ABSTRACT: While interrupted in-situ X-ray tomography can provide detailed 3D information about material behaviour, it has not been possible to achieve a similar level of detail for dynamic processes which cannot be interrupted. We develop a framework based on neural rendering which enables the reconstruction of 4D information for specimens undergoing complex spatio-temporal deformation. Using rich tomographic information at the beginning and end of deformation, and sparse X-ray projections during deformation, we demonstrate that it is possible to resolve dynamic events such as formation of crush bands in space and time with sub-second resolution. We further demonstrate that the method can be extended to stereo X-ray setting whereby two projections are used to reconstruct complex processes such as buckling in octet lattices and localisation in indentation experiments. This work can be used to drive the paradigm shift to enable the 4D reconstruction of dynamic events in materials engineering and beyond.

BIO: Dr. Ivan Grega obtained his PhD from the University of Cambridge under Prof. Vikram Deshpande where he focused on using machine learning methods to improve computational and experimental methods in materials engineering. During this period, he interned at Mila – Quebec AI Institute with Prof. Yoshua Bengio where he developed a physics-based machine learning for the optimisation of CO2 gas diffusion electrodes. He completed his undergraduate studies in Cambridge with a major in Mechanical and Aerospace Engineering.