Thursday, March 1, 2012
Mechanical and Civil Engineering Seminar
Computational Models for Multi-Scale Modeling of Deformation and Failure in Polycrystalline and Poly-phase Materials
Somnath Ghosh, Director, Computational Mechanics Research Laboratory, Department of Civil Engineering and Mechanical Engineering, Johns Hopkins University
This talk will discuss two thrusts in spatial and temporal multi-scale computational modeling of deformation, failure and fatigue of structural materials. The first thrust area is on crystal plasticity finite element (CPFE) modeling of polycrystalline metals and alloys (e.g. titanium alloys) for predicting cyclic deformation leading to fatigue. Image-based crystal plasticity FEM models are developed, incorporating statistically equivalent distribution functions of grain morphology and crystallographic orientations from data obtained from orientation imaging microscopy. A grain-based crack nucleation model evolves from considerations of energy needed to open a free surface in a hard grain surrounded by dislocation pileup in neighboring soft grains. A wavelet based multi-time scale methodology (WATMUS) is developed to significantly reduce the computational time in cyclic loading and deformation till crack initiation. The resulting finite element scheme is able to allow simulations to continue for a large number of cycles to fatigue crack nucleation. The second thrust area will be discussed in brief. It will summarize activities on concurrent multi-scale modeling of ductile failure in multi-phase materials such as cast aluminum alloys. Modules developed include: (i) multi-scale characterization based preprocessor; (ii) microstructural analysis module for ductile fracture; and (iii) homogenization-based continuum damage model for macroscopic analysis modules. Effective micro-mechanical modeling, incorporating particle fragmentation and ductile failure through matrix cracking, is accomplished by an adaptive, locally enriched Voronoi Cell finite element method or LE-VCFEM. Adaptive schemes and mesh refinement strategies are developed to create a hierarchy of computational sub-domains with varying resolution in the multi-scale model.