Modelling pavement performance is important to pavement asset management and to roadway safety and sustainability. While neural networks (NNs) have been employed to establish pavement performance models, past studies have mainly focused on achieving high model accuracy and generally failed to examine other model properties. This study calibrated feedforward neural networks with particle swarm optimization (PSO) to predict the rutting performance of asphalt pavement in the State of Idaho. The resulting models provided rutting predictions with higher accuracy than traditional mechanistic-empirical models. This study also employed various statistical methods in the pre-processing and post-analysis of PSO-NNs, such as the principal component analysis in the input selection and hypothesis tests in the model reproducibility and robustness evaluation. This study further explored the relationships between the model accuracy and updating efficiency and the number of calibrated parameters (hidden neurons). The selected PSO-NN structure struck the right balance between model accuracy, reproducibility and robustness of the models.

Since rutting is a major distress in flexible pavements at intermediate and high temperatures, rutting models with concise expressions and satisfying accuracy can have considerable contributions to the pavement design, maintenance and rehabilitation. In this circumstance, two mechanistic-empirical rutting models were proposed in this study. The models were adapted from traditional rutting models and involve advanced mechanical aspects such as strains and stress invariants under moving loads and a yield criterion for asphalt mixtures. This paper includes detailed explanations of model expressions, qualitative comparisons with current models and model applications in an innovative laboratory test. The laboratory test and finite element simulation of asphalt mixtures under moving loads were used to calibrate, validate and check the sensitivity of the proposed models. The results showed the proposed models successfully captured rutting development in asphalt mixtures; they also reasonably predicted rutting behaviors of layered structures under different loading and environmental conditions. In model applications, effects of different factors on material properties, structural strain and stress responses and rutting development were evaluated. Finally, the determination of model parameters in the proposed models was discussed and recommended for further research.

Ultrasonic testing can be used for a nondestructive and rapid determination of material properties. In this study, twelve asphalt mixture samples of four different types were fabricated and used in conventional material property tests and two ultrasonic wave tests. Physical properties such as bulk specific gravity and air void content, mechanical properties such as dynamic modulus and phase angle, and acoustic properties such as wave velocity were measured. Relationships between these properties were established and analyzed as a tool for the future material property determination. In addition, the dynamic modulus and phase angle, measured in a standard laboratory test, were used to construct two master curve models to predict their values at arbitrary temperatures and frequencies. Furthermore, a theoretical model for wave velocity in a linear isotropic viscoelastic material was utilized with measured density, Poisson’s ratio, phase angle and ultrasonic wave velocity to predict the dynamic modulus. Good agreement has been achieved between laboratory measurements and model predictions. It indicates that ultrasonic testing can serve as a rapid method for material property determination.

Field aging significantly accelerates distresses (fatigue cracking, water damage, etc.) of asphalt pavements due to exposure to the natural environment. Conventional studies explore field aging by laboratory tests of field cores, while little work is devoted to the inverse analysis of aging by nondestructive tests (e.g., falling weight deflectometer (FWD) test). To backcalulate the aging parameters accurately and automatically, an inverse analysis method of asphalt pavement field aging was proposed based on System Identification (SID) in this study. In this inverse procedure, a kinetics-based two-reaction field aging model was employed as the theoretical aging model and the parameter adjustment algorithm was developed based on sensitivity analysis. Fifty sections from four different climate zones in Long Term Pavement Performance (LTPP) database were selected for investigation, where backcalculated moduli from LTPP and field aging temperatures determined by a pavement hourly temperature model were utilized. The results indicated that SID was feasible to inversely analyze the field aging of asphalt pavements automatically with high accuracy and efficiency. To investigate the factors affecting aging parameters, the multiple linear regression analysis and one-dimensional sensitivity analysis were performed. Among those variables, shortwave radiation and cloud cover showed the dominant effects on field aging resistance.