Doctoral

Short Description

Traditional homogenization relies either on mean-field schemes with limited accuracy or full-field simulations with prohibitive computational cost. In this work, a deep material network is explored as a middle ground, combining the efficiency of mean-field models with the fidelity of full-field homogenization for nonlinear heterogeneous composites

Short Description

Multiscale modelling has been conventionally used with mean field or first statistical moments for micromechanical modelling of composites. In this work a novel approach would be adopted to compute higher order moments. Thereby statistical distribution of local field quantities can be estimated. This will be used for modelling inelastic behaviour of composites.

Short Description

Semi-analytically computed first and second statistical moments provides fluctuation of local fields such as stress and strain. In this work a novel approach would be adopted to identify localized damage in microstructures. This will be used for modelling failure initiation in composites.