResNet50+ MLP as the network structure, train and test the divided dataset; Step 15: Validate the model using the true values ​​of the low-temperature fracture characteristics in the validation set as the independent variable and the model predictions as the dependent variable. Analyze the performance of the ResNet50+MLP model. The fitting results are as follows: R 2 =0.942; Step 16: Based on the comparison results of the previous steps, the ResNet50+MLP network structure is selected, and a user interface is built. The skeleton structure image, asphalt grade, modifier type, modifier dosage, fracture characteristics, and rheological characteristics are used as inputs, and the low-temperature fracture characteristics of asphalt mixture are used as outputs.

[0041] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0042] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A smart design method for low-temperature, high-toughness asphalt pavement materials, characterized in that, Includes the following steps: Step 1: Select base asphalt of different grades to prepare modified asphalt with different styrene-butadiene (SBS) content and different rubber powder (CR) content; Step 2: Based on the single-sided notched beam bending SENB test and bending beam rheology BBR test, test the fracture characteristics and rheological characteristics of each asphalt grade under different temperature and heat preservation time conditions; Step 3: Establish a dataset with a one-to-one correspondence between asphalt grade, modifier type, modifier dosage, insulation time, fracture characteristics, and rheological characteristics; remove outlier data from the dataset; and mean the dataset to reduce the influence of dimensions. Step 4: Prepare asphalt mixtures with different gradations, namely AC, SMA, and OGFC, according to the specifications. Step 5: Use industrial CT to perform cross-sectional scanning of the above asphalt mixture, and perform image enhancement and segmentation on the CT images; Step 6: Reconstruct the longitudinal section of the asphalt mixture using 3D reconstruction software; Step 7: Divide the aggregates in the asphalt mixture according to their function and size, extract the microstructural features of the asphalt mixture, and simplify the aggregate contact structure using a ball-and-stick model. Step 8: Establish a quantitative microstructure characterization method based on topological theory, and statistically analyze the skeleton contact structure characteristics of asphalt mixtures with different gradations; Step 9: Based on topological statistical properties, develop a two-dimensional digital model of asphalt mixture and generate images of the skeleton structure of asphalt mixture with real microstructure for different gradations. Step 10: Use Python to perform secondary development on the finite element software Abaqus, conduct low-temperature freezing fracture numerical simulation tests on the microscale of asphalt mixtures, and obtain simulation data of each sample under different modified asphalt and different gradation combinations. Step 11: Use software to annotate the image of the asphalt mixture skeleton structure and ensure that the image is effectively input into the deep learning model; Step 12: Construct a cross-scale database of low-temperature toughness of asphalt materials, contact structure images of asphalt mixture skeleton, and low-temperature fracture performance of asphalt mixtures. Step 13: Divide the dataset from the previous steps into a training set and a test set. The training set is used to train the model, and the test set is used to verify the accuracy of the model. Step 14: Select a suitable convolutional neural network (CNN) and use a multilayer perceptron (MLP) to process physical parameters, and process the relationship between asphalt physical parameters, skeleton structure images and macroscopic properties of asphalt mixtures. Step 15: Validate and compare the models to determine their goodness of fit. R 2 As an indicator; Step 16: Based on the comparison results of the previous steps, select the best convolutional neural network, build a user interface, and take the skeleton structure image, asphalt grade, modifier type, modifier dosage, fracture characteristics, and rheological characteristics as inputs, and the low-temperature fracture characteristics of asphalt mixture as output.

2. The method according to claim 1, characterized in that, In step 1, the SBS doping levels are 4%, 5%, and 6%; the CR doping levels are 18%, 20%, and 22%. In step 2, the temperatures are -12℃, -18℃, and -24℃; The heat preservation time is: 3h, 6h, 12h, 24h; The fracture characteristics refer to the fracture energy obtained based on the SENB test; The rheological characteristics refer to the stiffness modulus obtained based on the BBR test. S and creep rate m .

3. The method according to claim 1, characterized in that, In step 4, the asphalt mixture gradation is AC-16, SMA-16 and OGFC-16; in step 5, the image enhancement method is brightness uniformization, histogram equalization enhancement and median filtering noise reduction; the image segmentation method is an image segmentation method based on U-Net network structure, which performs image segmentation training on CT images.

4. The method according to claim 1, characterized in that, In step 6, the reconstruction method involves rotating the asphalt mixture specimen around its center every 10°, from the initial position to 170°, resulting in 18 asphalt mixture specimens with different rotation angles. The longitudinal section image of the asphalt mixture is then obtained by fixing the slicing angle.

5. The method according to claim 1, characterized in that, In step 7, the aggregate is divided into large central particles, skeleton particles, interference particles, and filler particles.

6. The method according to claim 1, characterized in that, In step 8, the topological indices are degree distribution characteristics, isomatch coefficient, and average clustering coefficient.

7. The method according to claim 1, characterized in that, In step 10, the numerical simulation test is the constrained freeze-break test (TSRST), with the initial ambient temperature set at 20°C and the final temperature at -20°C; the cooling method is linear cooling, and the cooling rate is simulated by setting the analysis step duration. Furthermore, the simulated data refers to the indicators for evaluating low-temperature toughness, namely the total energy consumed throughout the entire test and the freezing point. The total energy is calculated as follows: in, Let be the strain energy density function. The critical strain. For stress components, For strain components.

8. The method according to claim 1, characterized in that, In step 11, the image processing method is as follows: by scaling proportionally and filling the background, the image with the original size of 246×369 is modified to an image with the size of 224×224. In step 13, the dataset is divided into 80% training set and 20% validation set. In step 15, the R 2 for In the formula: This represents the true value of the freeze-thaw temperature, which is the low-temperature toughness property of asphalt mixtures. This refers to the predicted freezing-break temperature value for the low-temperature toughness properties of asphalt mixtures. It is the average value of the data.

9. The method according to claim 1, characterized in that, Replace steps 14 and 15 with the following operations: Step 14: Train and test the divided dataset using 2DCNN+MLP as the network structure; Step 15: Validate the model, using the true values ​​of the low-temperature fracture characteristics in the validation set as the independent variable and the model predictions as the dependent variable. Analyze the performance of the 2DCNN+MLP model, and the fitting results are as follows: R 2 =0.

913.

10. The method according to claim 1, characterized in that, Replace steps 14, 15, and 16 with the following operations: Step 14: Train and test the divided dataset using ResNet50+ MLP as the network structure; Step 15: Validate the model using the true values ​​of the low-temperature fracture characteristics in the validation set as the independent variable and the model predictions as the dependent variable. Analyze the performance of the ResNet50+MLP model. The fitting results are as follows: R 2 =0.942; Step 16: The convolutional neural network is a ResNet50+MLP network structure.

Images