Stress and strain localization modeling of particle reinforced metal matrix composites.

摘要

Particle reinforced metal matrix composites (PR MMCs) are characterized by a distribution of particles in a continuous metal matrix. The objective of this thesis is to screen the effects of model features on local stress and strain fields and document how they lead to stress and strain localization.; A factorial design of experiments methodology (DOE) is used to systematically compare the effect that features of 2D and 3D unit cell finite element models have on the predicted response of PR MMCs and develop an understanding of the limitations imposed by these features. Common features of all unit cell models for the DOE study are a particle volume fraction of 0.2 and linear elastic particles. Six features are varied to create 16 test cases: model type (plane stress, plane strain, generalized plane strain, and three dimensional), loading condition (tension, compression, shear, and biaxial tension), particle shape (ellipsoid and block), matrix type (linear elastic and elastic-perfectly plastic), thermal residual stress (present and not present), and particle aspect ratio (1 and 3). The quantitative effects of the features are used to screen their relative importance for prediction of: apparent modulus, macroscopic stress, equivalent plastic strain, maximum principal stress, and hydrostatic stress. A 3D multi-particle model (3D-MP) of a random particle distribution is compared with plane stress, plane strain, and generalized plane strain models to study the effect of model selection for a realistic particle distribution. Finally, results from a 3D model containing a cubic array of spherical particle clusters (3D-CLUS) are compared with those for random and periodic microstructures.; The results of the DOE study indicate that model selection sensitivity varies for different loading conditions. Stress and plastic strain fields and the strength-differential effect are significantly influenced by model selection and particle aspect ratio. Comparison of 3D and 2D multi-particle models indicates that while the predicted overall stress-strain responses are very similar, the local stress and strain fields are much different. The 2D models predict plastic strain to localize in diagonal bands much more than does the 3D model. It is these local fields that are important for ductility. The model of a clustered microstructure predicts a stiffer response with more hardening than do the models of random and periodic microstructures. (Abstract shortened by UMI.)