Abstract: In order to solve problems of large amount of calculation and low efficiency in failure prediction of angle interlock composites, 4³ voxels in the original model are packed into one representative volume element by adopting the down-sampling voxel technique commonly used in image processing methods, thus greatly reducing the calculation amount and time. The full-voxel model is compared in terms of elastic property, damage state of composites and CPU time. Numerical results show that the model using down-sampling voxel technique is not only identical to the full-voxel model in stiffness prediction, but mimic the full-voxel model in damage state. At the same time, the CPU time is reduced by almost 85% in the model computation with tens of millions voxels, thus it is suitable for high-precision and high-speed prediction of engineering angle interlock composites.

Key words: downsampling voxel technique, angle interlock composite, strength prediction