Pavement materials that could enhance the mechanical properties of open-graded porous asphalt mixtures in long-term service periods could offer a solution to produce long-life pavements, causing a reduction of interventions' needs, as well as the associated disruptions to road users and user costs. One option to improve the longevity of open-graded porous mixtures is with the use of epoxy asphalt that, despite its high initial cost, offers enhanced longevity that might offset any future user and intervention costs. This study aimed to evaluate the durability of plant-produced epoxy-modified open-graded porous asphalt mixtures. A batch production plant was employed to produce loose mixtures, which were used to pave a test road in the Province of Gelderland, the Netherlands, and compact specimens in the laboratory. Control mixtures with a non-epoxy-modified asphalt binder were also produced in the same plant. The durability of laboratory- and field-compacted mixtures was evaluated by conducting indirect tensile tests before and after oven conditioning. Results illustrated that the epoxy-modified asphalt demonstrated the highest strength and stiffness values, while the strength was reduced after conditioning in a water bath with the retained strength within the allowable specification limits. This attribute was confirmed from drill cores obtained from the test road after one year in service. Also, the materials compacted in the field had slightly higher strength and stiffness values than the laboratory-produced mixtures. Although the results provided have illustrated the improvement of durability of open-graded porous asphalt with implementing epoxy modification, further evidence from the test road over the years is needed for validation.

Physics Informed Neural Networks (PINNs) have been rarely applied to solve multiphysics systems due to the inherent challenges in optimizing their complex loss functions, which typically incorporate multiple physics-based terms. This study presents a multistage PINN approach designed to efficiently solve coupled multiphysics systems with strong interdependencies. The multistage PINN progressively increases the complexity of the physical system being modeled, enabling more effective capture of coupling between different physics. The computational merits of this approach are demonstrated through two illustrative applications: prediction of asphalt aging and modeling of lid-driven cavity flow. Quantitative and qualitative comparisons with standard PINN and adaptive weight PINN approaches demonstrate the enhanced precision and computational efficiency of the proposed algorithm. The multistage PINN achieves a reduction in training time of more than 90% compared to standard PINNs while maintaining better alignment with the finite element method (FEM) solutions. The improvement in computational efficiency, coupled with enhanced accuracy, positions the multistage PINN as a powerful tool for addressing complex multiphysics problems across various engineering disciplines. The method’s ability to handle interactions between multiple physical processes, such as diffusion, chemical reactions, and fluid dynamics, makes it suitable for simulating long-term material behavior and complex fluid systems.

This research addresses the critical issue of load transfer efficiency (LTE) in jointed plain concrete pavements (JPCP), with a specific focus on the role of dowel bars in ensuring optimal load transfer and providing a comfortable ride for vehicles. While experimental studies have investigated factors like joint width, slab thickness, concrete strength, and dowel bar size that influence LTE, they are limited in their ability to accurately replicate real-world conditions and can be time-consuming. To overcome these limitations, finite element modelling (FEM) is employed as a powerful tool for simulating complex loading conditions and analyzing stress and strain distributions in pavements. The primary objective of this research is to develop an advanced FE model that incorporates the forklift tire-pavement interaction, enabling precise analysis of complex loading conditions in industrial pavements and the impact of various rigid pavement parameters on load transfer. By explicitly considering the interaction between the tire and pavement, the proposed model will provide an extensive and robust numerical tool for designers and engineers. Additionally, this study represents a novel framework to integrate concrete pavement dowel bars and complex tire modelling using FEM. The developed methodology holds significant promise in optimizing the design of dowel bar systems and back-calculating the pavement parameters for rolling weight deflectometers.

Scymol is a Python-based software package specifically designed to facilitate the setup and execution of molecular simulations in LAMMPS. It comes equipped with a user-friendly interface, which simplifies the process of initializing molecular systems and defining simulation parameters. Moreover, the software generates and executes LAMMPS simulation sequences, enabling researchers to establish comprehensive simulation schemes, such as heating or deformation cycles, in a single run. Through its successful application in diverse research projects and its modular design, Scymol demonstrates considerable promise as an indispensable tool for researchers aiming to carry out molecular dynamics simulations without sacrificing complexity or high-throughput capabilities in their methodologies.

A major contributor to GHG emissions is the transportation sector, particularly pavement transport. The limited understanding of tire-pavement interactions leads to inaccurate predictions of these emissions, particularly from rolling resistance (RR). Traditional methods for predicting RR are constrained by their limited applicability and inability to account for the complex dynamics of tire-pavement interactions, resulting in poor prediction accuracy. These limitations make it challenging for policymakers to make proper decisions, as existing methods are manual and labour-intensive. This study aims to develop an automated system to capture tire-pavement interaction data using the Laser Crack Measurement System (LCMS). To the best of the authors' knowledge, no robust technique currently exists for automatically calculating tire penetration-related information from LCMS data to predict RR. Therefore, this research explores machine learning (ML)-based models to reduce uncertainties in existing approaches and enhance RR predictions using automated LCMS data. It examines the relationships between RR, tire penetration volume, and the characteristics of the Dutch pavement network, comparing the results with those of commonly used RR prediction models. The study introduces an automatic tire penetration calculation approach using LCMS data to assess the impact of tire penetration volume and depth on RR in relation to surface properties. The findings reveal that traditional empirical models show poor correlations between RR and texture indicators, whereas ML-based models significantly improve the accuracy of RR predictions. These results could inform the development of strategies to reduce GHG emissions from pavement transport, supporting global efforts to combat climate change and achieve the goals of the Paris Agreement.

Asphalt mixtures show complex mechanical behavior due to their heterogeneous structure. Traditionally, the mechanical characterization of asphalt mixture is done through laboratory testing or micromechanical modeling. While laboratory tests and micromechanical models provide reliable measurements and physical interpretability, they are often resource-intensive and demand extensive calibration. Recent advances in machine learning address some of the above issues by enabling accurate predictions, though often lacking physical interpretability and stability. Hence, this study aims to present a novel micromechanics-infused neural network (MINN) framework for predicting asphalt mixture stiffness. The framework embeds micromechanical principles derived from the modified Hirsch model into the neural network's loss function, allowing the model to learn from experimental data while adhering to micromechanics-based constraints. In this study, feature selection is performed using BorutaShap, and Bayesian optimization is applied for hyperparameter tuning. Results show that MINN improves prediction accuracy, interpretability, and robustness.

The chemo-mechanical properties of bitumen undergo significant alternations during aging and rejuvenation, posing challenges for accurately evaluating and enhancing rejuvenation efficiency in asphalt recycling. This study investigates how bitumen source, aging degree, rejuvenator type and dosage influence the chemical and rheological performance of rejuvenated bitumen. Comprehensive characterizations are performed using saturate, aromatic, resin, and asphaltene (SARA) fractionation, elemental analysis, gel permeation chromatography (GPC), and dynamic shear rheometer (DSR) tests. To elucidate chemo-rheological correlations, statistical techniques (Pearson correlation, analysis of variance (ANOVA), and Chi-square tests) are combined with artificial neural networks (ANN). Results indicate that the NB bitumen with more colloidal stability and less sulfur content exhibits the highest resistance to long-term aging. FB bitumen with 4.3 % sulfur achieves the best high-temperature deformation resistance with rutting failure temperature (RFT) higher than 80 °C, and TB bitumen exhibits the longest fatigue life. Rejuvenation using bio-oil is most effective on reducing relaxation time by up to 60 % and increasing creep compliance (Jnr3.2) by 1.7–2.5 times, depending on bitumen type. Rejuvenator dosage sensitivity for relaxation stress follows the trend: bio-oil < engine-oil < naphthenic-oil, while aromatic-oil shows variability depending on its source. Among the tested rejuvenators, bio-oil proves most effective, particularly for rejuvenating TB and FB bitumen. The ANN model demonstrates strong predictive performance for rheological properties, achieving R2 values between 0.90 and 0.98, with the highest accuracy observed for relaxation indices, followed by fatigue and rutting properties.

Previous work demonstrated that Random Forest Regressors (RFRs) could estimate the physical properties of bitumen using molecular descriptors derived from Molecular Dynamics (MD) simulations, thereby reducing the need for computationally intensive simulations. However, due to their decision-tree structure, RFRs lack true predictive capabilities, particularly for interpolation and extrapolation beyond the training data.

This study advances that foundation by employing Artificial Neural Networks (ANNs), which—when properly trained—can capture complex relationships with greater continuity and generalizability. Beyond simply replacing RFRs, we develop a fully automated framework for constructing Machine Learning Models (MLMs) to predict density and thermal expansion coefficients of bitumen. Using Optuna for hyperparameter optimization, we ensure that the information extracted from MD simulations is utilized effectively.

The resulting ANN models accurately reproduce MD-predicted densities, achieving R2>0.99, MSEs below 0.1 %, and maximum absolute errors below 5 % on test data. In addition to reducing computational cost, the models exhibit improved interpolation and extrapolation capabilities, enabling reliable predictions for properties, ranges, and compositions not explicitly simulated.

Key aspects of our approach include:
• Transitioning from RFRs to ANNs, improving generalization, interpolation, and predictive accuracy.
• Automated hyperparameter optimization, leveraging Optuna to maximize model efficiency.
• Expanding applicability, enabling property prediction for unseen compositions without additional MD simulations.

Asphalt binders are critical for asphalt pavement performance, and understanding their rheological behavior is essential for designing durable roadways. The complex shear modulus (G⁎) and phase angle (δ) are primary parameters characterizing binder rheology. This study introduces a novel hybrid machine learning model combining deep neural networks (DNN) and Gaussian process regression (GPR) to predict G⁎ and δ for bituminous binders and binder-filler systems (mastics). DNN excel at capturing complex, nonlinear relationships among eleven binder and thirteen mastic input parameters, including aging conditions, chemical and physical properties, and test parameters. However, standalone DNN struggle with small datasets, common at the binder scale, and lack inherent uncertainty quantification, limiting reliability in engineering applications. GPR improves DNN by refining predictions through probabilistic modeling, while providing uncertainty estimates, and enhancing accuracy with limited or noisy data. The hybrid model leverages DNN's feature extraction capabilities and GPR's ability to smooth predictions, significantly improving performance over standalone DNN. The hybrid model achieves high prediction accuracy, with R2 values of 0.997 for G⁎ and 0.947 for δ for binders, and 0.993 for G⁎ and 0.972 for δ for mastics, reducing G⁎ prediction error from 22.7% to 0.031% for fresh asphalt binder compared to standalone DNN. Feature importance analysis using random forest and SHAP techniques identifies test temperature, aging conditions, and penetration as key influencers of G⁎ and δ. This hybrid approach enhances the characterization of complex asphalt materials, offering pavement engineers a robust, reliable tool for predicting material behavior under diverse conditions.

The low-temperature relaxation and fatigue cracking performance are two essential aspects in estimating the rejuvenation efficiency of recycling agents (RAs). This study aims to fundamentally investigate the effects of recycling agent type/dosage and aging degree of bitumen on thermodynamic and rheological properties of rejuvenated bitumen at low and intermediate temperatures. Molecular dynamics (MD) simulations are utilized to predict thermodynamic indices of rejuvenated bitumen, further linked to critical low-temperature and fatigue indicators from experiments. The results reveal that all RAs show a regeneration effect on fractional free volume (FFV), self-diffusion coefficient (DS), glass transition temperature (Tg), and surface free energy (γ). Bio-oil and engine-oil exhibit higher rejuvenation efficiency on these thermodynamic properties than naphthenic-oil and aromatic-oil. The aging degree of bitumen and temperature show significant effects on rejuvenation efficiency. It is recommended to use the FFV parameter to predict the relaxation properties of rejuvenated bitumen. However, these thermodynamic indicators inadequately differentiate between rejuvenators and softeners. Based on crossover parameter results, most recycling agents (bio-oil, engine oil, and naphthenic oil) in this study display softening characteristics. Only aromatic oil effectively rejuvenates the crossover modulus (Gc) of aged binder. Notably, engine oil demonstrates the least rejuvenation in crossover parameters for the recovery of aged bitumen. Further, γ demonstrates a strong association with both Glover-Rowe (G-R) and fatigue crack width C500 indices across all cases involving rejuvenated bitumen. This work will build a multi-scale evaluation framework on the rejuvenation effectiveness of recycling agents on low-temperature and fatigue performance of aged bitumen.

The ever-growing need to build roads to meet the necessary transportation demands is challenging, especially for developing countries. Low-volume roads (LVRs) are usually the backbone of catalyzing economic growth in these countries. With impediments surrounding Petroleum bitumen (price fluctuations) and environmental concerns, scientists are putting their effort into finding an alternative. The presented research is an attempt to check if Natural asphalt can be used as a full or partial replacement of the Petroleum bitumen. To the best of the authors' knowledge, only limited studies have focused on characterizing and understanding the engineering properties of Natural asphalt. The available techniques do not provide reliable information to the road authorities and hence they are discouraged from using it in practice. Particularly for countries, where the Natural asphalt source is available, the overall dependence on importing the Bitumen could be substantially reduced. Empirical and experience-based design criteria may not be sufficient as the standards were never developed for such materials, hence, a scientific approach is required before bringing it into practice. In this research, Natural asphalt sourced from different locations in Nigeria was assessed. Before performing the mixture level tests using Marshall and Cantabro design methods, the rheological and fatigue properties of the extracted Natural bitumen were examined in the laboratory. In the design of the experiment, various percentages of Natural asphalt were added between 0 % and 20 % by total mix weight; implying that the remaining required fraction of binder was fulfilled by the addition of petroleum bitumen. By using a ranking system (supported by statistics), an optimal design of mixture was obtained which was used in the field (exposed to normal traffic) at 30 different sections.

The linear viscoelastic behavior of materials is represented using mechanical models of choice, which are further utilized in different numerical investigations, such as finite element simulations and discrete element simulations. Burger's model is one of the widely adopted mechanical models and remains highly favored in contemporary research due to its multiple advantages. Specifically, it excels in representing long-term creep and stress relaxation behavior in a relatively simplified manner. Accurate identification of the long-term behavior for the viscoelastic material, particularly asphalt concrete, is crucial, as it serves as a key indicator of asphalt pavement performance over its service life. However, past research studies show that the parameters of Burger's model should be back-calculated from experimental data only within a limited range of frequency, otherwise, the parameters fail to represent the true material behavior. To the best of the authors’ knowledge, there is no approach for researchers to obtain the critical frequency range in which the experiments should be performed. Therefore, this study proposes a novel framework to find the critical frequency range to obtain appropriate model parameters of Burger's model, to better characterize the viscoelastic behavior of the materials. To examine the framework, asphalt concrete mixtures are used as examples in this study. Necessary laboratory tests including complex modulus tests and stress relaxation tests, are performed on two distinctive types of asphalt concrete mixtures. The generalized Maxwell model with different number of Maxwell chains are used to evaluate the performance of Burger's model. Furthermore, since commercially available finite element packages generally do not have a direct built-in Burger's model, the article shows a way of implementing Burger's model in finite element simulation. The simulations corresponding to the laboratory tests are carried out in both frequency domain and time domain to thoroughly evaluate the performance of Burger's model. The optimal frequency range of 0.1–20 Hz for the examined mixtures is found to significantly improve the accuracy of the descriptive master curve. The results also suggest that the generalized Maxwell model requires a minimum of four Maxwell chains to maintain good performance in accurately characterizing the behavior of asphalt mixtures. However, adding more Maxwell chains beyond a critical limit may not provide significant benefits. Finite element simulations demonstrate that the stress relaxation behavior predicted by the obtained Burger's model parameters aligns more closely with experimental data over longer time intervals. This makes Burger's model a strong choice for aiding in the design of simulations for studies focused on the long-term behavior of materials.

This study explores the use of chemical descriptors derived from force field atom types to predict Fickian diffusion coefficients of rejuvenators in bitumen, utilizing machine learning models trained on data from 240 non-equilibrium molecular dynamics simulations. The simulations cover three bitumen types (NO, TO, FO), five aging degrees, and four temperatures (60 °C, 120 °C, 160 °C, 200 °C), capturing diffusion coefficients ranging from 0.0068e-10 m2/s in highly aged bitumens at 60 °C to 4.35e-10 m2/s in fresher samples at 200 °C. The MLM, built with 18 chemical descriptors for bitumen and rejuvenator sides, achieves an R2 of 0.97, accurately predicting diffusion across varied conditions. This approach abstracts away from the need for repeated MD simulations, enabling diffusion predictions even for systems outside the original dataset. The manuscript presents three case studies to illustrate how the model can be used for the iterative design of rejuvenators by optimizing molecular structures based on critical chemical features, such as rejuvenator oxygen content, bitumen sulfur content, and molecular weights. It also demonstrates how the model offers a practical framework for understanding the diffusion and performance of rejuvenators by linking time-dependent factors—such as concentration, depth, and rejuvenation time—with the bulk properties of bitumen-rejuvenator systems, facilitating industrial applications.

This study employs strain-controlled oscillatory deformations in Molecular Dynamics (MD) simulations to evaluate the dynamic properties of all-atom molecular systems, specifically targeting the SARA fractions of bitumen. Twelve molecular systems representing these fractions were modeled using the PCFF force field. The simulations effectively captured their viscoelastic properties across multiple frequency domains, including Elastic, Glassy, Rubbery, and Viscous responses. Reported storage and loss moduli range from thousands to tens of megapascals, with viscosities from tens to near-zero Pascal-seconds across various frequencies and temperatures, aligning well with experimental observations. Saturates and Aromatics were identified as the softest and most thermally susceptible fractions, while Resins and Asphaltenes were the stiffest and least susceptible. The study reveals that the relaxation time of all-atom molecular systems is significantly shorter than in experimental setups, necessitating careful comparison of stress-related phenomena across equivalent relaxation times. Although this allows for the exploration of response profiles in computationally tractable simulations, the nature of all-atom force fields and simulation algorithms introduces spatiotemporal scale discrepancies that must be addressed in future simulations involving the study of stress-related phenomena using MD.

Developed by Delft University of Technology, the tri-component polyurethane modified cold binder (PMCB) displays impressive durability and strength in asphalt mixtures, showing promise as a reliable binder for cold in-place recycling. However, when applying PMCB for rapid, in-situ recycling, the presence of moisture in reclaimed asphalt pavement (RAP) poses a significant challenge. To address this, an innovative approach involving treatment of the wet RAP with Calcium dioxide (CaO) prior to the integration of PMCB was tested. Evaluation methods used included the Indirect Tensile Test (ITT), followed by the calculation of the Indirect Tensile Strength Ratio (ITSR) to assess moisture susceptibility. Furthermore, Cantabro tests were performed to determine the material loss under abrasion and weathering conditions. These assessments underscored the feasibility of this approach. The treatment of wet RAP with CaO has proven a viable strategy for rapid in-situ recycling with PMCB, contributing to sustainable pavement construction. In addition, the research identified that a 5.5% concentration of the PMCB binder maximizes structural integrity and performance in the considered RAP.

The advent of computational techniques, particularly atomistic simulations, has lessened the dependency on physical experiments in various scientific fields. Yet, the preparation complexity for simulations using platforms like LAMMPS and GROMACS persists. We introduce SMI2PDB, a Python tool that automates molecular systems assembly from SMILES to PDB format, easing molecular dynamics simulation setups. SMI2PDB manages molecule configuration and quantification effortlessly, establishes stable conformers, applies random rotations, and positions them in a simulation box with a Sobol sequence to reduce overlaps. This script facilitates the rapid preparation of complex organic mixtures for use in simulations, enhancing the exploration of novel materials.

This paper presents a United Atom (UA) force field for simulating hydrocarbon molecules in bituminous materials, integrating explicit hydrogens into beads with their parent atom. This method simplifies all-atom molecular models, significantly accelerating Molecular Dynamics (MD) simulations of bitumen by 10 to 100 times. Key advantages include halving the particle count, eliminating complex hydrogen interactions, and decreasing the degrees of freedom of the molecules. Developed by mapping forces from an all-atom model to the centers of mass of UA model beads, the force field ensures accurate replication of energies, forces, and molecular conformations, mirroring properties like pressure and density. It features 17 bead types and 287 interaction types, encompassing various hydrocarbon molecules. The UA force field's stability, surpassing all-atom models, is a notable achievement. This stability, stemming from smoother potential energy surfaces, leads to consistent property measurements and improved stress tensor accuracy. It enables the extension of MD simulations to larger spatiotemporal scales, crucial for understanding complex phenomena such as phase separation in bituminous materials. This foundational work sets the stage for future developments, including refining parameters and introducing new bead types, to enhance the modeling capabilities of the force field, thereby advancing the application and understanding of bituminous materials.

This study enhances the molecular analysis of bitumen by transitioning from traditional chemical descriptors, such as SARA (Saturates, Aromatics, Resins, and Asphaltenes) fractions and elemental compositions, to specific force field atom types in Molecular Dynamics (MD) models. This shift improves the precision in predicting material properties critical for bituminous material characterization. Machine Learning Models (MLMs) were developed to use these atom types as input features, inherently reflecting fundamental chemical characteristics. Trained on data from over 1,770 LAMMPS simulations of diverse bitumen types and conditions, these MLMs enable the prediction of properties like density, heat capacity, solubility parameters, and thermal expansion coefficients without the need for additional MD simulations. The models utilize 30 chemical descriptors corresponding to specific atom types in the PCFF force field, which collectively account for over 95% of the influence on these properties. By accurately predicting fundamental, thermodynamic, and kinetic properties, the use of MLMs and force field atom types allows researchers to efficiently tweak the chemical nature of organic molecules and mixtures to achieve desired properties. With near-instantaneous prediction times, these MLMs offer valuable insights for advancing bitumen research in the construction and petroleum industries, reducing the need for more intensive simulation techniques.

Inconsistencies between performance data from laboratory-prepared and field samples have been widely reported. These inconsistencies often result in inaccurate condition prediction, which leads to inefficient maintenance planning. Traditional pavement management systems (PMS) do not have the appropriate means (e.g., mechanistic solutions, extensive data handling facilities, etc.) to consider these data inconsistencies. With the growing demand for sustainable materials, there is a need for more self-learning systems that could quickly transfer laboratory-based information to field-based information inside the PMS. The article aims to present a future-ready machine learning-based framework for analyzing the differences between laboratory and field-prepared samples. Developed on the basis of data obtained from field and laboratory data, the gradient-boosting decision trees-based framework was able to establish a good relationship between laboratory performance and field performance (R2test > 80 for all models). At the same time, the framework could also show more complex relationships that are often not considered in practice.

Atomistic simulations are crucial for understanding material properties at the molecular level but are limited by high computational costs, especially for large, complex systems like bituminous materials. Our team developed a Force-matched United Atom (UA) Coarse Graining (CG) force field to enhance computational efficiency while retaining atomic detail. However, converting all-atom models to CG models is complex, requiring detailed atom-to-bead mapping and compatibility with molecular dynamics (MD) engines like LAMMPS. To address this, we introduce AA2UA, an open-source software that simplifies the conversion of PDB files into LAMMPS-readable structure topology files, facilitating broader use of the developed UA force field.