On the selection of constitutive equation for modeling the friction stir processes of twin roll cast wrought AZ31B

Abstract

Proper numerical modeling of the Friction Stir Processes (FSPs) requires the identification of a suitable constitutive equation which accurately describes the stress-strain material behavior under an applicable range of strains, strain rates, and temperatures. While some such equations may be perfectly suitable to simulate processes characterized by low (or high) strains and temperatures, FSPs are widely recognized for their relatively moderate ranges of such state variables. In this work, a number of constitutive equations for describing flow stress in metals were screened for their suitability for modeling Friction Stir Processes of twin roll cast (TRC) wrought magnesium Mg–AL–Zn (AZ31B) alloy. Considered were 4 different reported variations of the popular Johnson–Cook equation and one Sellars–Tegart equation along with their literature–reported coefficients for fitting AZ31B stress–strain behavior. In addition, 6 variations of the (rarely used in FSPs simulations) Zerilli–Armstrong equation were also considered along with their literature–reported coefficients. The screening assessment was based on how well the considered constitutive equations fit experimental tensile stress–strain data of twin roll cast wrought AZ31B. Goodness of fit and residual sum of squares were the two statistical criteria utilized in the quantitative assessment whereas a ‘visual ’ measure was used as a qualitative measure. Initial screening resulted in the selection of one best fitting constitutive equation representing one of each of the Johnson–Cook, Sellars–Tegart, and Zerilli–Armstrong equations. An HCP–specific Zerilli–Armstrong constitutive equation (dubbed here as ZA6 ) was found to have the best quantitative and qualitative fit results with an R2 value of 0.967 compared to values of 0.934 and 0.826 for the Johnson–Cook and Sellars–Tegart constitutive equations, respectively. Additionally, a 3D thermo–mechanically coupled FEM model was built in DEFORM 3D to simulate the experimental tensile test from which the experimental load–deflection data was obtained. The three ‘finalist ’ equations were fed into the FEM simulations and were compared based on the 1) simulations’ running times and 2) goodness of fit of the simulation results to the experimental load–deflection data. It was found that the ZA6 constitutive equation exhibited favorable run times even when contrasted against the simpler mathematical form of the Sellars–Tegart equation. On average, the ZA6 equation showed improvements in solution time by 5.4% as compared with the Johnson–Cook equation and almost identical solution time (0.9% increase) with that of the ST equation. This result indicates that the proposed equation is not numerically expensive and can be safely adopted in such FEM simulations. Based on the favorable running times and goodness of fit, it was concluded that the HCP–specific Zerilli–Armstrong constitutive equation ZA6 holds an advantage over all other considered equations and was, therefore, selected as most suitable for the numerical modeling of FSP of twin roll cast AZ31B.