Efficient Nonlinear Multiscale Analysis Using Sparse Sampling-Based Model Order Reduction Method

In this study, we conducted to improve the computational efficiency of the classical FE2 method by introducing micro-level reduced order modeling technique. For the classical FE2 method, multiple repetitive computations in microscopic representative volume element are required considering nonlinearities of such unit cells. Therefore, a great amount of computational resource is required for the multiscale analysis considering the nonlinearities in both macro- and microscopic domains. We propose to introduce reduced-order modeling of the representative volume element model using sparse sampling-based nonlinear reduced order modeling to improve the efficiency of FE2 analysis. We verify the proposed method comparing accuracy and efficiency with those of full FE2 analysis investigating several microscopic and associated macroscopic models.