Multiscale analyses have been extensively used to virtually test how a material will respond linearly and nonlinearly, due to the initiation and evolution of damage, to a variety of loads and environmental conditions. This work improved several components of a multiscale framework. At the microscale, elastic properties were determined for four types of graphite fibers, including AS4, IM7, T300, and T650, along with a type of glass fiber, E-glass 21xK43, using an inverse method. Homogenization methods used in the inverse analyses include:finite element analysis (FEA) with a hexagonal microstructure, FEA with a random microstructure, and Mori-Tanaka averaging scheme. Fiber properties determined using FEA with a hexagonal microstructure and the Mori-Tanaka averaging scheme were very similar, while using FEA with a random arrangement of fibers resulted in significantly different properties. The predicted longitudinal shear modulus, G_(12), of the graphite fiber was observed to almost linearly depend on the minimum spacing between fibers, while the other engineering constants did not depend on the minimum space between fibers. The predicted properties for the glass fiber were shown to be insensitive to the homogenization method used. At the mesoscale, two types of continuum damage models, a cohesive zone model, and a combination of the two types were compared using a [0/90]_(s) laminate under uniaxial tension and in-plane shear loads. The volume average stress-strain response, the crack density evolution, and a metric developed using two-point correlation functions were used to quantify the similarities and differences of the progressive damage models. For a laminate under uniaxial tension, a continuum damage model that degrades the material on an element basis predicted a progression of damage similar to the cohesive zone model. A continuum damage model that degrades the material on a quadrature point basis predicted a lower applied strain for final failure and a higher crack density. Under in-plane shear, the continuum damage models predicted damage growth across fibers, which is unrealistic. Cohesive zone elements can be placed where damage is expected, but when placed in all directions, the cohesive zone model predicted the same unrealistic damage growth across fibers.
