Multiscale characterization of viscoelastic construction materials using microscopic experiments and machine learning

Viscoelastic materials are used widely in construction of built infrastructure, most notably in the case of asphalt binders used in road construction. The characterization of these materials presents substantial challenges due to their complex molecular makeup and time, temperature, and geometry-dependent failure mechanisms. Specifically, understanding the mechanism underlying failure due to fatigue and fracture has been a challenge. Meanwhile, at larger scales, the engineering of asphalt mixtures involves complex manipulation of volumetric and other variables. To address these challenges, a Bayesian Neural Network (BNN) was established to predict the dynamic complex modulus of asphalt mixtures based on a range of data from existing literature and our laboratory. The BNN was shown to have comparable errors to those observed experimentally when enough training and validation datasets were used. However, nonlinear viscoelastic problems provide more substantial challenges due to both lack of data and repeatability. Microscopic observations of the cavitation phenomenon underlying asphalt binder fracture were therefore observed using in-situ testing and darkfield microscopy. Future work involves predicting these cavitation events using physicsinformed neural networks.

Bio

Ramez Hajj is an Assistant Professor in the department of Civil and Environmental Engineering at the University of Illinois at Urbana-Champaign. Prior to joining UIUC, he completed his Ph.D. and Master's in Civil Engineering at the University of Texas at Austin. He obtained his Bachelor's Degree in Civil Engineering with a minor in Engineering Science and Mechanics from Virginia Tech. Dr. Hajj's research spans everything related to asphalt materials and flexible pavements, from as small as investigating asphalt binder behavior at the molecular, nano-, and micro-scale, to developing novel materials for roadway pavement and railroad track applications, to solving network-level transportation infrastructure problems. Primarily, his research focuses on asphalt binders, mixtures, and pavements, but more broadly, these research areas are situated in wider research areas of viscoelasticity, composites, machine learning, and computational modeling. So far, his work includes fundamental scientific discoveries into the origins of cracking in asphalt binder microstructure, the development of predictive machine learning algorithms for linear viscoelastic properties, and practical contributions to the asphalt industry, state Departments of Transportation, and local transportation agencies. To date, he has authored more than 25 peer-reviewed journal articles in these areas and presented his work at many international conferences.