A Comparative Study of Constitutive and Neural Network Models for Flow Behavior of Mg-5.9Zn-1.6Zr-1.6Nd-0.9Y Alloy and Processing Maps

Abstract

Isothermal and constant strain rate compression tests of as-homogenized Mg-5.9Zn-1.6Zr-1.6Nd-0.9Y alloy were performed by using Gleeble-3800 thermo-mechanical simulator at temperature between 523 and 723 K and strain rate range from 0.001 to 1 s−1, respectively. The deformation behavior of the alloy at elevated temperature was investigated. A strain-dependent constitutive model based on the Arrhenius model and back-propagation neural network (BPNN) model of the flow stress was established. The predictability of the two models was evaluated by average absolute relative error (AARE) and correlation coefficient (R c), respectively. The results show that the BPNN model has more accurate predictability than strain-dependent constitutive model. Processing maps are obtained according to the principles of the dynamic materials modeling and microstructures of the compressed samples. The microstructure of the samples was observed by optical microscope and electron backscattered diffraction. Processing maps show that the instability domain is at a strain rate of range of 0.015-1 s−1 and a temperature between 523 and 623 K. The alloy has good workability at 630-700 K and 0.008-0.1 s−1, wherein dynamic recrystallization occurs.