Physically-based simulations of the cyclic behavior of FCC polycrystals

Two homogenization approaches are used in order predict the cyclic elastic-plastic behaviour of 316L(N) polycrystals, either self-consistent mean-field approaches such as the Kroner and the Hill-Hutchinson models or crystalline finite element computations carried out on large aggregates. Various shapes of grains have been used: cube, Voronoi polyhedron or even more realistic grain shapes. Either homogeneous grain size or measured grain size distributions have been taken into account. Finally, a 3D mesh containing 1370 grains has been built based on repeated EBSD measurements and re-polishing. This most realistic aggregate contains numerous twins. The used crystalline plasticity laws are based on the mechanisms of cyclic plasticity observed in FCC metals and alloys. The parameter adjustment uses single crystal data only. The macroscopic cyclic stress-strain curves predicted by the various models are generally close to experimental data. Whatever the used homogenization model, the distribution of the mean grain plastic strain becomes narrower with increasing macroscopic plastic strain amplitudes whereas the mean grain axial stress one becomes broader. These results are in a qualitative agreement with many experimental observations and measurements. Nevertheless, mean-field approaches neglect the effect of the neighbour grains and the scatter they predict remains about two times lower than the one predicted by crystalline finite element calculations using simple basic cubic grains. Additionally, taking into account very heterogeneous grain sizes and twins leads to slightly broader distributions. The effect of the homogenization procedure and microstructure on microcrack initiation is finally discussed. (authors)