Data-driven computing for nonlinear problems of composite structures based on sub-domain search technique

Abstract

The distance-minimizing data-driven algorithm brings new insight in computing nonlinear problems of composite structures, where the calculation is directly carried out by finding the material data that closest to equilibrium constraints. However, it is challenging to efficiently find the desired material data in a tremendous database. Towards this end, a sub-domain search technique is proposed to construct a sub-database with adaptive size for each computational step, which enables to significantly speed up the data search process in the data-driven algorithm. The databases of the composite materials are collected by the offline homogenization over the representative volume element. Several examples, including the nonlinear geometric behavior of sandwich beam, the internal instability of fiber reinforced composite plate and the large deformation of composite beam, are performed to validate the reliability and efficiency of the proposed technique. The results show that compared with the classical multiscale finite element method, the data-driven method with the sub-domain search technique is able to predict the nonlinear behaviors of composite structures with good accuracy and efficiency.