Abstract
A generic 3-D boundary-element substructuring algorithm, based on iterative solvers, is applied to solve problems involving composite materials. The substructuring technique is optimal regarding memory requirements for storing the resulting sparse matrices, and its efficiency is increased by implementing structured matrix-vector products in the built-in Krylov solver. Moreover, special integration techniques are employed to effect the modeling of thin-walled parts of the composites, such as thin layers or reinforcing elements. To test the performance of the algorithm, effective mechanical properties of carbon-nanotube-reinforced composites are estimated by analyzing representative volume elements. The advantages of the coupling algorithm for developing parallel BE codes are also highlighted.