Stable rubber asphalt material analysis method

Abstract

The application discloses a kind of stable rubber asphalt material analysis methods, it is related to traffic transport industry asphalt material and data analysis technical field, the method includes: first group high flux test is carried out to stable rubber asphalt material, to obtain the cross-scale characteristic data of the stable rubber asphalt material;Second group high flux test is carried out to the stable rubber asphalt material, to obtain the digital image data of the stable rubber asphalt material;According to the cross-scale characteristic data and digital image data, extract the relationship model between structure and performance in the stable rubber asphalt material;The relationship model is verified, and the relationship model is optimized according to verification result;According to the relationship model after optimization, stable rubber asphalt material is analyzed.The application can be more comprehensive and systematic by the cross and fusion of multi-field technology to build the key problems of high flux characterization and analysis of stable rubber asphalt material.

Description

Technical Field

[0001] This invention relates to the field of asphalt materials and data analysis technology in the transportation industry, and in particular to a method for analyzing stable rubber asphalt materials. Background Technology

[0002] Asphalt materials are characterized by their cross-scale nature, diverse composition, complex structure, and variable service environment. Their performance is influenced by the multi-level material genes of "composition-structure-process-performance" across the "macro-fine-micro-nano" scale. The traditional "experiment-guided experiment" material development model is difficult to meet the needs of higher-level material design.

[0003] In recent years, domestic and international research has also been conducted on high-throughput testing of materials. For example, China University of Petroleum and Michigan Technological University have used nuclear magnetic resonance (NMR) and small-angle scattering (SAS) techniques to measure the molecular composition and structure of asphalt and cement; the Highway Research Institute of the Ministry of Transport and Imperial College London have used X-ray diffraction to measure the mineral composition and morphological characteristics of aggregates; Hefei University of Technology and the University of California have used industrial CT scanning to characterize the microstructure, spatial distribution, and other characteristic parameters of materials; China University of Petroleum, Southeast University, Harbin Institute of Technology, and the University of Missouri have established databases for modulus, strength, and other parameters for different asphalt materials and asphalt concrete, and proposed material performance prediction models; Shanghai Jiao Tong University and the University of Technology Sydney have also conducted research on the single properties of cement concrete. Therefore, high-throughput digital characterization and intelligent design of transportation infrastructure materials represent a new research and development model for the "theoretical prediction and experimental verification" of future transportation construction and maintenance materials.

[0004] Stabilized rubber asphalt is a high-performance road material produced by compounding and modifying waste tire rubber powder with base asphalt under specific conditions. It solves problems that traditional rubber asphalt products cannot address, such as "unstable storage" and "strong odor," fully embodying the characteristics of low carbon, green, and environmentally friendly properties. The application of this product reduces engineering costs and extends the service life of road surfaces.

[0005] However, current research on stabilized rubber asphalt mainly suffers from problems such as "low efficiency of traditional experimental methods and unclear characteristic parameters, making digital characterization difficult." Therefore, it is necessary to develop an analytical method for stabilized rubber asphalt. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide an analytical method for stable rubber asphalt materials, which can realize high-throughput characterization and analysis of stable rubber asphalt materials.

[0007] To address the aforementioned technical problems, this invention provides a method for analyzing stable rubber asphalt materials, comprising: conducting a first set of high-throughput tests on the stable rubber asphalt material to obtain cross-scale characteristic data of the stable rubber asphalt material; conducting a second set of high-throughput tests on the stable rubber asphalt material to obtain digital image data of the stable rubber asphalt material; extracting a relationship model between structure and properties in the stable rubber asphalt material based on the cross-scale characteristic data and the digital image data; verifying the relationship model and optimizing the relationship model based on the verification results; and analyzing the stable rubber asphalt material based on the optimized relationship model.

[0008] As an improvement to the above scheme, the step of conducting a first set of high-throughput tests on the stabilized rubber asphalt material to obtain cross-scale characteristic data of the stabilized rubber asphalt material includes: collecting the mesoscopic and microscopic characteristic parameters of the stabilized rubber asphalt material using a chromatography-mass spectrometry system to characterize the asphalt composition of the stabilized rubber asphalt material; collecting the nanoscopic characteristic parameters of the stabilized rubber asphalt material using an infrared scanner and nuclear magnetic resonance spectrometer to characterize the asphalt structure of the stabilized rubber asphalt material; and collecting the macroscopic, mesoscopic, and microscopic characteristic parameters of the stabilized rubber asphalt material using a CT scanner to characterize the asphalt properties of the stabilized rubber asphalt material.

[0009] As an improvement to the above scheme, the steps of collecting the microscopic and macroscopic characteristic parameters of the stable rubber asphalt material by chromatography-mass spectrometry include: analyzing volatile products by Py-GC-MS technology to generate a spectrum; selectively extracting polar compounds generated during aging from the spectrum by supercritical fluid extraction technology; and processing the spectrum of the extracted polar compounds by a pre-trained convolutional neural network to identify overlapping peaks.

[0010] As an improvement to the above scheme, the step of acquiring the nanoscopic characteristic parameters of the stabilized rubber asphalt material using an infrared scanner includes: uniformly mixing embrittled rubber particles with asphalt that has lost its stickiness to form a sample of the stabilized rubber asphalt material using low-temperature cryogenic pulverization technology; performing infrared spectral testing on the sample using ATR-FTIR technology to generate an infrared spectrum; separating overlapping peaks in the infrared spectrum using second-derivative spectroscopy; constructing a rubber asphalt spectral library to record standard spectra of each component and aging products in the stabilized rubber asphalt material; analyzing the dynamic spectral changes of the stabilized rubber asphalt material under different temperature gradients using two-dimensional correlation spectroscopy; and comparing the dynamic spectral changes with the rubber asphalt spectral library to identify the aging-sensitive bands of the stabilized rubber asphalt material.

[0011] As an improvement to the above scheme, the step of acquiring the nanoscopic characteristic parameters of the stabilized rubber asphalt material using a nuclear magnetic resonance (NMR) spectrometer includes: analyzing the aging degree of the asphalt binder based on the transverse relaxation time distribution using low-field NMR technology; detecting the sulfur crosslinking network density using multi-quantum NMR technology; and further... Weighted imaging technology was used to characterize the interfacial compatibility between the rigid and flexible phases in SBS modified asphalt; a high-temperature and high-pressure in-situ cavity was constructed and shear rates were applied simultaneously to simulate the asphalt pumping process; and molecular dynamics information was measured using fast-field cyclic nuclear magnetic resonance technology.

[0012] As an improvement to the above scheme, the step of acquiring the macroscopic, mesoscopic, and microscopic characteristic parameters of the stable rubber asphalt material using a CT scanner includes: obtaining the volume fraction of rubber particles in asphalt using high-energy microfocus CT; capturing the deformation state of rubber particles under temperature cycling using a dynamic scanning mode; quantifying the three-dimensional deformation of rubber particles under temperature cycling based on the volume fraction and deformation state using CT three-dimensional reconstruction technology to generate 3D data; and segmenting the 3D data using a pre-trained deep learning segmentation model to identify the rubber phase and the asphalt phase.

[0013] As an improvement to the above scheme, the step of performing a second set of high-throughput tests on the stabilized rubber asphalt material to obtain digital image data of the stabilized rubber asphalt material includes: performing a second set of high-throughput tests on the stabilized rubber asphalt material to acquire cross-scale feature images of the stabilized rubber asphalt material; and performing denoising, data augmentation, and feature structure association processing on the cross-scale feature images to obtain digital image data characterizing the structural features of the stabilized rubber asphalt material.

[0014] As an improvement to the above scheme, the step of extracting the relationship model between structure and performance in the stable rubber asphalt material based on the cross-scale feature data and digital image data includes: classifying the cross-scale feature data and digital image data according to computer vision technology and multimodal learning technology to generate classification results; and constructing the relationship model between structure and performance in the stable rubber asphalt material based on the classification results according to multimodal learning technology, semantic segmentation technology and object detection technology.

[0015] As an improvement to the above scheme, the step of constructing a model of the relationship between structure and properties in the stable rubber asphalt material based on the classification results includes: associating structural parameters with molecular mobility to establish a structure-rheological property prediction model.

[0016] As an improvement to the above scheme, the method for analyzing stable rubber asphalt materials further includes: optimizing the testing process based on the optimized relationship model.

[0017] This invention provides an analytical method for stable rubber asphalt materials that encompasses technologies from multiple fields, including materials science, mechanics, physics, chemistry, testing, algorithms, and software. Through the intersection and integration of these technologies, it comprehensively and systematically addresses the key issues of high-throughput characterization and analysis of stable rubber asphalt materials. Accordingly, implementing this invention offers the following beneficial effects: (1) For high-throughput digital characterization, the use of big data-based multi-level cross-scale experimental and simulation collaborative technology for materials can effectively shorten the material experiment cycle; (2) Based on the data intelligent mining algorithm, the integrity of material characteristic parameter data can be significantly improved; (3) Based on the typical performance gene data of materials, the correlation analysis theory can be used to select characteristic genes of stable rubber asphalt, identify gene units and classify them. (4) Based on the key factor analysis algorithm, it is possible to determine the control factors of typical material properties. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating an embodiment of the analytical method for stabilized rubber asphalt materials of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. It is hereby declared that the directional terms such as up, down, left, right, front, back, inside, and outside used in this text are based solely on the accompanying drawings and are not intended to specifically limit the invention.

[0020] See Figure 1 , Figure 1 The flowchart illustrating an embodiment of the present invention's method for analyzing stable rubber asphalt materials includes: S101, the first set of high-throughput tests was conducted on the stabilized rubber asphalt material to obtain cross-scale characteristic data of the stabilized rubber asphalt material; It should be noted that high-throughput testing can be used to obtain characteristic parameters of stable rubber asphalt materials at macroscopic, mesoscopic, microscopic, and nanoscopic scales, and to form a multi-level cross-scale characteristic parameter set from the perspectives of material composition, material structure, and material properties, so as to characterize stable rubber asphalt materials.

[0021] Accordingly, the steps for conducting the first set of high-throughput tests on the stabilized rubber asphalt materials to obtain cross-scale characteristic data of the stabilized rubber asphalt materials include: (1) The microscopic and macroscopic characteristic parameters of the stabilized rubber asphalt material were collected by chromatography-mass spectrometry to characterize the asphalt composition of the stabilized rubber asphalt material; For the fine and microscopic characteristic parameters of the four components, molecular weight, and distribution of stabilized rubber asphalt materials, high-throughput characterization of asphalt components can be achieved based on the basic principles of chromatography. Specific steps include: (1.1) Volatile products were analyzed using Py-GC-MS technology to generate spectra; (1.2) Polar compounds generated during aging were selectively extracted from the spectrum using supercritical fluid extraction technology; Because high molecular weight components (such as SBS and rubber powder) in rubber asphalt are difficult to vaporize, GC separation fails. Therefore, this invention introduces Py-GC-MS to analyze volatile products (such as styrene monomers and isoprene fragments) after pyrolysis at 600℃; simultaneously, supercritical fluid extraction (SFE-CO2) is used to selectively extract polar compounds (such as ketones and carbonyl compounds) generated during aging; thus, the rubber powder content can be quantitatively determined by Py-GC-MS fingerprinting (characteristic peak: isoprene derivative m / z=68).

[0022] (1.3) The spectra of the extracted polar compounds are processed by a pre-trained convolutional neural network to identify overlapping peaks.

[0023] Since the analysis of complex mixture spectra relies on human experience, it is inefficient. Therefore, this invention constructs a database of characteristic ions for rubber asphalt containing markers such as sulfides (m / z=64) and antioxidants (e.g., Irganox 1010, m / z=1177), and applies a convolutional neural network (CNN) to automatically identify overlapping peaks (e.g., phthalate plasticizers and asphalt oxidation products).

[0024] (2) Nanoscale characteristic parameters of stable rubber asphalt materials were collected by infrared scanner and nuclear magnetic resonance spectrometer to characterize the asphalt structure of stable rubber asphalt materials; For nanoscale characteristic parameters such as molecular functional groups, element types, and content distribution of stabilized rubber asphalt materials, high-throughput characterization of asphalt structural parameters can be developed based on the fundamental theory of spectroscopic testing. Specifically: The steps for acquiring nanoscopic characteristic parameters of stabilized rubber asphalt materials using an infrared scanner include: (2.1.1) The brittle rubber particles and the asphalt that has lost its stickiness are uniformly mixed by low-temperature cryogenic pulverization technology (-196℃ liquid nitrogen environment) to form a sample of stable rubber asphalt material; (2.1.2) The infrared spectrum of the sample was measured using ATR-FTIR technology to generate an infrared spectrum; (2.1.3) Separate overlapping peaks in the infrared spectrum using the second derivative spectroscopy method; The high viscosity of stabilized rubber-modified asphalt materials makes transmission mode sample preparation difficult, and the uneven distribution of rubber particles may obscure characteristic peaks. Therefore, this invention utilizes cryogenic pulverization technology to embrittle rubber particles and then uniformly mix them with asphalt; simultaneously, an attenuated total reflectance (ATR) attachment is used to directly test the surface of the untreated sample, avoiding slicing errors; thus, combined with second-derivative spectroscopy, styrene (~699 cm⁻¹) in rubber-modified composite asphalt can be separated. -1 ) and asphalt (~1600cm) -1 The overlapping peaks of ).

[0025] (2.1.4) Construct a rubber asphalt spectral library to record the standard spectra of each component and aging products in stable rubber asphalt materials; (2.1.5) The dynamic spectral changes of the stabilized rubber asphalt material under different temperature gradients were analyzed by two-dimensional correlation spectroscopy (2D-COS); (2.1.6) Compare the dynamic spectrum changes with the rubber asphalt spectral library to identify the aging-sensitive bands of stable rubber asphalt materials.

[0026] Because the spectral peaks of oxidation products (such as carbonyl and sulfoxide groups) and rubber degradation products during the aging process are confused, this invention constructs a standard spectrum containing SBS, rubber powder, matrix asphalt, and their aging products; and uses two-dimensional correlation spectroscopy to analyze the dynamic spectral changes under temperature gradients to identify aging-sensitive bands.

[0027] Meanwhile, the steps for acquiring nanoscopic characteristic parameters of stabilized rubber asphalt materials using nuclear magnetic resonance imaging include: (2.2.1) The aging degree of asphalt binder was analyzed based on the transverse relaxation time distribution using low-field nuclear magnetic resonance (LF-NMR, 0.5~2MHz); (2.2.2) The density of sulfur crosslinking network (intersection between crosslinking points <10 nm) was detected by multi-quantum nuclear magnetic resonance technology. (2.2.3) Through Weighted imaging technology characterizes the interfacial compatibility between the rigid and flexible phases in SBS modified bitumen; The restricted movement of polymer chains in rubberized asphalt results in weak conventional NMR signals. Therefore, this invention employs low-field portable NMR to analyze the aging degree of asphalt binder through transverse relaxation time distribution; simultaneously, a multi-quantum coherence (MQ-NMR) pulse sequence is developed to detect the density of the sulfur crosslinking network; thereby enabling the analysis of the aging degree of asphalt binder through... Weighted imaging characterizes the interfacial compatibility between styrene blocks (rigid phase) and butadiene blocks (flexible phase) in SBS-modified asphalt.

[0028] (2.2.4) Construct a high-temperature and high-pressure in-situ cavity and simultaneously apply shear rates (0.1~100s). -1 To simulate the asphalt pumping process; (2.2.5) Molecular dynamics information was measured using fast field cyclic nuclear magnetic resonance (NMR) technology.

[0029] Traditional nuclear magnetic resonance (NMR) cannot monitor molecular motion under high-temperature (>100℃) flow dynamics in real time. Therefore, this invention designs a high-temperature, high-pressure in-situ cavity to simultaneously apply shear rates to simulate the asphalt pumping process; at the same time, it combines fast field cycling (FFC) NMR to study the molecular dynamics transition under temperature-stress coupling.

[0030] (3) Macroscopic, mesoscopic and microscopic characteristic parameters of stabilized rubber asphalt materials are collected by CT scanner to characterize the asphalt properties of stabilized rubber asphalt materials.

[0031] For macroscopic characteristic parameters such as aggregate spatial distribution and skeleton structure, microscopic porosity parameters such as asphalt saturation and pore volume size distribution (i.e., microscopic characteristic parameters), and microscopic structural parameters such as asphalt film thickness and interface structure (i.e., microscopic characteristic parameters) of stabilized rubber asphalt materials, high-throughput characterization of asphalt performance parameters of stabilized rubber asphalt materials can be constructed based on X-ray cross-sectional scanning imaging and three-dimensional reconstruction technology. Specific steps include: (3.1) The volume fraction of rubber particles in asphalt was obtained by high-energy microfocus CT (voltage > 200 kV, resolution < 1 μm); (3.2) Capture the deformation state of rubber particles under temperature cycling (-20℃~160℃) using dynamic scanning mode; (3.3) The three-dimensional deformation of rubber particles under temperature cycling is quantified based on volume fraction and deformation state using CT three-dimensional reconstruction technology to generate 3D data; Due to the high density of rubber asphalt (containing SBS or rubber powder), X-ray penetration is low, resulting in insufficient resolution. Therefore, this invention upgrades high-energy microfocus CT to penetrate the sulfur-crosslinked rubber phase; simultaneously, a dynamic scanning mode is used to capture the expansion / contraction behavior of rubber particles under temperature cycling; thus, the volume fraction and spatial distribution uniformity of rubber particles in asphalt are quantified through 3D reconstruction using CT (using Minkowski functional analysis).

[0032] (3.4) The 3D data is segmented using a pre-trained deep learning segmentation model to identify the rubber phase and the asphalt phase.

[0033] Because the rubber phase and the asphalt phase have similar densities, their grayscale contrast is low. Therefore, in constructing the deep learning segmentation model in this invention, nanoscale contrast agents (bismuth iodide nanoparticles) are injected to selectively label the rubber phase; wherein, the deep learning segmentation model adopts the U-Net++ architecture, and the training dataset contains manually labeled rubber / asphalt phase boundaries.

[0034] Therefore, by using testing methods such as infrared scanners, CT scanners, nuclear magnetic resonance spectrometers, and chromatography-mass spectrometry, a multi-level, multi-scale characteristic parameter set of stable rubber asphalt materials can be obtained, and a high-throughput experimental and characterization technique for stable rubber asphalt materials can be established.

[0035] S102, A second set of high-throughput tests was conducted on the stabilized rubber asphalt material to obtain digital image data of the stabilized rubber asphalt material; Accordingly, the steps for conducting a second set of high-throughput tests on the stabilized rubber asphalt material to obtain digital image data of the stabilized rubber asphalt material include: (1) A second set of high-throughput tests was conducted on the stabilized rubber asphalt material to collect cross-scale characteristic images of the stabilized rubber asphalt material; When conducting the second set of high-throughput tests, cross-scale characteristic images of the material composition and structural features of stable rubber asphalt materials can be obtained through techniques such as photoelectric microscopy, CT non-destructive scanning, and infrared spectroscopy.

[0036] (2) Denoising, data enhancement and feature structure association processing are performed on the cross-scale feature images to obtain digital image data that characterizes the structural features of stable rubber asphalt materials.

[0037] Accordingly, image denoising, data augmentation, and feature structure association can be used to process cross-scale feature images to obtain high-quality digital samples (i.e., digital image data) that characterize the structural features of stable rubber asphalt materials.

[0038] The following sections provide further detailed descriptions of image denoising, data augmentation, and feature structure association: I. Image Denoising On the one hand, the sources of noise in microscopic images of rubber asphalt (such as SEM and industrial CT) are complex, mainly including inherent material noise (X-ray scattering noise caused by the density difference between rubber particles and asphalt matrix) and equipment noise (motion artifacts when scanning high-viscosity samples, such as trailing shadows in dynamic CT scans).

[0039] Therefore, this invention introduces multimodal noise modeling. Specifically, a noise hybrid model is constructed by combining a composite denoising algorithm of Gaussian noise (equipment noise) and Poisson noise (low photon count noise); at the same time, a material-adaptive deep learning denoising network is developed; in practical applications, for industrial CT images, nonlocal mean filtering (NLM) combined with deep learning denoising is used to improve the accuracy of pore identification to 92%.

[0040] On the other hand, general denoising algorithms may destroy key structural features of materials (such as rubber particle boundaries and microcracks).

[0041] Therefore, this invention introduces physical constraints. Specifically, morphological constraints are added to the loss function, that is, discontinuous particle boundaries are penalized through corrosion-expansion operations; at the same time, prior material knowledge is used to pre-label the size distribution range of rubber particles to constrain the physical rationality of the denoised structure.

[0042] II. Data Augmentation On the one hand, traditional data augmentation (rotation / flipping) cannot reflect real material deformation (such as shear-induced rubber particle orientation).

[0043] Therefore, this invention introduces enhancement based on a physical model. Specifically, it implements finite element simulation enhancement, that is, it uses COMSOL to simulate the distribution of rubber particles under different process parameters (temperature, shear rate) to generate synthetic images; at the same time, it designs a Physics-informed GAN, using rheological equations (such as the Cross model) as generator constraints to ensure that the enhanced data conforms to actual rheological behavior; in practical applications, CycleGAN generates FTIR spectral-microscopic image pairs with different aging degrees (0%~20% oxidation).

[0044] On the other hand, single-modal data augmentation is difficult to cover multi-scale features (such as the correlation between molecular crosslinking degree and macroscopic mechanical properties).

[0045] Therefore, this invention introduces cross-modal enhancement. Specifically, it achieves multimodal joint enhancement, that is, simultaneously generating CT images (mesoscopic structure) and NMR relaxation time distributions (molecular motion), learning the joint distribution of CT-NMR through a variational autoencoder (VAE), and sampling to generate new data; at the same time, it achieves knowledge graph-guided enhancement, that is, generating data combinations that conform to chemical rules based on a material ontology library (such as the relationship between rubber type and glass transition temperature).

[0046] III. Feature Structure Association On the one hand, traditional characteristics (such as porosity) cannot describe the chemical bonding state of the rubber-asphalt interface.

[0047] Therefore, this invention introduces multi-scale feature extraction. Specifically, it achieves cross-scale feature fusion: Microscopic: The distribution of interfacial adhesion forces (nanoscale) was extracted using atomic force microscopy (AFM).

[0048] Mesoscopic: Modeling the network topology of rubber particles (micrometer scale) using Graph Neural Networks (GNN).

[0049] Macroscopic: Combining dynamic shear rheology (DSR) data (millimeter scale).

[0050] For example, regression analysis was performed on the number of particle contact points output by GNN and the complex modulus of DSR (R²>0.85).

[0051] On the other hand, static feature association ignores time-varying effects during aging / service.

[0052] Therefore, this invention introduces dynamic correlation modeling. Specifically, a temporal graph convolutional network (T-GCN) is constructed to model the dynamic correlation between rubber particle aggregation and performance degradation during the aging process; simultaneously, a physical information neural network (PINN) is constructed, in which aging dynamics equations (such as the Arrhenius equation) are embedded in the loss function to improve extrapolation reliability.

[0053] S103, Based on cross-scale feature data and digital image data, extract the relationship model between structure and performance in stable rubber asphalt materials; Step S103 is a crucial step in data mining, with its core objective being to extract implicit physicochemical features and correlations from multimodal data such as images, text, spectra, and experimental reports. Specific steps include: (1) Based on computer vision technology and multimodal learning technology, classify cross-scale feature data and digital image data to generate classification results; Accordingly, computer vision techniques include computer vision feature extraction, for example: Traditional characteristics: -Texture features: Grain orientation entropy is extracted from the gray-level co-occurrence matrix (GLCM); - Morphological characteristics: grain size distribution (Feret diameter), porosity calculation.

[0054] Deep learning characteristics: - Self-supervised pre-training: The SimCLR framework is used to learn the representation vectors of microscopic images; -Key region detection: Based on Grad-CAM visualization of attention regions (such as crack initiation points).

[0055] (2) Based on multimodal learning technology, semantic segmentation technology and target detection technology, a relationship model between structure and performance in stable rubber asphalt materials is constructed based on the classification results.

[0056] Correspondingly, semantic segmentation techniques include text semantic feature extraction, such as fine-tuning the SciBERT model to identify material synthesis steps (e.g., the "sol-gel method"); and constructing a material failure mode ontology library (fatigue, creep, stress corrosion).

[0057] In practical applications, various relationship models can be constructed to achieve cross-scale correlation analysis of multimodal data.

[0058] For example, structural parameters can be correlated with molecular mobility to establish a structure-rheological property prediction model. This involves correlating structural parameters (porosity, rubber distribution) obtained by CT scans with molecular mobility (T2 distribution peak width) measured by NMR to establish a structure-rheological property prediction model.

[0059] Therefore, based on the theoretical foundations of computer vision technology, multimodal learning framework, semantic segmentation technology and object detection technology, data mining algorithms centered on cross-scale feature data and digital image data can be established.

[0060] Furthermore, based on the relationship model between structure and performance in stabilized rubber asphalt materials, high-throughput tests can be dynamically supplemented and updated.

[0061] S104, validate the relational model and optimize the relational model based on the validation results; It should be noted that by studying the correlation between structure and performance obtained from data mining, extracting confidence intervals, and combining reliability testing methods such as manual verification, experimental verification, and repeated testing, the data mining algorithm in step S103 can be verified, improving the integrity and reliability of cross-scale feature data and digital image data. At the same time, a feedback correction mechanism for the data mining algorithm can be established to continuously optimize the unstructured data mining algorithm.

[0062] Specifically, for experimental data on stabilized rubber asphalt materials, mathematical methods such as simple random sampling, stratified random sampling, and systematic sampling can be used to establish sampling methods for material characteristic parameters, extract small samples of multi-level, cross-scale data, conduct corresponding material experiments, and verify the effectiveness, reliability, and stability of high-throughput experimental techniques. At the same time, a feedback correction mechanism for high-throughput experiments can be established to address the errors between high-throughput experimental data and verification data, and to continuously optimize high-throughput experimental and characterization techniques. For example, SHAP value analysis can identify key structural features that affect rutting resistance (such as a 40% increase in contribution when the aspect ratio of rubber particles is >2.5).

[0063] For example, the correlation between the crosslinking density measured by NMR and the interface area calculated by image analysis was compared (Pearson coefficient > 0.9).

[0064] S105, Analysis of stable rubber asphalt materials based on the optimized relationship model.

[0065] Specific analytical applications include: (1) Prediction of aging lifespan: For example, by fusing denoised microscopic images with time-series FTIR data, the error in predicting the life of rubber asphalt pavement is less than 8%.

[0066] (2) Root cause analysis of defects: For example, feature correlation revealed that the pore aggregation area (>5μm) in CT images corresponds to an abnormal concentration of volatile substances detected by GC-MS (styrene loss rate >12%).

[0067] Furthermore, the testing process can be optimized based on the optimized relationship model.

[0068] For example, by training a GAN model based on augmented data, the optimal rubber dispersion structure can be generated, which then guides the processing temperature to be reduced from 200°C to 185°C, resulting in a 15% reduction in energy consumption. For example, based on industrial CT monitoring of rubber particle dispersion, data enhancement can generate optimized process parameters, which are then fed back to the colloid mill in the processing unit to adjust the shear rate and temperature.

[0069] Therefore, this invention optimizes the production / testing process by constructing a closed loop of "image denoising -> data enhancement -> feature association -> process feedback"; it can also significantly improve the accuracy and efficiency of rubber asphalt in microstructure characterization, performance prediction and process optimization, providing reliable support for the design of high-durability pavement materials.

[0070] In summary, the analytical method for stabilized rubber asphalt materials of this invention encompasses technologies from multiple fields, including materials science, mechanics, physics, chemistry, testing, algorithms, and software. Through the intersection and integration of these technologies, it comprehensively and systematically addresses key issues in the high-throughput characterization and design of stabilized rubber asphalt materials. Accordingly, implementing this invention has the following beneficial effects: (1) For high-throughput digital characterization, the use of big data-based multi-level cross-scale experimental and simulation collaborative technology for materials can effectively shorten the material experiment cycle; (2) Based on the data intelligent mining algorithm, the integrity of material characteristic parameter data can be significantly improved; (3) Based on the typical performance gene data of materials, the correlation analysis theory can be used to select characteristic genes of stable rubber asphalt, identify gene units and classify them. (4) Based on the key factor analysis algorithm, it is possible to determine the control factors of typical material properties.

[0071] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for analyzing stable rubber asphalt materials, characterized in that, include: The first set of high-throughput tests was conducted on the stabilized rubber asphalt material to obtain cross-scale characteristic data of the stabilized rubber asphalt material; A second set of high-throughput tests was conducted on the stabilized rubber asphalt material to obtain digital image data of the stabilized rubber asphalt material; Based on the cross-scale feature data and digital image data, a model of the relationship between structure and performance in the stable rubber asphalt material is extracted; Validate the relation model and optimize the relation model based on the validation results; The stable rubber asphalt material was analyzed based on the optimized relationship model. The step of performing a first set of high-throughput tests on the stabilized rubber asphalt material to obtain cross-scale characteristic data of the stabilized rubber asphalt material includes: collecting the mesoscopic and microscopic characteristic parameters of the stabilized rubber asphalt material using a chromatography-mass spectrometry system to characterize the asphalt composition of the stabilized rubber asphalt material; collecting the nanoscopic characteristic parameters of the stabilized rubber asphalt material using an infrared scanner and nuclear magnetic resonance spectrometer to characterize the asphalt structure of the stabilized rubber asphalt material; and collecting the macroscopic, mesoscopic, and microscopic characteristic parameters of the stabilized rubber asphalt material using a CT scanner to characterize the asphalt properties of the stabilized rubber asphalt material. The step of acquiring nanoscopic characteristic parameters of the stable rubber asphalt material using a nuclear magnetic resonance (NMR) spectrometer includes: analyzing the aging degree of the asphalt binder based on the transverse relaxation time distribution using low-field NMR technology; detecting the sulfur crosslinking network density using multi-quantum NMR technology; and further... Weighted imaging technology was used to characterize the interfacial compatibility between the rigid and flexible phases in SBS modified asphalt; a high-temperature and high-pressure in-situ cavity was constructed and shear rates were applied simultaneously to simulate the asphalt pumping process; and molecular dynamics information was measured using fast-field cyclic nuclear magnetic resonance technology.

2. The method for analyzing stabilized rubber asphalt materials as described in claim 1, characterized in that, The steps for collecting the microscopic and macroscopic characteristic parameters of the stabilized rubber asphalt material using chromatography-mass spectrometry include: Volatile products were analyzed using Py-GC-MS technology to generate spectra; Polar compounds generated during aging were selectively extracted from the spectrum using supercritical fluid extraction technology. The spectra of the extracted polar compounds are processed using a pre-trained convolutional neural network to identify overlapping peaks.

3. The method for analyzing stabilized rubber asphalt materials as described in claim 1, characterized in that, The steps for acquiring the nanoscopic characteristic parameters of the stabilized rubber asphalt material using an infrared scanner include: The embrittled rubber particles are uniformly mixed with the asphalt that has lost its stickiness by low-temperature cryogenic pulverization technology to form a sample of the stabilized rubber asphalt material. The sample was subjected to infrared spectroscopy using ATR-FTIR technology to generate an infrared spectrum. Overlapping peaks in the infrared spectrum were separated using second-order derivative spectroscopy. A rubber asphalt spectral library was constructed to record the standard spectra of each component and aging products in the stabilized rubber asphalt material. The dynamic spectral changes of the stabilized rubber asphalt material under different temperature gradients were analyzed by two-dimensional correlation spectroscopy. The dynamic spectral changes are compared with a rubber asphalt spectral library to identify the aging-sensitive bands of the stable rubber asphalt material.

4. The method for analyzing stabilized rubber asphalt materials as described in claim 1, characterized in that, The steps of acquiring the macroscopic, mesoscopic, and microscopic characteristic parameters of the stabilized rubber asphalt material using a CT scanner include: The volume fraction of rubber particles in asphalt was obtained by high-energy microfocus CT. The deformation state of rubber particles under temperature cycling is captured by dynamic scanning mode; The 3D deformation of rubber particles under temperature cycling is quantified based on the volume fraction and deformation state using CT 3D reconstruction technology to generate 3D data. The 3D data is segmented using a pre-trained deep learning segmentation model to identify the rubber phase and the asphalt phase.

5. The method for analyzing stabilized rubber asphalt materials as described in claim 1, characterized in that, The step of performing a second set of high-throughput tests on the stabilized rubber asphalt material to obtain digital image data of the stabilized rubber asphalt material includes: A second set of high-throughput tests was conducted on the stabilized rubber asphalt material to acquire cross-scale feature images of the stabilized rubber asphalt material; The cross-scale feature images are subjected to denoising, data augmentation, and feature structure association processing to obtain digital image data characterizing the structural features of the stable rubber asphalt material.

6. The method for analyzing stabilized rubber asphalt materials as described in claim 1, characterized in that, The step of extracting the relationship model between structure and properties in the stable rubber asphalt material based on the cross-scale feature data and digital image data includes: Based on computer vision technology and multimodal learning technology, the cross-scale feature data and digital image data are classified to generate classification results; Based on multimodal learning, semantic segmentation, and object detection techniques, a model relating structure and properties in the stable rubber asphalt material is constructed according to the classification results.

7. The method for analyzing stabilized rubber asphalt materials as described in claim 6, characterized in that, The steps of constructing a model of the relationship between structure and properties in the stable rubber asphalt material based on the classification results include: associating structural parameters with molecular mobility and establishing a structure-rheological property prediction model.

8. The method for analyzing stabilized rubber asphalt materials as described in claim 1, characterized in that, Also includes: The testing process is optimized based on the optimized relationship model.

Images