This paper explores a convergent approach that combines advanced informatics and computational science to develop road-paving materials. It also analyzes research trends that apply artificial-intelligence technologies to propose research directions for developing new materials and optimizing them for road pavements. This paper reviews various research trends in material design and development, including studies on materials and substances, quantitative structure–activity/property relationship (QSAR/QSPR) research, molecular data, and descriptors, and their applications in the fields of biomedicine, composite materials, and road-construction materials. Data representation is crucial for applying deep learning to construction-material data. Moreover, selecting significant variables for training is important, and the importance of these variables can be evaluated using Pearson’s correlation coefficients or ensemble techniques. In selecting training data and applying appropriate prediction models, the author intends to conduct future research on property prediction and apply string-based representations and generative adversarial networks (GANs). The convergence of artificial intelligence and computational science has enabled transformative changes in the field of material development, contributing significantly to enhancing the performance of road-paving materials. The future impacts of discovering new materials and optimizing research outcomes are highly anticipated.

PURPOSES : This study aimed to develop a quantitative structure property relationships (QSPR) model to predict the density from the molecular structure information of the asphalt binder AAA1, a non-full connected structure mixed with a total of 12 molecules. METHODS : The partial least squares regression (PLSR) model, which models the relationship between predictions and responses and the structure of these variables, was applied to predict the density of a binder with molecule descriptors. The PLSR model could also analyze data with collinear, noisy, and multiple dimensional independent variables. The density and additive-free AAA1 binder’s molecule systems generated by an asphalt binder’s molecules-related study were used to fit the PLSR model with the molecular descriptors produced using alvaDesc software. In addition to developing the relationship, a systematic feature selection framework (i.e., the V-WSP- and PLSR-modelbased genetic algorithm (GA)) was applied to explore sets of predictors which contributed to predicting the physical property. RESULTS : The PLSR model accurately predicted the density for the AAA1 binder’s molecules using the condition of the temperature and aging level (R2 was 0.9537, RMSE was 0.00424, and MAP was 0.00323 for the test data) and provided a set of features which correlated well to the property. CONCLUSIONS : Through the establishment of the physical property prediction model, it was possible to evaluate the physical properties of construction materials without limited experiments or simulations, and it could be used to comprehensively design the modified material composition.

PURPOSES : This study aims to determine whether machine learning techniques based on the results of chemical analysis experiments can be rationally applied to evaluate the aging of various asphalt binders used throughout the country. METHODS : We conducted chemical experiments such as FT-IR, H-NMR, C- NMR, and GPC for the three-stage aging levels of eight types of asphalt binders used in the country and utilized two artificial neural network models to determine valid chemical experimentation and conditions for the use of neural modeling through predictions. RESULTS : The M-prop model, which combined the findings from each neural network model into a single artificial neural network model, demonstrated superior predictive performance compared with the M-base model, which assessed aging using two cluster layers. In addition, the minimum amount of data required to achieve 100% predictive accuracy for the target asphalt binders, regardless of the artificial neural network model, was 18, and the amount of training data decreased to less than 50%. CONCLUSIONS : The predictive accuracy of the aging of asphalt binders was significantly enhanced when GPC data was used, indicating that GPC should be prioritized in evaluating the aging of asphalt binders.