A computational study on efficient yield surface calibrations using a crystal plasticity spectral solver

Abstract

A computational framework is presented, capable of calculating virtual loads using the spectral solver in the DAMASK software for crystal plasticity simulations in desired stress directions. Calculations are used for the calibration of yield surfaces. The required spatial resolution is assessed based on a comparison with the previously published crystal plasticity finite-element method (CPFEM) and experimental results for three different aluminum alloys (AA1050, AA3103O, and AA3103H18) with 1000 and 2500 grains in a representative volume element. The results of the crystal plasticity fast Fourier transform (CPFFT) method agree well with CPFEM. The elongated grain morphology of the AA3103H18 alloy was found to have a small effect on predicted anisotropy. An analysis was made of how many tests are required for proper calibration of the Yld2004-18p orthotropic yield surface. It was found that 32 virtual tests, along either uniformly distributed strain rate or stress directions but obeying the orthotropic symmetry of the Yld2004-18p yield surface, make a good compromise between accuracy and computation time. Randomly chosen directions have a significantly larger error and require more virtual tests for a similarly good calibration of the yield surface. Since a preselected set of strain–rate directions does not require extra iterations, it is the preferred choice for the calibration of the full stress-based Yld2004-18p.