HPC Computing in Micromechanics

What is it about?

The paper concerns linear micromechanics exploiting CT scans for determination of microstructure and numerical homogenization with a focus on a specific application - analysis of fiber-reinforced concrete. The analysis includes an identification problem and stochastic uncertainty, which brings new dimensions and enhances the need for fast solvers and ultrascale computations. Fiber-reinforced concrete with steel fibers has many applications in civil and geotechnical engineering. It is less expensive than hand-tied rebar, while still increasing the tensile strength many times. The shape, dimension, and length (standard 1 mm diameter, 45 mm length) of the fiber together with fiber volume amount and distribution are important parameters influencing the tensile strength of concrete.

Featured Image

Why is it important?

The paper demonstrates the need for high performance computing by focusing on one engineering application - investigation of fiber-reinforced concrete. The primary analysis is solving a microscale problem for homogenization within the range of linear material behavior. This basis problem can be modified (extended) in several directions and any of them substantially increases the computational demands. One extension, roughly described in this paper, is the solution of the inverse problem of identification of local material parameters or some level of deboning of the matrix and fibers. This problem is solved by an optimization procedure which requires repeated solution of the basic problem. The computational demands can increase about hundred times.

Perspectives